Upload folder using huggingface_hub

checkpoint-12208/config.json

@@ -0,0 +1,39 @@
+{
+  "_name_or_path": "meta-llama/Llama-3.1-8B-Instruct",
+  "architectures": [
+    "LlamaForCausalLM"
+  ],
+  "attention_bias": false,
+  "attention_dropout": 0.0,
+  "bos_token_id": 128000,
+  "eos_token_id": [
+    128001,
+    128008,
+    128009
+  ],
+  "hidden_act": "silu",
+  "hidden_size": 4096,
+  "initializer_range": 0.02,
+  "intermediate_size": 14336,
+  "max_position_embeddings": 131072,
+  "mlp_bias": false,
+  "model_type": "llama",
+  "num_attention_heads": 32,
+  "num_hidden_layers": 32,
+  "num_key_value_heads": 8,
+  "pretraining_tp": 1,
+  "rms_norm_eps": 1e-05,
+  "rope_scaling": {
+    "factor": 8.0,
+    "high_freq_factor": 4.0,
+    "low_freq_factor": 1.0,
+    "original_max_position_embeddings": 8192,
+    "rope_type": "llama3"
+  },
+  "rope_theta": 500000.0,
+  "tie_word_embeddings": false,
+  "torch_dtype": "float32",
+  "transformers_version": "4.44.0.dev0",
+  "use_cache": true,
+  "vocab_size": 128256
+}

checkpoint-12208/generation_config.json

@@ -0,0 +1,12 @@
+{
+  "bos_token_id": 128000,
+  "do_sample": true,
+  "eos_token_id": [
+    128001,
+    128008,
+    128009
+  ],
+  "temperature": 0.6,
+  "top_p": 0.9,
+  "transformers_version": "4.44.0.dev0"
+}

checkpoint-12208/model-00001-of-00007.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e883d4a61ef5df0ad9892e74d7dfd3be79d8f95ad2904612bff4a6ce8d2eefd0
+size 4886466168

checkpoint-12208/model-00002-of-00007.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:23fa5aaf8ac56c17c821980c9aee44fcbc2a617fc570e875e6fdb93d92886c72
+size 4832007448

checkpoint-12208/model-00003-of-00007.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1a223e87240e50dfb05634ea909cf1ec715d37a12b681b14a3a6e97f28688f2b
+size 4999813112

checkpoint-12208/model-00004-of-00007.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3047d199c611c6d12b1cca780f033ad89df9c076f742674f7ac4a892ca9573b3
+size 4999813128

checkpoint-12208/model-00005-of-00007.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0a2f91d749d664ecd22c71f3d54488966fc6899a03924b8fea3613c8f83398a8
+size 4832007496

checkpoint-12208/model-00006-of-00007.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9d532c5f7c4eac826ccdacf4eb79d6ddf528d667bbe818e32de351a81a3b9c57
+size 4999813120

checkpoint-12208/model-00007-of-00007.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:81b2e9b86a71241114f0e84b77acd6af49a3ec3180b37e495547f70d06e705ef
+size 2571158184

checkpoint-12208/model.safetensors.index.json

@@ -0,0 +1,298 @@
+{
+  "metadata": {
+    "total_size": 32121044992
+  },
+  "weight_map": {
+    "lm_head.weight": "model-00007-of-00007.safetensors",
+    "model.embed_tokens.weight": "model-00001-of-00007.safetensors",
+    "model.layers.0.input_layernorm.weight": "model-00001-of-00007.safetensors",
+    "model.layers.0.mlp.down_proj.weight": "model-00001-of-00007.safetensors",
+    "model.layers.0.mlp.gate_proj.weight": "model-00001-of-00007.safetensors",
+    "model.layers.0.mlp.up_proj.weight": "model-00001-of-00007.safetensors",
+    "model.layers.0.post_attention_layernorm.weight": "model-00001-of-00007.safetensors",
+    "model.layers.0.self_attn.k_proj.weight": "model-00001-of-00007.safetensors",
+    "model.layers.0.self_attn.o_proj.weight": "model-00001-of-00007.safetensors",
+    "model.layers.0.self_attn.q_proj.weight": "model-00001-of-00007.safetensors",
+    "model.layers.0.self_attn.v_proj.weight": "model-00001-of-00007.safetensors",
+    "model.layers.1.input_layernorm.weight": "model-00001-of-00007.safetensors",
+    "model.layers.1.mlp.down_proj.weight": "model-00001-of-00007.safetensors",
+    "model.layers.1.mlp.gate_proj.weight": "model-00001-of-00007.safetensors",
+    "model.layers.1.mlp.up_proj.weight": "model-00001-of-00007.safetensors",
+    "model.layers.1.post_attention_layernorm.weight": "model-00001-of-00007.safetensors",
+    "model.layers.1.self_attn.k_proj.weight": "model-00001-of-00007.safetensors",
+    "model.layers.1.self_attn.o_proj.weight": "model-00001-of-00007.safetensors",
+    "model.layers.1.self_attn.q_proj.weight": "model-00001-of-00007.safetensors",
+    "model.layers.1.self_attn.v_proj.weight": "model-00001-of-00007.safetensors",
+    "model.layers.10.input_layernorm.weight": "model-00003-of-00007.safetensors",
+    "model.layers.10.mlp.down_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.10.mlp.gate_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.10.mlp.up_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.10.post_attention_layernorm.weight": "model-00003-of-00007.safetensors",
+    "model.layers.10.self_attn.k_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.10.self_attn.o_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.10.self_attn.q_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.10.self_attn.v_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.11.input_layernorm.weight": "model-00003-of-00007.safetensors",
+    "model.layers.11.mlp.down_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.11.mlp.gate_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.11.mlp.up_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.11.post_attention_layernorm.weight": "model-00003-of-00007.safetensors",
+    "model.layers.11.self_attn.k_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.11.self_attn.o_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.11.self_attn.q_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.11.self_attn.v_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.12.input_layernorm.weight": "model-00003-of-00007.safetensors",
+    "model.layers.12.mlp.down_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.12.mlp.gate_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.12.mlp.up_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.12.post_attention_layernorm.weight": "model-00003-of-00007.safetensors",
+    "model.layers.12.self_attn.k_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.12.self_attn.o_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.12.self_attn.q_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.12.self_attn.v_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.13.input_layernorm.weight": "model-00003-of-00007.safetensors",
+    "model.layers.13.mlp.down_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.13.mlp.gate_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.13.mlp.up_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.13.post_attention_layernorm.weight": "model-00003-of-00007.safetensors",
+    "model.layers.13.self_attn.k_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.13.self_attn.o_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.13.self_attn.q_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.13.self_attn.v_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.14.input_layernorm.weight": "model-00004-of-00007.safetensors",
+    "model.layers.14.mlp.down_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.14.mlp.gate_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.14.mlp.up_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.14.post_attention_layernorm.weight": "model-00004-of-00007.safetensors",
+    "model.layers.14.self_attn.k_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.14.self_attn.o_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.14.self_attn.q_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.14.self_attn.v_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.15.input_layernorm.weight": "model-00004-of-00007.safetensors",
+    "model.layers.15.mlp.down_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.15.mlp.gate_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.15.mlp.up_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.15.post_attention_layernorm.weight": "model-00004-of-00007.safetensors",
+    "model.layers.15.self_attn.k_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.15.self_attn.o_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.15.self_attn.q_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.15.self_attn.v_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.16.input_layernorm.weight": "model-00004-of-00007.safetensors",
+    "model.layers.16.mlp.down_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.16.mlp.gate_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.16.mlp.up_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.16.post_attention_layernorm.weight": "model-00004-of-00007.safetensors",
+    "model.layers.16.self_attn.k_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.16.self_attn.o_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.16.self_attn.q_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.16.self_attn.v_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.17.input_layernorm.weight": "model-00004-of-00007.safetensors",
+    "model.layers.17.mlp.down_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.17.mlp.gate_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.17.mlp.up_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.17.post_attention_layernorm.weight": "model-00004-of-00007.safetensors",
+    "model.layers.17.self_attn.k_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.17.self_attn.o_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.17.self_attn.q_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.17.self_attn.v_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.18.input_layernorm.weight": "model-00004-of-00007.safetensors",
+    "model.layers.18.mlp.down_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.18.mlp.gate_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.18.mlp.up_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.18.post_attention_layernorm.weight": "model-00004-of-00007.safetensors",
+    "model.layers.18.self_attn.k_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.18.self_attn.o_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.18.self_attn.q_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.18.self_attn.v_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.19.input_layernorm.weight": "model-00004-of-00007.safetensors",
+    "model.layers.19.mlp.down_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.19.mlp.gate_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.19.mlp.up_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.19.post_attention_layernorm.weight": "model-00004-of-00007.safetensors",
+    "model.layers.19.self_attn.k_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.19.self_attn.o_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.19.self_attn.q_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.19.self_attn.v_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.2.input_layernorm.weight": "model-00001-of-00007.safetensors",
+    "model.layers.2.mlp.down_proj.weight": "model-00001-of-00007.safetensors",
+    "model.layers.2.mlp.gate_proj.weight": "model-00001-of-00007.safetensors",
+    "model.layers.2.mlp.up_proj.weight": "model-00001-of-00007.safetensors",
+    "model.layers.2.post_attention_layernorm.weight": "model-00001-of-00007.safetensors",
+    "model.layers.2.self_attn.k_proj.weight": "model-00001-of-00007.safetensors",
+    "model.layers.2.self_attn.o_proj.weight": "model-00001-of-00007.safetensors",
+    "model.layers.2.self_attn.q_proj.weight": "model-00001-of-00007.safetensors",
+    "model.layers.2.self_attn.v_proj.weight": "model-00001-of-00007.safetensors",
+    "model.layers.20.input_layernorm.weight": "model-00005-of-00007.safetensors",
+    "model.layers.20.mlp.down_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.20.mlp.gate_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.20.mlp.up_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.20.post_attention_layernorm.weight": "model-00005-of-00007.safetensors",
+    "model.layers.20.self_attn.k_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.20.self_attn.o_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.20.self_attn.q_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.20.self_attn.v_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.21.input_layernorm.weight": "model-00005-of-00007.safetensors",
+    "model.layers.21.mlp.down_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.21.mlp.gate_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.21.mlp.up_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.21.post_attention_layernorm.weight": "model-00005-of-00007.safetensors",
+    "model.layers.21.self_attn.k_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.21.self_attn.o_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.21.self_attn.q_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.21.self_attn.v_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.22.input_layernorm.weight": "model-00005-of-00007.safetensors",
+    "model.layers.22.mlp.down_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.22.mlp.gate_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.22.mlp.up_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.22.post_attention_layernorm.weight": "model-00005-of-00007.safetensors",
+    "model.layers.22.self_attn.k_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.22.self_attn.o_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.22.self_attn.q_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.22.self_attn.v_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.23.input_layernorm.weight": "model-00005-of-00007.safetensors",
+    "model.layers.23.mlp.down_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.23.mlp.gate_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.23.mlp.up_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.23.post_attention_layernorm.weight": "model-00005-of-00007.safetensors",
+    "model.layers.23.self_attn.k_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.23.self_attn.o_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.23.self_attn.q_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.23.self_attn.v_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.24.input_layernorm.weight": "model-00005-of-00007.safetensors",
+    "model.layers.24.mlp.down_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.24.mlp.gate_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.24.mlp.up_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.24.post_attention_layernorm.weight": "model-00005-of-00007.safetensors",
+    "model.layers.24.self_attn.k_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.24.self_attn.o_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.24.self_attn.q_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.24.self_attn.v_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.25.input_layernorm.weight": "model-00006-of-00007.safetensors",
+    "model.layers.25.mlp.down_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.25.mlp.gate_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.25.mlp.up_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.25.post_attention_layernorm.weight": "model-00006-of-00007.safetensors",
+    "model.layers.25.self_attn.k_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.25.self_attn.o_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.25.self_attn.q_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.25.self_attn.v_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.26.input_layernorm.weight": "model-00006-of-00007.safetensors",
+    "model.layers.26.mlp.down_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.26.mlp.gate_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.26.mlp.up_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.26.post_attention_layernorm.weight": "model-00006-of-00007.safetensors",
+    "model.layers.26.self_attn.k_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.26.self_attn.o_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.26.self_attn.q_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.26.self_attn.v_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.27.input_layernorm.weight": "model-00006-of-00007.safetensors",
+    "model.layers.27.mlp.down_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.27.mlp.gate_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.27.mlp.up_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.27.post_attention_layernorm.weight": "model-00006-of-00007.safetensors",
+    "model.layers.27.self_attn.k_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.27.self_attn.o_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.27.self_attn.q_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.27.self_attn.v_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.28.input_layernorm.weight": "model-00006-of-00007.safetensors",
+    "model.layers.28.mlp.down_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.28.mlp.gate_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.28.mlp.up_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.28.post_attention_layernorm.weight": "model-00006-of-00007.safetensors",
+    "model.layers.28.self_attn.k_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.28.self_attn.o_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.28.self_attn.q_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.28.self_attn.v_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.29.input_layernorm.weight": "model-00006-of-00007.safetensors",
+    "model.layers.29.mlp.down_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.29.mlp.gate_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.29.mlp.up_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.29.post_attention_layernorm.weight": "model-00006-of-00007.safetensors",
+    "model.layers.29.self_attn.k_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.29.self_attn.o_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.29.self_attn.q_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.29.self_attn.v_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.3.input_layernorm.weight": "model-00002-of-00007.safetensors",
+    "model.layers.3.mlp.down_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.3.mlp.gate_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.3.mlp.up_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.3.post_attention_layernorm.weight": "model-00002-of-00007.safetensors",
+    "model.layers.3.self_attn.k_proj.weight": "model-00001-of-00007.safetensors",
+    "model.layers.3.self_attn.o_proj.weight": "model-00001-of-00007.safetensors",
+    "model.layers.3.self_attn.q_proj.weight": "model-00001-of-00007.safetensors",
+    "model.layers.3.self_attn.v_proj.weight": "model-00001-of-00007.safetensors",
+    "model.layers.30.input_layernorm.weight": "model-00006-of-00007.safetensors",
+    "model.layers.30.mlp.down_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.30.mlp.gate_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.30.mlp.up_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.30.post_attention_layernorm.weight": "model-00006-of-00007.safetensors",
+    "model.layers.30.self_attn.k_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.30.self_attn.o_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.30.self_attn.q_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.30.self_attn.v_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.31.input_layernorm.weight": "model-00007-of-00007.safetensors",
+    "model.layers.31.mlp.down_proj.weight": "model-00007-of-00007.safetensors",
+    "model.layers.31.mlp.gate_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.31.mlp.up_proj.weight": "model-00007-of-00007.safetensors",
+    "model.layers.31.post_attention_layernorm.weight": "model-00007-of-00007.safetensors",
+    "model.layers.31.self_attn.k_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.31.self_attn.o_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.31.self_attn.q_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.31.self_attn.v_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.4.input_layernorm.weight": "model-00002-of-00007.safetensors",
+    "model.layers.4.mlp.down_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.4.mlp.gate_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.4.mlp.up_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.4.post_attention_layernorm.weight": "model-00002-of-00007.safetensors",
+    "model.layers.4.self_attn.k_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.4.self_attn.o_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.4.self_attn.q_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.4.self_attn.v_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.5.input_layernorm.weight": "model-00002-of-00007.safetensors",
+    "model.layers.5.mlp.down_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.5.mlp.gate_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.5.mlp.up_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.5.post_attention_layernorm.weight": "model-00002-of-00007.safetensors",
+    "model.layers.5.self_attn.k_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.5.self_attn.o_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.5.self_attn.q_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.5.self_attn.v_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.6.input_layernorm.weight": "model-00002-of-00007.safetensors",
+    "model.layers.6.mlp.down_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.6.mlp.gate_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.6.mlp.up_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.6.post_attention_layernorm.weight": "model-00002-of-00007.safetensors",
+    "model.layers.6.self_attn.k_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.6.self_attn.o_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.6.self_attn.q_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.6.self_attn.v_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.7.input_layernorm.weight": "model-00002-of-00007.safetensors",
+    "model.layers.7.mlp.down_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.7.mlp.gate_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.7.mlp.up_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.7.post_attention_layernorm.weight": "model-00002-of-00007.safetensors",
+    "model.layers.7.self_attn.k_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.7.self_attn.o_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.7.self_attn.q_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.7.self_attn.v_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.8.input_layernorm.weight": "model-00003-of-00007.safetensors",
+    "model.layers.8.mlp.down_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.8.mlp.gate_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.8.mlp.up_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.8.post_attention_layernorm.weight": "model-00003-of-00007.safetensors",
+    "model.layers.8.self_attn.k_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.8.self_attn.o_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.8.self_attn.q_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.8.self_attn.v_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.9.input_layernorm.weight": "model-00003-of-00007.safetensors",
+    "model.layers.9.mlp.down_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.9.mlp.gate_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.9.mlp.up_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.9.post_attention_layernorm.weight": "model-00003-of-00007.safetensors",
+    "model.layers.9.self_attn.k_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.9.self_attn.o_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.9.self_attn.q_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.9.self_attn.v_proj.weight": "model-00003-of-00007.safetensors",
+    "model.norm.weight": "model-00007-of-00007.safetensors"
+  }
+}

checkpoint-12208/optimizer.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a1fad81575b8221b28ab1ac09c6d2e7ec197e3f5cc5c34462361718fba288a68
+size 15385036334

checkpoint-12208/rng_state.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9d9cd6a0487226e5bd30d1846894c82af483733ab4381b75bae9c0745e05d405
+size 14244

checkpoint-12208/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:abe644ed33a3c4139223f0857a985127f3e6fbaa8c89fa14b57671b49ca52c21
+size 1064

checkpoint-12208/trainer_state.json

@@ -0,0 +1,2784 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.16267789563456106,
+  "eval_steps": 500,
+  "global_step": 12208,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.00041309098662118223,
+      "grad_norm": 3.6193394660949707,
+      "learning_rate": 1.0157273918741808e-06,
+      "loss": 0.7111,
+      "step": 31
+    },
+    {
+      "epoch": 0.0008261819732423645,
+      "grad_norm": 3.099701404571533,
+      "learning_rate": 2.0314547837483616e-06,
+      "loss": 0.6721,
+      "step": 62
+    },
+    {
+      "epoch": 0.0012392729598635467,
+      "grad_norm": 2.8040711879730225,
+      "learning_rate": 3.0471821756225426e-06,
+      "loss": 0.653,
+      "step": 93
+    },
+    {
+      "epoch": 0.001652363946484729,
+      "grad_norm": 2.99450945854187,
+      "learning_rate": 4.062909567496723e-06,
+      "loss": 0.647,
+      "step": 124
+    },
+    {
+      "epoch": 0.002065454933105911,
+      "grad_norm": 2.6550698280334473,
+      "learning_rate": 5.078636959370905e-06,
+      "loss": 0.6351,
+      "step": 155
+    },
+    {
+      "epoch": 0.0024785459197270935,
+      "grad_norm": 2.857128143310547,
+      "learning_rate": 6.094364351245085e-06,
+      "loss": 0.629,
+      "step": 186
+    },
+    {
+      "epoch": 0.002891636906348276,
+      "grad_norm": 2.6330196857452393,
+      "learning_rate": 7.110091743119267e-06,
+      "loss": 0.6349,
+      "step": 217
+    },
+    {
+      "epoch": 0.003304727892969458,
+      "grad_norm": 2.5196075439453125,
+      "learning_rate": 8.125819134993446e-06,
+      "loss": 0.6322,
+      "step": 248
+    },
+    {
+      "epoch": 0.0037178188795906402,
+      "grad_norm": 2.8110358715057373,
+      "learning_rate": 9.141546526867629e-06,
+      "loss": 0.6335,
+      "step": 279
+    },
+    {
+      "epoch": 0.004130909866211822,
+      "grad_norm": 2.4525060653686523,
+      "learning_rate": 1.015727391874181e-05,
+      "loss": 0.6385,
+      "step": 310
+    },
+    {
+      "epoch": 0.004544000852833005,
+      "grad_norm": 2.6030209064483643,
+      "learning_rate": 1.117300131061599e-05,
+      "loss": 0.6388,
+      "step": 341
+    },
+    {
+      "epoch": 0.004957091839454187,
+      "grad_norm": 2.976867914199829,
+      "learning_rate": 1.218872870249017e-05,
+      "loss": 0.632,
+      "step": 372
+    },
+    {
+      "epoch": 0.005370182826075369,
+      "grad_norm": 2.6810641288757324,
+      "learning_rate": 1.3204456094364351e-05,
+      "loss": 0.6304,
+      "step": 403
+    },
+    {
+      "epoch": 0.005783273812696552,
+      "grad_norm": 2.627316474914551,
+      "learning_rate": 1.4220183486238533e-05,
+      "loss": 0.6231,
+      "step": 434
+    },
+    {
+      "epoch": 0.006196364799317734,
+      "grad_norm": 3.0429155826568604,
+      "learning_rate": 1.5235910878112714e-05,
+      "loss": 0.6376,
+      "step": 465
+    },
+    {
+      "epoch": 0.006609455785938916,
+      "grad_norm": 2.3426945209503174,
+      "learning_rate": 1.6251638269986893e-05,
+      "loss": 0.6366,
+      "step": 496
+    },
+    {
+      "epoch": 0.007022546772560098,
+      "grad_norm": 2.640401840209961,
+      "learning_rate": 1.7267365661861077e-05,
+      "loss": 0.6375,
+      "step": 527
+    },
+    {
+      "epoch": 0.0074356377591812805,
+      "grad_norm": 2.8303098678588867,
+      "learning_rate": 1.8283093053735257e-05,
+      "loss": 0.6371,
+      "step": 558
+    },
+    {
+      "epoch": 0.007848728745802462,
+      "grad_norm": 2.2159361839294434,
+      "learning_rate": 1.9298820445609438e-05,
+      "loss": 0.6265,
+      "step": 589
+    },
+    {
+      "epoch": 0.008261819732423644,
+      "grad_norm": 2.60260272026062,
+      "learning_rate": 2.031454783748362e-05,
+      "loss": 0.6291,
+      "step": 620
+    },
+    {
+      "epoch": 0.008674910719044826,
+      "grad_norm": 2.245054006576538,
+      "learning_rate": 2.13302752293578e-05,
+      "loss": 0.6233,
+      "step": 651
+    },
+    {
+      "epoch": 0.00908800170566601,
+      "grad_norm": 2.550356864929199,
+      "learning_rate": 2.234600262123198e-05,
+      "loss": 0.6296,
+      "step": 682
+    },
+    {
+      "epoch": 0.009501092692287192,
+      "grad_norm": 2.1571903228759766,
+      "learning_rate": 2.336173001310616e-05,
+      "loss": 0.6239,
+      "step": 713
+    },
+    {
+      "epoch": 0.009914183678908374,
+      "grad_norm": 2.3740170001983643,
+      "learning_rate": 2.437745740498034e-05,
+      "loss": 0.6275,
+      "step": 744
+    },
+    {
+      "epoch": 0.010327274665529556,
+      "grad_norm": 2.0983521938323975,
+      "learning_rate": 2.5393184796854525e-05,
+      "loss": 0.6311,
+      "step": 775
+    },
+    {
+      "epoch": 0.010740365652150738,
+      "grad_norm": 2.818451404571533,
+      "learning_rate": 2.6408912188728702e-05,
+      "loss": 0.6383,
+      "step": 806
+    },
+    {
+      "epoch": 0.01115345663877192,
+      "grad_norm": 1.9579570293426514,
+      "learning_rate": 2.7424639580602886e-05,
+      "loss": 0.6297,
+      "step": 837
+    },
+    {
+      "epoch": 0.011566547625393104,
+      "grad_norm": 2.228142261505127,
+      "learning_rate": 2.8440366972477066e-05,
+      "loss": 0.6203,
+      "step": 868
+    },
+    {
+      "epoch": 0.011979638612014286,
+      "grad_norm": 2.132103681564331,
+      "learning_rate": 2.9456094364351244e-05,
+      "loss": 0.6248,
+      "step": 899
+    },
+    {
+      "epoch": 0.012392729598635467,
+      "grad_norm": 1.9388266801834106,
+      "learning_rate": 3.0471821756225428e-05,
+      "loss": 0.6271,
+      "step": 930
+    },
+    {
+      "epoch": 0.01280582058525665,
+      "grad_norm": 2.023773431777954,
+      "learning_rate": 3.148754914809961e-05,
+      "loss": 0.6291,
+      "step": 961
+    },
+    {
+      "epoch": 0.013218911571877831,
+      "grad_norm": 2.125810146331787,
+      "learning_rate": 3.2503276539973785e-05,
+      "loss": 0.6332,
+      "step": 992
+    },
+    {
+      "epoch": 0.013632002558499013,
+      "grad_norm": 2.299865484237671,
+      "learning_rate": 3.351900393184797e-05,
+      "loss": 0.6276,
+      "step": 1023
+    },
+    {
+      "epoch": 0.014045093545120195,
+      "grad_norm": 2.183851957321167,
+      "learning_rate": 3.453473132372215e-05,
+      "loss": 0.628,
+      "step": 1054
+    },
+    {
+      "epoch": 0.014458184531741379,
+      "grad_norm": 2.0226590633392334,
+      "learning_rate": 3.555045871559633e-05,
+      "loss": 0.6237,
+      "step": 1085
+    },
+    {
+      "epoch": 0.014871275518362561,
+      "grad_norm": 1.9166854619979858,
+      "learning_rate": 3.6566186107470514e-05,
+      "loss": 0.6282,
+      "step": 1116
+    },
+    {
+      "epoch": 0.015284366504983743,
+      "grad_norm": 1.8313781023025513,
+      "learning_rate": 3.7581913499344695e-05,
+      "loss": 0.6227,
+      "step": 1147
+    },
+    {
+      "epoch": 0.015697457491604925,
+      "grad_norm": 1.7466970682144165,
+      "learning_rate": 3.8597640891218876e-05,
+      "loss": 0.6219,
+      "step": 1178
+    },
+    {
+      "epoch": 0.016110548478226107,
+      "grad_norm": 1.744147777557373,
+      "learning_rate": 3.9613368283093056e-05,
+      "loss": 0.6264,
+      "step": 1209
+    },
+    {
+      "epoch": 0.01652363946484729,
+      "grad_norm": 2.1382932662963867,
+      "learning_rate": 4.062909567496724e-05,
+      "loss": 0.6186,
+      "step": 1240
+    },
+    {
+      "epoch": 0.01693673045146847,
+      "grad_norm": 1.9084594249725342,
+      "learning_rate": 4.164482306684142e-05,
+      "loss": 0.6295,
+      "step": 1271
+    },
+    {
+      "epoch": 0.017349821438089653,
+      "grad_norm": 1.9866138696670532,
+      "learning_rate": 4.26605504587156e-05,
+      "loss": 0.6284,
+      "step": 1302
+    },
+    {
+      "epoch": 0.017762912424710835,
+      "grad_norm": 1.7818348407745361,
+      "learning_rate": 4.367627785058978e-05,
+      "loss": 0.6217,
+      "step": 1333
+    },
+    {
+      "epoch": 0.01817600341133202,
+      "grad_norm": 1.9320199489593506,
+      "learning_rate": 4.469200524246396e-05,
+      "loss": 0.6263,
+      "step": 1364
+    },
+    {
+      "epoch": 0.018589094397953202,
+      "grad_norm": 1.9324169158935547,
+      "learning_rate": 4.570773263433814e-05,
+      "loss": 0.6185,
+      "step": 1395
+    },
+    {
+      "epoch": 0.019002185384574384,
+      "grad_norm": 1.887138843536377,
+      "learning_rate": 4.672346002621232e-05,
+      "loss": 0.6239,
+      "step": 1426
+    },
+    {
+      "epoch": 0.019415276371195566,
+      "grad_norm": 1.8720016479492188,
+      "learning_rate": 4.77391874180865e-05,
+      "loss": 0.6223,
+      "step": 1457
+    },
+    {
+      "epoch": 0.019828367357816748,
+      "grad_norm": 1.784279465675354,
+      "learning_rate": 4.875491480996068e-05,
+      "loss": 0.6267,
+      "step": 1488
+    },
+    {
+      "epoch": 0.02024145834443793,
+      "grad_norm": 2.594247817993164,
+      "learning_rate": 4.977064220183487e-05,
+      "loss": 0.6214,
+      "step": 1519
+    },
+    {
+      "epoch": 0.020654549331059112,
+      "grad_norm": 1.9205982685089111,
+      "learning_rate": 4.9999915451558777e-05,
+      "loss": 0.6167,
+      "step": 1550
+    },
+    {
+      "epoch": 0.021067640317680294,
+      "grad_norm": 1.6222068071365356,
+      "learning_rate": 4.999955597496219e-05,
+      "loss": 0.6197,
+      "step": 1581
+    },
+    {
+      "epoch": 0.021480731304301476,
+      "grad_norm": 1.5908927917480469,
+      "learning_rate": 4.9998914381774255e-05,
+      "loss": 0.6268,
+      "step": 1612
+    },
+    {
+      "epoch": 0.021893822290922658,
+      "grad_norm": 2.100959062576294,
+      "learning_rate": 4.999799067923527e-05,
+      "loss": 0.6224,
+      "step": 1643
+    },
+    {
+      "epoch": 0.02230691327754384,
+      "grad_norm": 1.6441237926483154,
+      "learning_rate": 4.999678487776908e-05,
+      "loss": 0.6189,
+      "step": 1674
+    },
+    {
+      "epoch": 0.02272000426416502,
+      "grad_norm": 1.680490255355835,
+      "learning_rate": 4.9995296990983006e-05,
+      "loss": 0.617,
+      "step": 1705
+    },
+    {
+      "epoch": 0.023133095250786207,
+      "grad_norm": 1.6946313381195068,
+      "learning_rate": 4.999352703566763e-05,
+      "loss": 0.6179,
+      "step": 1736
+    },
+    {
+      "epoch": 0.02354618623740739,
+      "grad_norm": 1.6758601665496826,
+      "learning_rate": 4.999147503179668e-05,
+      "loss": 0.6257,
+      "step": 1767
+    },
+    {
+      "epoch": 0.02395927722402857,
+      "grad_norm": 1.7444614171981812,
+      "learning_rate": 4.998914100252672e-05,
+      "loss": 0.6129,
+      "step": 1798
+    },
+    {
+      "epoch": 0.024372368210649753,
+      "grad_norm": 1.6522071361541748,
+      "learning_rate": 4.998652497419696e-05,
+      "loss": 0.6133,
+      "step": 1829
+    },
+    {
+      "epoch": 0.024785459197270935,
+      "grad_norm": 1.520391821861267,
+      "learning_rate": 4.9983626976328927e-05,
+      "loss": 0.6113,
+      "step": 1860
+    },
+    {
+      "epoch": 0.025198550183892117,
+      "grad_norm": 1.6573857069015503,
+      "learning_rate": 4.998044704162613e-05,
+      "loss": 0.6277,
+      "step": 1891
+    },
+    {
+      "epoch": 0.0256116411705133,
+      "grad_norm": 1.6087454557418823,
+      "learning_rate": 4.9976985205973705e-05,
+      "loss": 0.6121,
+      "step": 1922
+    },
+    {
+      "epoch": 0.02602473215713448,
+      "grad_norm": 1.545267939567566,
+      "learning_rate": 4.997324150843799e-05,
+      "loss": 0.5984,
+      "step": 1953
+    },
+    {
+      "epoch": 0.026437823143755663,
+      "grad_norm": 1.6027580499649048,
+      "learning_rate": 4.99692159912661e-05,
+      "loss": 0.6096,
+      "step": 1984
+    },
+    {
+      "epoch": 0.026850914130376845,
+      "grad_norm": 1.510338544845581,
+      "learning_rate": 4.996490869988546e-05,
+      "loss": 0.6019,
+      "step": 2015
+    },
+    {
+      "epoch": 0.027264005116998027,
+      "grad_norm": 1.6014549732208252,
+      "learning_rate": 4.996031968290326e-05,
+      "loss": 0.6049,
+      "step": 2046
+    },
+    {
+      "epoch": 0.02767709610361921,
+      "grad_norm": 1.8664464950561523,
+      "learning_rate": 4.995544899210594e-05,
+      "loss": 0.5994,
+      "step": 2077
+    },
+    {
+      "epoch": 0.02809018709024039,
+      "grad_norm": 1.6542178392410278,
+      "learning_rate": 4.9950296682458583e-05,
+      "loss": 0.6078,
+      "step": 2108
+    },
+    {
+      "epoch": 0.028503278076861576,
+      "grad_norm": 1.6745108366012573,
+      "learning_rate": 4.994486281210429e-05,
+      "loss": 0.6007,
+      "step": 2139
+    },
+    {
+      "epoch": 0.028916369063482758,
+      "grad_norm": 1.4461243152618408,
+      "learning_rate": 4.9939147442363566e-05,
+      "loss": 0.6057,
+      "step": 2170
+    },
+    {
+      "epoch": 0.02932946005010394,
+      "grad_norm": 1.455697774887085,
+      "learning_rate": 4.9933150637733574e-05,
+      "loss": 0.6015,
+      "step": 2201
+    },
+    {
+      "epoch": 0.029742551036725122,
+      "grad_norm": 1.550209641456604,
+      "learning_rate": 4.992687246588743e-05,
+      "loss": 0.5937,
+      "step": 2232
+    },
+    {
+      "epoch": 0.030155642023346304,
+      "grad_norm": 1.7507678270339966,
+      "learning_rate": 4.992031299767347e-05,
+      "loss": 0.601,
+      "step": 2263
+    },
+    {
+      "epoch": 0.030568733009967486,
+      "grad_norm": 1.5312328338623047,
+      "learning_rate": 4.9913472307114386e-05,
+      "loss": 0.6069,
+      "step": 2294
+    },
+    {
+      "epoch": 0.030981823996588668,
+      "grad_norm": 1.5791345834732056,
+      "learning_rate": 4.9906350471406446e-05,
+      "loss": 0.6011,
+      "step": 2325
+    },
+    {
+      "epoch": 0.03139491498320985,
+      "grad_norm": 1.6071161031723022,
+      "learning_rate": 4.989894757091861e-05,
+      "loss": 0.6043,
+      "step": 2356
+    },
+    {
+      "epoch": 0.031808005969831035,
+      "grad_norm": 1.6286126375198364,
+      "learning_rate": 4.989126368919158e-05,
+      "loss": 0.6074,
+      "step": 2387
+    },
+    {
+      "epoch": 0.032221096956452214,
+      "grad_norm": 1.3929970264434814,
+      "learning_rate": 4.988329891293693e-05,
+      "loss": 0.6037,
+      "step": 2418
+    },
+    {
+      "epoch": 0.0326341879430734,
+      "grad_norm": 1.426689624786377,
+      "learning_rate": 4.987505333203608e-05,
+      "loss": 0.6037,
+      "step": 2449
+    },
+    {
+      "epoch": 0.03304727892969458,
+      "grad_norm": 1.4528875350952148,
+      "learning_rate": 4.9866527039539276e-05,
+      "loss": 0.6027,
+      "step": 2480
+    },
+    {
+      "epoch": 0.03346036991631576,
+      "grad_norm": 1.359504222869873,
+      "learning_rate": 4.9857720131664594e-05,
+      "loss": 0.5901,
+      "step": 2511
+    },
+    {
+      "epoch": 0.03387346090293694,
+      "grad_norm": 1.5734764337539673,
+      "learning_rate": 4.9848632707796773e-05,
+      "loss": 0.5981,
+      "step": 2542
+    },
+    {
+      "epoch": 0.03428655188955813,
+      "grad_norm": 1.7405669689178467,
+      "learning_rate": 4.9839264870486155e-05,
+      "loss": 0.5973,
+      "step": 2573
+    },
+    {
+      "epoch": 0.034699642876179305,
+      "grad_norm": 1.5589804649353027,
+      "learning_rate": 4.9829616725447526e-05,
+      "loss": 0.6041,
+      "step": 2604
+    },
+    {
+      "epoch": 0.03511273386280049,
+      "grad_norm": 1.260091781616211,
+      "learning_rate": 4.981968838155888e-05,
+      "loss": 0.6009,
+      "step": 2635
+    },
+    {
+      "epoch": 0.03552582484942167,
+      "grad_norm": 1.5448715686798096,
+      "learning_rate": 4.980947995086024e-05,
+      "loss": 0.5945,
+      "step": 2666
+    },
+    {
+      "epoch": 0.035938915836042855,
+      "grad_norm": 1.8747330904006958,
+      "learning_rate": 4.979899154855234e-05,
+      "loss": 0.592,
+      "step": 2697
+    },
+    {
+      "epoch": 0.03635200682266404,
+      "grad_norm": 1.553969144821167,
+      "learning_rate": 4.9788223292995386e-05,
+      "loss": 0.5892,
+      "step": 2728
+    },
+    {
+      "epoch": 0.03676509780928522,
+      "grad_norm": 1.351070523262024,
+      "learning_rate": 4.977717530570768e-05,
+      "loss": 0.5964,
+      "step": 2759
+    },
+    {
+      "epoch": 0.037178188795906404,
+      "grad_norm": 1.3044447898864746,
+      "learning_rate": 4.976584771136425e-05,
+      "loss": 0.5948,
+      "step": 2790
+    },
+    {
+      "epoch": 0.03759127978252758,
+      "grad_norm": 1.4545929431915283,
+      "learning_rate": 4.975424063779547e-05,
+      "loss": 0.5873,
+      "step": 2821
+    },
+    {
+      "epoch": 0.03800437076914877,
+      "grad_norm": 1.4722667932510376,
+      "learning_rate": 4.974235421598557e-05,
+      "loss": 0.5886,
+      "step": 2852
+    },
+    {
+      "epoch": 0.03841746175576995,
+      "grad_norm": 1.3582379817962646,
+      "learning_rate": 4.973018858007122e-05,
+      "loss": 0.598,
+      "step": 2883
+    },
+    {
+      "epoch": 0.03883055274239113,
+      "grad_norm": 1.4251989126205444,
+      "learning_rate": 4.9717743867339963e-05,
+      "loss": 0.5839,
+      "step": 2914
+    },
+    {
+      "epoch": 0.03924364372901231,
+      "grad_norm": 1.4309897422790527,
+      "learning_rate": 4.9705020218228695e-05,
+      "loss": 0.5936,
+      "step": 2945
+    },
+    {
+      "epoch": 0.039656734715633496,
+      "grad_norm": 1.5348546504974365,
+      "learning_rate": 4.969201777632205e-05,
+      "loss": 0.5806,
+      "step": 2976
+    },
+    {
+      "epoch": 0.040069825702254674,
+      "grad_norm": 1.3221420049667358,
+      "learning_rate": 4.9678736688350846e-05,
+      "loss": 0.5863,
+      "step": 3007
+    },
+    {
+      "epoch": 0.04048291668887586,
+      "grad_norm": 1.302721619606018,
+      "learning_rate": 4.966517710419033e-05,
+      "loss": 0.5854,
+      "step": 3038
+    },
+    {
+      "epoch": 0.04089600767549704,
+      "grad_norm": 1.354509949684143,
+      "learning_rate": 4.965133917685858e-05,
+      "loss": 0.5896,
+      "step": 3069
+    },
+    {
+      "epoch": 0.041309098662118224,
+      "grad_norm": 1.2658582925796509,
+      "learning_rate": 4.9637223062514714e-05,
+      "loss": 0.5839,
+      "step": 3100
+    },
+    {
+      "epoch": 0.04172218964873941,
+      "grad_norm": 1.5425275564193726,
+      "learning_rate": 4.962282892045718e-05,
+      "loss": 0.5885,
+      "step": 3131
+    },
+    {
+      "epoch": 0.04213528063536059,
+      "grad_norm": 1.4991977214813232,
+      "learning_rate": 4.9608156913121904e-05,
+      "loss": 0.5888,
+      "step": 3162
+    },
+    {
+      "epoch": 0.04254837162198177,
+      "grad_norm": 1.6346509456634521,
+      "learning_rate": 4.959320720608049e-05,
+      "loss": 0.5911,
+      "step": 3193
+    },
+    {
+      "epoch": 0.04296146260860295,
+      "grad_norm": 1.368919849395752,
+      "learning_rate": 4.9577979968038354e-05,
+      "loss": 0.5877,
+      "step": 3224
+    },
+    {
+      "epoch": 0.04337455359522414,
+      "grad_norm": 1.324533224105835,
+      "learning_rate": 4.956247537083282e-05,
+      "loss": 0.5857,
+      "step": 3255
+    },
+    {
+      "epoch": 0.043787644581845316,
+      "grad_norm": 1.315446138381958,
+      "learning_rate": 4.9546693589431145e-05,
+      "loss": 0.5881,
+      "step": 3286
+    },
+    {
+      "epoch": 0.0442007355684665,
+      "grad_norm": 1.3033638000488281,
+      "learning_rate": 4.9530634801928595e-05,
+      "loss": 0.5826,
+      "step": 3317
+    },
+    {
+      "epoch": 0.04461382655508768,
+      "grad_norm": 1.4759029150009155,
+      "learning_rate": 4.9514299189546395e-05,
+      "loss": 0.5815,
+      "step": 3348
+    },
+    {
+      "epoch": 0.045026917541708865,
+      "grad_norm": 1.4898096323013306,
+      "learning_rate": 4.949768693662973e-05,
+      "loss": 0.5755,
+      "step": 3379
+    },
+    {
+      "epoch": 0.04544000852833004,
+      "grad_norm": 1.3367149829864502,
+      "learning_rate": 4.948079823064559e-05,
+      "loss": 0.5841,
+      "step": 3410
+    },
+    {
+      "epoch": 0.04585309951495123,
+      "grad_norm": 1.5252976417541504,
+      "learning_rate": 4.946363326218074e-05,
+      "loss": 0.5882,
+      "step": 3441
+    },
+    {
+      "epoch": 0.046266190501572414,
+      "grad_norm": 1.3493881225585938,
+      "learning_rate": 4.9446192224939525e-05,
+      "loss": 0.5896,
+      "step": 3472
+    },
+    {
+      "epoch": 0.04667928148819359,
+      "grad_norm": 1.275595784187317,
+      "learning_rate": 4.942847531574167e-05,
+      "loss": 0.5852,
+      "step": 3503
+    },
+    {
+      "epoch": 0.04709237247481478,
+      "grad_norm": 1.4550764560699463,
+      "learning_rate": 4.941048273452008e-05,
+      "loss": 0.5702,
+      "step": 3534
+    },
+    {
+      "epoch": 0.04750546346143596,
+      "grad_norm": 1.2401679754257202,
+      "learning_rate": 4.9392214684318605e-05,
+      "loss": 0.5841,
+      "step": 3565
+    },
+    {
+      "epoch": 0.04791855444805714,
+      "grad_norm": 1.3234368562698364,
+      "learning_rate": 4.93736713712897e-05,
+      "loss": 0.5751,
+      "step": 3596
+    },
+    {
+      "epoch": 0.04833164543467832,
+      "grad_norm": 1.4088999032974243,
+      "learning_rate": 4.9354853004692124e-05,
+      "loss": 0.5758,
+      "step": 3627
+    },
+    {
+      "epoch": 0.048744736421299506,
+      "grad_norm": 1.3294583559036255,
+      "learning_rate": 4.93357597968886e-05,
+      "loss": 0.5855,
+      "step": 3658
+    },
+    {
+      "epoch": 0.049157827407920685,
+      "grad_norm": 1.4270315170288086,
+      "learning_rate": 4.931639196334338e-05,
+      "loss": 0.5751,
+      "step": 3689
+    },
+    {
+      "epoch": 0.04957091839454187,
+      "grad_norm": 1.3925172090530396,
+      "learning_rate": 4.9296749722619826e-05,
+      "loss": 0.5724,
+      "step": 3720
+    },
+    {
+      "epoch": 0.04998400938116305,
+      "grad_norm": 1.2499849796295166,
+      "learning_rate": 4.9276833296377966e-05,
+      "loss": 0.5833,
+      "step": 3751
+    },
+    {
+      "epoch": 0.050397100367784234,
+      "grad_norm": 1.2049689292907715,
+      "learning_rate": 4.925664290937196e-05,
+      "loss": 0.5808,
+      "step": 3782
+    },
+    {
+      "epoch": 0.05081019135440541,
+      "grad_norm": 1.604710578918457,
+      "learning_rate": 4.9236178789447576e-05,
+      "loss": 0.5829,
+      "step": 3813
+    },
+    {
+      "epoch": 0.0512232823410266,
+      "grad_norm": 1.3707163333892822,
+      "learning_rate": 4.921544116753962e-05,
+      "loss": 0.5813,
+      "step": 3844
+    },
+    {
+      "epoch": 0.05163637332764778,
+      "grad_norm": 1.2728792428970337,
+      "learning_rate": 4.919443027766935e-05,
+      "loss": 0.5764,
+      "step": 3875
+    },
+    {
+      "epoch": 0.05204946431426896,
+      "grad_norm": 1.197830319404602,
+      "learning_rate": 4.91731463569418e-05,
+      "loss": 0.5713,
+      "step": 3906
+    },
+    {
+      "epoch": 0.05246255530089015,
+      "grad_norm": 1.3592743873596191,
+      "learning_rate": 4.915158964554312e-05,
+      "loss": 0.5877,
+      "step": 3937
+    },
+    {
+      "epoch": 0.052875646287511326,
+      "grad_norm": 1.2711323499679565,
+      "learning_rate": 4.912976038673786e-05,
+      "loss": 0.5737,
+      "step": 3968
+    },
+    {
+      "epoch": 0.05328873727413251,
+      "grad_norm": 1.2760255336761475,
+      "learning_rate": 4.9107658826866254e-05,
+      "loss": 0.5806,
+      "step": 3999
+    },
+    {
+      "epoch": 0.05370182826075369,
+      "grad_norm": 1.5267741680145264,
+      "learning_rate": 4.908528521534139e-05,
+      "loss": 0.5794,
+      "step": 4030
+    },
+    {
+      "epoch": 0.054114919247374875,
+      "grad_norm": 1.318003535270691,
+      "learning_rate": 4.906263980464644e-05,
+      "loss": 0.5683,
+      "step": 4061
+    },
+    {
+      "epoch": 0.05452801023399605,
+      "grad_norm": 1.4435062408447266,
+      "learning_rate": 4.903972285033178e-05,
+      "loss": 0.5772,
+      "step": 4092
+    },
+    {
+      "epoch": 0.05494110122061724,
+      "grad_norm": 1.4204825162887573,
+      "learning_rate": 4.901653461101213e-05,
+      "loss": 0.5794,
+      "step": 4123
+    },
+    {
+      "epoch": 0.05535419220723842,
+      "grad_norm": 1.222264289855957,
+      "learning_rate": 4.8993075348363626e-05,
+      "loss": 0.5681,
+      "step": 4154
+    },
+    {
+      "epoch": 0.0557672831938596,
+      "grad_norm": 1.3331806659698486,
+      "learning_rate": 4.896934532712084e-05,
+      "loss": 0.5761,
+      "step": 4185
+    },
+    {
+      "epoch": 0.05618037418048078,
+      "grad_norm": 1.2175304889678955,
+      "learning_rate": 4.8945344815073846e-05,
+      "loss": 0.5718,
+      "step": 4216
+    },
+    {
+      "epoch": 0.05659346516710197,
+      "grad_norm": 1.2743812799453735,
+      "learning_rate": 4.892107408306516e-05,
+      "loss": 0.5801,
+      "step": 4247
+    },
+    {
+      "epoch": 0.05700655615372315,
+      "grad_norm": 1.2461763620376587,
+      "learning_rate": 4.889653340498669e-05,
+      "loss": 0.5756,
+      "step": 4278
+    },
+    {
+      "epoch": 0.05741964714034433,
+      "grad_norm": 1.3406884670257568,
+      "learning_rate": 4.8871723057776664e-05,
+      "loss": 0.5787,
+      "step": 4309
+    },
+    {
+      "epoch": 0.057832738126965516,
+      "grad_norm": 1.3336981534957886,
+      "learning_rate": 4.8846643321416476e-05,
+      "loss": 0.568,
+      "step": 4340
+    },
+    {
+      "epoch": 0.058245829113586695,
+      "grad_norm": 1.194069743156433,
+      "learning_rate": 4.882129447892753e-05,
+      "loss": 0.5764,
+      "step": 4371
+    },
+    {
+      "epoch": 0.05865892010020788,
+      "grad_norm": 1.3039817810058594,
+      "learning_rate": 4.8795676816368076e-05,
+      "loss": 0.5712,
+      "step": 4402
+    },
+    {
+      "epoch": 0.05907201108682906,
+      "grad_norm": 1.2447880506515503,
+      "learning_rate": 4.876979062282995e-05,
+      "loss": 0.5722,
+      "step": 4433
+    },
+    {
+      "epoch": 0.059485102073450244,
+      "grad_norm": 1.3227744102478027,
+      "learning_rate": 4.8743636190435325e-05,
+      "loss": 0.5756,
+      "step": 4464
+    },
+    {
+      "epoch": 0.05989819306007142,
+      "grad_norm": 1.2006003856658936,
+      "learning_rate": 4.871721381433344e-05,
+      "loss": 0.5582,
+      "step": 4495
+    },
+    {
+      "epoch": 0.06031128404669261,
+      "grad_norm": 1.5150734186172485,
+      "learning_rate": 4.869052379269719e-05,
+      "loss": 0.5773,
+      "step": 4526
+    },
+    {
+      "epoch": 0.060724375033313786,
+      "grad_norm": 1.1430795192718506,
+      "learning_rate": 4.866356642671985e-05,
+      "loss": 0.5712,
+      "step": 4557
+    },
+    {
+      "epoch": 0.06113746601993497,
+      "grad_norm": 1.181968331336975,
+      "learning_rate": 4.8636342020611634e-05,
+      "loss": 0.5778,
+      "step": 4588
+    },
+    {
+      "epoch": 0.06155055700655616,
+      "grad_norm": 1.0638952255249023,
+      "learning_rate": 4.860885088159626e-05,
+      "loss": 0.5646,
+      "step": 4619
+    },
+    {
+      "epoch": 0.061963647993177336,
+      "grad_norm": 1.2407903671264648,
+      "learning_rate": 4.858109331990751e-05,
+      "loss": 0.5694,
+      "step": 4650
+    },
+    {
+      "epoch": 0.06237673897979852,
+      "grad_norm": 1.1231423616409302,
+      "learning_rate": 4.855306964878567e-05,
+      "loss": 0.5732,
+      "step": 4681
+    },
+    {
+      "epoch": 0.0627898299664197,
+      "grad_norm": 1.324874997138977,
+      "learning_rate": 4.8524780184474084e-05,
+      "loss": 0.5677,
+      "step": 4712
+    },
+    {
+      "epoch": 0.06320292095304088,
+      "grad_norm": 1.2522512674331665,
+      "learning_rate": 4.8496225246215496e-05,
+      "loss": 0.5692,
+      "step": 4743
+    },
+    {
+      "epoch": 0.06361601193966207,
+      "grad_norm": 1.3129041194915771,
+      "learning_rate": 4.8467405156248505e-05,
+      "loss": 0.5718,
+      "step": 4774
+    },
+    {
+      "epoch": 0.06402910292628325,
+      "grad_norm": 1.2068370580673218,
+      "learning_rate": 4.843832023980392e-05,
+      "loss": 0.5701,
+      "step": 4805
+    },
+    {
+      "epoch": 0.06444219391290443,
+      "grad_norm": 1.2644529342651367,
+      "learning_rate": 4.840897082510106e-05,
+      "loss": 0.5711,
+      "step": 4836
+    },
+    {
+      "epoch": 0.0648552848995256,
+      "grad_norm": 1.3398981094360352,
+      "learning_rate": 4.8379357243344084e-05,
+      "loss": 0.5588,
+      "step": 4867
+    },
+    {
+      "epoch": 0.0652683758861468,
+      "grad_norm": 1.208992600440979,
+      "learning_rate": 4.8349479828718236e-05,
+      "loss": 0.5634,
+      "step": 4898
+    },
+    {
+      "epoch": 0.06568146687276798,
+      "grad_norm": 1.1965893507003784,
+      "learning_rate": 4.8319338918386075e-05,
+      "loss": 0.5606,
+      "step": 4929
+    },
+    {
+      "epoch": 0.06609455785938916,
+      "grad_norm": 1.0229936838150024,
+      "learning_rate": 4.828893485248369e-05,
+      "loss": 0.5708,
+      "step": 4960
+    },
+    {
+      "epoch": 0.06650764884601033,
+      "grad_norm": 1.3772759437561035,
+      "learning_rate": 4.825826797411682e-05,
+      "loss": 0.5671,
+      "step": 4991
+    },
+    {
+      "epoch": 0.06692073983263153,
+      "grad_norm": 1.2997850179672241,
+      "learning_rate": 4.822733862935702e-05,
+      "loss": 0.5615,
+      "step": 5022
+    },
+    {
+      "epoch": 0.0673338308192527,
+      "grad_norm": 1.2622663974761963,
+      "learning_rate": 4.819614716723775e-05,
+      "loss": 0.5628,
+      "step": 5053
+    },
+    {
+      "epoch": 0.06774692180587388,
+      "grad_norm": 0.981414794921875,
+      "learning_rate": 4.8164693939750425e-05,
+      "loss": 0.5573,
+      "step": 5084
+    },
+    {
+      "epoch": 0.06816001279249508,
+      "grad_norm": 1.3478119373321533,
+      "learning_rate": 4.813297930184042e-05,
+      "loss": 0.5663,
+      "step": 5115
+    },
+    {
+      "epoch": 0.06857310377911625,
+      "grad_norm": 1.2769936323165894,
+      "learning_rate": 4.810100361140314e-05,
+      "loss": 0.5648,
+      "step": 5146
+    },
+    {
+      "epoch": 0.06898619476573743,
+      "grad_norm": 1.181077003479004,
+      "learning_rate": 4.8068767229279885e-05,
+      "loss": 0.5699,
+      "step": 5177
+    },
+    {
+      "epoch": 0.06939928575235861,
+      "grad_norm": 1.2607613801956177,
+      "learning_rate": 4.8036270519253854e-05,
+      "loss": 0.5609,
+      "step": 5208
+    },
+    {
+      "epoch": 0.0698123767389798,
+      "grad_norm": 1.242676854133606,
+      "learning_rate": 4.8003513848046e-05,
+      "loss": 0.5609,
+      "step": 5239
+    },
+    {
+      "epoch": 0.07022546772560098,
+      "grad_norm": 1.1886576414108276,
+      "learning_rate": 4.79704975853109e-05,
+      "loss": 0.5665,
+      "step": 5270
+    },
+    {
+      "epoch": 0.07063855871222216,
+      "grad_norm": 1.1143399477005005,
+      "learning_rate": 4.793722210363262e-05,
+      "loss": 0.5573,
+      "step": 5301
+    },
+    {
+      "epoch": 0.07105164969884334,
+      "grad_norm": 1.1033484935760498,
+      "learning_rate": 4.7903687778520414e-05,
+      "loss": 0.5566,
+      "step": 5332
+    },
+    {
+      "epoch": 0.07146474068546453,
+      "grad_norm": 1.2626802921295166,
+      "learning_rate": 4.7869894988404593e-05,
+      "loss": 0.5677,
+      "step": 5363
+    },
+    {
+      "epoch": 0.07187783167208571,
+      "grad_norm": 1.2453737258911133,
+      "learning_rate": 4.783584411463221e-05,
+      "loss": 0.5615,
+      "step": 5394
+    },
+    {
+      "epoch": 0.07229092265870689,
+      "grad_norm": 1.2548526525497437,
+      "learning_rate": 4.780153554146274e-05,
+      "loss": 0.5611,
+      "step": 5425
+    },
+    {
+      "epoch": 0.07270401364532808,
+      "grad_norm": 1.2009012699127197,
+      "learning_rate": 4.7766969656063766e-05,
+      "loss": 0.5648,
+      "step": 5456
+    },
+    {
+      "epoch": 0.07311710463194926,
+      "grad_norm": 1.1282159090042114,
+      "learning_rate": 4.773214684850662e-05,
+      "loss": 0.5584,
+      "step": 5487
+    },
+    {
+      "epoch": 0.07353019561857044,
+      "grad_norm": 1.1493905782699585,
+      "learning_rate": 4.769706751176193e-05,
+      "loss": 0.5655,
+      "step": 5518
+    },
+    {
+      "epoch": 0.07394328660519162,
+      "grad_norm": 1.1352925300598145,
+      "learning_rate": 4.7661732041695264e-05,
+      "loss": 0.5641,
+      "step": 5549
+    },
+    {
+      "epoch": 0.07435637759181281,
+      "grad_norm": 1.1498860120773315,
+      "learning_rate": 4.762614083706258e-05,
+      "loss": 0.5609,
+      "step": 5580
+    },
+    {
+      "epoch": 0.07476946857843399,
+      "grad_norm": 1.3136110305786133,
+      "learning_rate": 4.759029429950581e-05,
+      "loss": 0.5565,
+      "step": 5611
+    },
+    {
+      "epoch": 0.07518255956505517,
+      "grad_norm": 1.1240262985229492,
+      "learning_rate": 4.7554192833548235e-05,
+      "loss": 0.5605,
+      "step": 5642
+    },
+    {
+      "epoch": 0.07559565055167634,
+      "grad_norm": 1.4061548709869385,
+      "learning_rate": 4.751783684659e-05,
+      "loss": 0.5637,
+      "step": 5673
+    },
+    {
+      "epoch": 0.07600874153829754,
+      "grad_norm": 1.2112833261489868,
+      "learning_rate": 4.748122674890348e-05,
+      "loss": 0.564,
+      "step": 5704
+    },
+    {
+      "epoch": 0.07642183252491871,
+      "grad_norm": 1.1065545082092285,
+      "learning_rate": 4.7444362953628654e-05,
+      "loss": 0.553,
+      "step": 5735
+    },
+    {
+      "epoch": 0.0768349235115399,
+      "grad_norm": 1.3150484561920166,
+      "learning_rate": 4.7407245876768424e-05,
+      "loss": 0.56,
+      "step": 5766
+    },
+    {
+      "epoch": 0.07724801449816109,
+      "grad_norm": 1.0889962911605835,
+      "learning_rate": 4.736987593718397e-05,
+      "loss": 0.557,
+      "step": 5797
+    },
+    {
+      "epoch": 0.07766110548478226,
+      "grad_norm": 1.0983312129974365,
+      "learning_rate": 4.733225355658999e-05,
+      "loss": 0.5517,
+      "step": 5828
+    },
+    {
+      "epoch": 0.07807419647140344,
+      "grad_norm": 1.1129499673843384,
+      "learning_rate": 4.7294379159549926e-05,
+      "loss": 0.5572,
+      "step": 5859
+    },
+    {
+      "epoch": 0.07848728745802462,
+      "grad_norm": 1.1074233055114746,
+      "learning_rate": 4.725625317347119e-05,
+      "loss": 0.5548,
+      "step": 5890
+    },
+    {
+      "epoch": 0.07890037844464581,
+      "grad_norm": 1.3234531879425049,
+      "learning_rate": 4.7217876028600374e-05,
+      "loss": 0.5536,
+      "step": 5921
+    },
+    {
+      "epoch": 0.07931346943126699,
+      "grad_norm": 1.2597479820251465,
+      "learning_rate": 4.717924815801832e-05,
+      "loss": 0.5647,
+      "step": 5952
+    },
+    {
+      "epoch": 0.07972656041788817,
+      "grad_norm": 1.2188916206359863,
+      "learning_rate": 4.714036999763532e-05,
+      "loss": 0.5588,
+      "step": 5983
+    },
+    {
+      "epoch": 0.08013965140450935,
+      "grad_norm": 1.2447947263717651,
+      "learning_rate": 4.7101241986186116e-05,
+      "loss": 0.566,
+      "step": 6014
+    },
+    {
+      "epoch": 0.08055274239113054,
+      "grad_norm": 1.1489174365997314,
+      "learning_rate": 4.7061864565225e-05,
+      "loss": 0.5536,
+      "step": 6045
+    },
+    {
+      "epoch": 0.08096583337775172,
+      "grad_norm": 1.119452953338623,
+      "learning_rate": 4.702223817912081e-05,
+      "loss": 0.5496,
+      "step": 6076
+    },
+    {
+      "epoch": 0.0813789243643729,
+      "grad_norm": 1.109541416168213,
+      "learning_rate": 4.698236327505195e-05,
+      "loss": 0.5463,
+      "step": 6107
+    },
+    {
+      "epoch": 0.08179201535099408,
+      "grad_norm": 1.1216788291931152,
+      "learning_rate": 4.694224030300127e-05,
+      "loss": 0.5602,
+      "step": 6138
+    },
+    {
+      "epoch": 0.08220510633761527,
+      "grad_norm": 1.1995866298675537,
+      "learning_rate": 4.690186971575107e-05,
+      "loss": 0.5576,
+      "step": 6169
+    },
+    {
+      "epoch": 0.08261819732423645,
+      "grad_norm": 1.0980093479156494,
+      "learning_rate": 4.6861251968877916e-05,
+      "loss": 0.5546,
+      "step": 6200
+    },
+    {
+      "epoch": 0.08303128831085763,
+      "grad_norm": 1.0906014442443848,
+      "learning_rate": 4.68203875207476e-05,
+      "loss": 0.5523,
+      "step": 6231
+    },
+    {
+      "epoch": 0.08344437929747882,
+      "grad_norm": 1.1709624528884888,
+      "learning_rate": 4.677927683250983e-05,
+      "loss": 0.5475,
+      "step": 6262
+    },
+    {
+      "epoch": 0.0838574702841,
+      "grad_norm": 0.9864281415939331,
+      "learning_rate": 4.6737920368093156e-05,
+      "loss": 0.5671,
+      "step": 6293
+    },
+    {
+      "epoch": 0.08427056127072118,
+      "grad_norm": 1.1931488513946533,
+      "learning_rate": 4.669631859419965e-05,
+      "loss": 0.5607,
+      "step": 6324
+    },
+    {
+      "epoch": 0.08468365225734235,
+      "grad_norm": 1.1696871519088745,
+      "learning_rate": 4.6654471980299676e-05,
+      "loss": 0.5632,
+      "step": 6355
+    },
+    {
+      "epoch": 0.08509674324396355,
+      "grad_norm": 1.1524802446365356,
+      "learning_rate": 4.661238099862658e-05,
+      "loss": 0.5446,
+      "step": 6386
+    },
+    {
+      "epoch": 0.08550983423058472,
+      "grad_norm": 1.0331172943115234,
+      "learning_rate": 4.657004612417138e-05,
+      "loss": 0.5496,
+      "step": 6417
+    },
+    {
+      "epoch": 0.0859229252172059,
+      "grad_norm": 1.132674217224121,
+      "learning_rate": 4.6527467834677374e-05,
+      "loss": 0.5583,
+      "step": 6448
+    },
+    {
+      "epoch": 0.08633601620382708,
+      "grad_norm": 1.140329360961914,
+      "learning_rate": 4.648464661063478e-05,
+      "loss": 0.549,
+      "step": 6479
+    },
+    {
+      "epoch": 0.08674910719044827,
+      "grad_norm": 1.0936356782913208,
+      "learning_rate": 4.6441582935275264e-05,
+      "loss": 0.5504,
+      "step": 6510
+    },
+    {
+      "epoch": 0.08716219817706945,
+      "grad_norm": 1.0445035696029663,
+      "learning_rate": 4.6398277294566586e-05,
+      "loss": 0.5524,
+      "step": 6541
+    },
+    {
+      "epoch": 0.08757528916369063,
+      "grad_norm": 1.013694167137146,
+      "learning_rate": 4.6354730177207e-05,
+      "loss": 0.551,
+      "step": 6572
+    },
+    {
+      "epoch": 0.08798838015031182,
+      "grad_norm": 1.2154170274734497,
+      "learning_rate": 4.6310942074619787e-05,
+      "loss": 0.5563,
+      "step": 6603
+    },
+    {
+      "epoch": 0.088401471136933,
+      "grad_norm": 1.0196465253829956,
+      "learning_rate": 4.626691348094777e-05,
+      "loss": 0.5604,
+      "step": 6634
+    },
+    {
+      "epoch": 0.08881456212355418,
+      "grad_norm": 1.2061710357666016,
+      "learning_rate": 4.622264489304762e-05,
+      "loss": 0.5515,
+      "step": 6665
+    },
+    {
+      "epoch": 0.08922765311017536,
+      "grad_norm": 1.3378483057022095,
+      "learning_rate": 4.617813681048434e-05,
+      "loss": 0.5466,
+      "step": 6696
+    },
+    {
+      "epoch": 0.08964074409679655,
+      "grad_norm": 1.1906929016113281,
+      "learning_rate": 4.61333897355256e-05,
+      "loss": 0.5484,
+      "step": 6727
+    },
+    {
+      "epoch": 0.09005383508341773,
+      "grad_norm": 1.1115672588348389,
+      "learning_rate": 4.608840417313604e-05,
+      "loss": 0.5534,
+      "step": 6758
+    },
+    {
+      "epoch": 0.09046692607003891,
+      "grad_norm": 1.0681543350219727,
+      "learning_rate": 4.6043180630971646e-05,
+      "loss": 0.549,
+      "step": 6789
+    },
+    {
+      "epoch": 0.09088001705666009,
+      "grad_norm": 1.111663579940796,
+      "learning_rate": 4.599771961937391e-05,
+      "loss": 0.5529,
+      "step": 6820
+    },
+    {
+      "epoch": 0.09129310804328128,
+      "grad_norm": 1.111009120941162,
+      "learning_rate": 4.5952021651364204e-05,
+      "loss": 0.549,
+      "step": 6851
+    },
+    {
+      "epoch": 0.09170619902990246,
+      "grad_norm": 1.075073480606079,
+      "learning_rate": 4.590608724263786e-05,
+      "loss": 0.5502,
+      "step": 6882
+    },
+    {
+      "epoch": 0.09211929001652364,
+      "grad_norm": 1.2174952030181885,
+      "learning_rate": 4.585991691155845e-05,
+      "loss": 0.5499,
+      "step": 6913
+    },
+    {
+      "epoch": 0.09253238100314483,
+      "grad_norm": 1.0371384620666504,
+      "learning_rate": 4.581351117915188e-05,
+      "loss": 0.5481,
+      "step": 6944
+    },
+    {
+      "epoch": 0.09294547198976601,
+      "grad_norm": 1.1694340705871582,
+      "learning_rate": 4.5766870569100534e-05,
+      "loss": 0.5543,
+      "step": 6975
+    },
+    {
+      "epoch": 0.09335856297638719,
+      "grad_norm": 1.0203725099563599,
+      "learning_rate": 4.571999560773736e-05,
+      "loss": 0.5537,
+      "step": 7006
+    },
+    {
+      "epoch": 0.09377165396300836,
+      "grad_norm": 1.1278784275054932,
+      "learning_rate": 4.5672886824039915e-05,
+      "loss": 0.5451,
+      "step": 7037
+    },
+    {
+      "epoch": 0.09418474494962956,
+      "grad_norm": 1.2053276300430298,
+      "learning_rate": 4.5625544749624435e-05,
+      "loss": 0.5526,
+      "step": 7068
+    },
+    {
+      "epoch": 0.09459783593625073,
+      "grad_norm": 1.021291971206665,
+      "learning_rate": 4.5577969918739794e-05,
+      "loss": 0.5419,
+      "step": 7099
+    },
+    {
+      "epoch": 0.09501092692287191,
+      "grad_norm": 1.146509051322937,
+      "learning_rate": 4.5530162868261486e-05,
+      "loss": 0.5514,
+      "step": 7130
+    },
+    {
+      "epoch": 0.09542401790949309,
+      "grad_norm": 1.055192232131958,
+      "learning_rate": 4.548212413768558e-05,
+      "loss": 0.5572,
+      "step": 7161
+    },
+    {
+      "epoch": 0.09583710889611428,
+      "grad_norm": 1.0573087930679321,
+      "learning_rate": 4.543385426912261e-05,
+      "loss": 0.555,
+      "step": 7192
+    },
+    {
+      "epoch": 0.09625019988273546,
+      "grad_norm": 1.189032793045044,
+      "learning_rate": 4.53853538072915e-05,
+      "loss": 0.5421,
+      "step": 7223
+    },
+    {
+      "epoch": 0.09666329086935664,
+      "grad_norm": 1.0390995740890503,
+      "learning_rate": 4.533662329951336e-05,
+      "loss": 0.5442,
+      "step": 7254
+    },
+    {
+      "epoch": 0.09707638185597782,
+      "grad_norm": 1.0945408344268799,
+      "learning_rate": 4.528766329570536e-05,
+      "loss": 0.5529,
+      "step": 7285
+    },
+    {
+      "epoch": 0.09748947284259901,
+      "grad_norm": 1.2713547945022583,
+      "learning_rate": 4.523847434837447e-05,
+      "loss": 0.5452,
+      "step": 7316
+    },
+    {
+      "epoch": 0.09790256382922019,
+      "grad_norm": 1.123363733291626,
+      "learning_rate": 4.518905701261128e-05,
+      "loss": 0.5534,
+      "step": 7347
+    },
+    {
+      "epoch": 0.09831565481584137,
+      "grad_norm": 1.0784560441970825,
+      "learning_rate": 4.5139411846083715e-05,
+      "loss": 0.5471,
+      "step": 7378
+    },
+    {
+      "epoch": 0.09872874580246256,
+      "grad_norm": 1.1390719413757324,
+      "learning_rate": 4.508953940903073e-05,
+      "loss": 0.5526,
+      "step": 7409
+    },
+    {
+      "epoch": 0.09914183678908374,
+      "grad_norm": 0.9794678688049316,
+      "learning_rate": 4.5039440264255994e-05,
+      "loss": 0.542,
+      "step": 7440
+    },
+    {
+      "epoch": 0.09955492777570492,
+      "grad_norm": 1.1383626461029053,
+      "learning_rate": 4.498911497712155e-05,
+      "loss": 0.5396,
+      "step": 7471
+    },
+    {
+      "epoch": 0.0999680187623261,
+      "grad_norm": 1.0440157651901245,
+      "learning_rate": 4.493856411554142e-05,
+      "loss": 0.5461,
+      "step": 7502
+    },
+    {
+      "epoch": 0.10038110974894729,
+      "grad_norm": 1.1624081134796143,
+      "learning_rate": 4.4887788249975206e-05,
+      "loss": 0.5439,
+      "step": 7533
+    },
+    {
+      "epoch": 0.10079420073556847,
+      "grad_norm": 1.0614362955093384,
+      "learning_rate": 4.4836787953421656e-05,
+      "loss": 0.55,
+      "step": 7564
+    },
+    {
+      "epoch": 0.10120729172218965,
+      "grad_norm": 1.2120076417922974,
+      "learning_rate": 4.478556380141218e-05,
+      "loss": 0.5517,
+      "step": 7595
+    },
+    {
+      "epoch": 0.10162038270881082,
+      "grad_norm": 0.9873495697975159,
+      "learning_rate": 4.4734116372004375e-05,
+      "loss": 0.5436,
+      "step": 7626
+    },
+    {
+      "epoch": 0.10203347369543202,
+      "grad_norm": 0.9908550381660461,
+      "learning_rate": 4.4682446245775477e-05,
+      "loss": 0.5412,
+      "step": 7657
+    },
+    {
+      "epoch": 0.1024465646820532,
+      "grad_norm": 1.065071940422058,
+      "learning_rate": 4.463055400581586e-05,
+      "loss": 0.5433,
+      "step": 7688
+    },
+    {
+      "epoch": 0.10285965566867437,
+      "grad_norm": 1.1011327505111694,
+      "learning_rate": 4.4578440237722374e-05,
+      "loss": 0.5469,
+      "step": 7719
+    },
+    {
+      "epoch": 0.10327274665529557,
+      "grad_norm": 1.1491224765777588,
+      "learning_rate": 4.452610552959183e-05,
+      "loss": 0.545,
+      "step": 7750
+    },
+    {
+      "epoch": 0.10368583764191674,
+      "grad_norm": 1.0363606214523315,
+      "learning_rate": 4.447355047201428e-05,
+      "loss": 0.5458,
+      "step": 7781
+    },
+    {
+      "epoch": 0.10409892862853792,
+      "grad_norm": 1.137918472290039,
+      "learning_rate": 4.4420775658066414e-05,
+      "loss": 0.5358,
+      "step": 7812
+    },
+    {
+      "epoch": 0.1045120196151591,
+      "grad_norm": 1.1191500425338745,
+      "learning_rate": 4.436778168330484e-05,
+      "loss": 0.5414,
+      "step": 7843
+    },
+    {
+      "epoch": 0.1049251106017803,
+      "grad_norm": 1.10321044921875,
+      "learning_rate": 4.4314569145759353e-05,
+      "loss": 0.5439,
+      "step": 7874
+    },
+    {
+      "epoch": 0.10533820158840147,
+      "grad_norm": 1.0266538858413696,
+      "learning_rate": 4.42611386459262e-05,
+      "loss": 0.5418,
+      "step": 7905
+    },
+    {
+      "epoch": 0.10575129257502265,
+      "grad_norm": 1.050444483757019,
+      "learning_rate": 4.420749078676133e-05,
+      "loss": 0.5462,
+      "step": 7936
+    },
+    {
+      "epoch": 0.10616438356164383,
+      "grad_norm": 1.0405640602111816,
+      "learning_rate": 4.4153626173673516e-05,
+      "loss": 0.5464,
+      "step": 7967
+    },
+    {
+      "epoch": 0.10657747454826502,
+      "grad_norm": 1.1248273849487305,
+      "learning_rate": 4.409954541451762e-05,
+      "loss": 0.5436,
+      "step": 7998
+    },
+    {
+      "epoch": 0.1069905655348862,
+      "grad_norm": 1.1844524145126343,
+      "learning_rate": 4.404524911958764e-05,
+      "loss": 0.5506,
+      "step": 8029
+    },
+    {
+      "epoch": 0.10740365652150738,
+      "grad_norm": 0.9842180013656616,
+      "learning_rate": 4.399073790160989e-05,
+      "loss": 0.5408,
+      "step": 8060
+    },
+    {
+      "epoch": 0.10781674750812857,
+      "grad_norm": 1.0924715995788574,
+      "learning_rate": 4.393601237573607e-05,
+      "loss": 0.539,
+      "step": 8091
+    },
+    {
+      "epoch": 0.10822983849474975,
+      "grad_norm": 1.0117294788360596,
+      "learning_rate": 4.388107315953628e-05,
+      "loss": 0.5474,
+      "step": 8122
+    },
+    {
+      "epoch": 0.10864292948137093,
+      "grad_norm": 0.9410596489906311,
+      "learning_rate": 4.382592087299212e-05,
+      "loss": 0.5413,
+      "step": 8153
+    },
+    {
+      "epoch": 0.1090560204679921,
+      "grad_norm": 1.1258411407470703,
+      "learning_rate": 4.377055613848964e-05,
+      "loss": 0.5378,
+      "step": 8184
+    },
+    {
+      "epoch": 0.1094691114546133,
+      "grad_norm": 1.0934258699417114,
+      "learning_rate": 4.3714979580812355e-05,
+      "loss": 0.5429,
+      "step": 8215
+    },
+    {
+      "epoch": 0.10988220244123448,
+      "grad_norm": 1.0256693363189697,
+      "learning_rate": 4.365919182713416e-05,
+      "loss": 0.5521,
+      "step": 8246
+    },
+    {
+      "epoch": 0.11029529342785566,
+      "grad_norm": 1.1147851943969727,
+      "learning_rate": 4.360319350701226e-05,
+      "loss": 0.5367,
+      "step": 8277
+    },
+    {
+      "epoch": 0.11070838441447683,
+      "grad_norm": 1.065674066543579,
+      "learning_rate": 4.3546985252380115e-05,
+      "loss": 0.5399,
+      "step": 8308
+    },
+    {
+      "epoch": 0.11112147540109803,
+      "grad_norm": 1.096153736114502,
+      "learning_rate": 4.349056769754021e-05,
+      "loss": 0.5358,
+      "step": 8339
+    },
+    {
+      "epoch": 0.1115345663877192,
+      "grad_norm": 0.9637765884399414,
+      "learning_rate": 4.3433941479156994e-05,
+      "loss": 0.5411,
+      "step": 8370
+    },
+    {
+      "epoch": 0.11194765737434038,
+      "grad_norm": 1.0165079832077026,
+      "learning_rate": 4.3377107236249647e-05,
+      "loss": 0.5404,
+      "step": 8401
+    },
+    {
+      "epoch": 0.11236074836096156,
+      "grad_norm": 1.0558301210403442,
+      "learning_rate": 4.332006561018488e-05,
+      "loss": 0.5421,
+      "step": 8432
+    },
+    {
+      "epoch": 0.11277383934758275,
+      "grad_norm": 0.9278895854949951,
+      "learning_rate": 4.3262817244669683e-05,
+      "loss": 0.5309,
+      "step": 8463
+    },
+    {
+      "epoch": 0.11318693033420393,
+      "grad_norm": 1.0022321939468384,
+      "learning_rate": 4.3205362785744083e-05,
+      "loss": 0.5432,
+      "step": 8494
+    },
+    {
+      "epoch": 0.11360002132082511,
+      "grad_norm": 1.0881727933883667,
+      "learning_rate": 4.314770288177384e-05,
+      "loss": 0.5324,
+      "step": 8525
+    },
+    {
+      "epoch": 0.1140131123074463,
+      "grad_norm": 1.1254336833953857,
+      "learning_rate": 4.308983818344313e-05,
+      "loss": 0.5485,
+      "step": 8556
+    },
+    {
+      "epoch": 0.11442620329406748,
+      "grad_norm": 0.9436982274055481,
+      "learning_rate": 4.3031769343747206e-05,
+      "loss": 0.5402,
+      "step": 8587
+    },
+    {
+      "epoch": 0.11483929428068866,
+      "grad_norm": 1.0071033239364624,
+      "learning_rate": 4.297349701798505e-05,
+      "loss": 0.5493,
+      "step": 8618
+    },
+    {
+      "epoch": 0.11525238526730984,
+      "grad_norm": 1.1261751651763916,
+      "learning_rate": 4.2915021863751916e-05,
+      "loss": 0.5327,
+      "step": 8649
+    },
+    {
+      "epoch": 0.11566547625393103,
+      "grad_norm": 1.0863945484161377,
+      "learning_rate": 4.285634454093198e-05,
+      "loss": 0.5429,
+      "step": 8680
+    },
+    {
+      "epoch": 0.11607856724055221,
+      "grad_norm": 1.0068254470825195,
+      "learning_rate": 4.279746571169086e-05,
+      "loss": 0.5387,
+      "step": 8711
+    },
+    {
+      "epoch": 0.11649165822717339,
+      "grad_norm": 1.219416618347168,
+      "learning_rate": 4.2738386040468136e-05,
+      "loss": 0.536,
+      "step": 8742
+    },
+    {
+      "epoch": 0.11690474921379457,
+      "grad_norm": 1.0604240894317627,
+      "learning_rate": 4.2679106193969866e-05,
+      "loss": 0.5326,
+      "step": 8773
+    },
+    {
+      "epoch": 0.11731784020041576,
+      "grad_norm": 1.0701854228973389,
+      "learning_rate": 4.261962684116106e-05,
+      "loss": 0.5386,
+      "step": 8804
+    },
+    {
+      "epoch": 0.11773093118703694,
+      "grad_norm": 1.18992018699646,
+      "learning_rate": 4.2559948653258145e-05,
+      "loss": 0.5442,
+      "step": 8835
+    },
+    {
+      "epoch": 0.11814402217365812,
+      "grad_norm": 1.0788195133209229,
+      "learning_rate": 4.250007230372134e-05,
+      "loss": 0.5448,
+      "step": 8866
+    },
+    {
+      "epoch": 0.11855711316027931,
+      "grad_norm": 1.2016948461532593,
+      "learning_rate": 4.2439998468247126e-05,
+      "loss": 0.5407,
+      "step": 8897
+    },
+    {
+      "epoch": 0.11897020414690049,
+      "grad_norm": 0.9844837188720703,
+      "learning_rate": 4.2379727824760566e-05,
+      "loss": 0.541,
+      "step": 8928
+    },
+    {
+      "epoch": 0.11938329513352167,
+      "grad_norm": 0.9805975556373596,
+      "learning_rate": 4.231926105340768e-05,
+      "loss": 0.5437,
+      "step": 8959
+    },
+    {
+      "epoch": 0.11979638612014284,
+      "grad_norm": 0.9650508761405945,
+      "learning_rate": 4.225859883654776e-05,
+      "loss": 0.5234,
+      "step": 8990
+    },
+    {
+      "epoch": 0.12020947710676404,
+      "grad_norm": 1.111871600151062,
+      "learning_rate": 4.219774185874569e-05,
+      "loss": 0.5354,
+      "step": 9021
+    },
+    {
+      "epoch": 0.12062256809338522,
+      "grad_norm": 1.1532142162322998,
+      "learning_rate": 4.213669080676418e-05,
+      "loss": 0.5464,
+      "step": 9052
+    },
+    {
+      "epoch": 0.1210356590800064,
+      "grad_norm": 0.9762258529663086,
+      "learning_rate": 4.2075446369556056e-05,
+      "loss": 0.5416,
+      "step": 9083
+    },
+    {
+      "epoch": 0.12144875006662757,
+      "grad_norm": 1.082337498664856,
+      "learning_rate": 4.201400923825648e-05,
+      "loss": 0.5298,
+      "step": 9114
+    },
+    {
+      "epoch": 0.12186184105324877,
+      "grad_norm": 0.9814473390579224,
+      "learning_rate": 4.195238010617511e-05,
+      "loss": 0.5427,
+      "step": 9145
+    },
+    {
+      "epoch": 0.12227493203986994,
+      "grad_norm": 0.9980778694152832,
+      "learning_rate": 4.1890559668788344e-05,
+      "loss": 0.5408,
+      "step": 9176
+    },
+    {
+      "epoch": 0.12268802302649112,
+      "grad_norm": 0.9563232660293579,
+      "learning_rate": 4.1828548623731405e-05,
+      "loss": 0.5407,
+      "step": 9207
+    },
+    {
+      "epoch": 0.12310111401311231,
+      "grad_norm": 1.090119481086731,
+      "learning_rate": 4.1766347670790506e-05,
+      "loss": 0.5377,
+      "step": 9238
+    },
+    {
+      "epoch": 0.12351420499973349,
+      "grad_norm": 1.140044093132019,
+      "learning_rate": 4.170395751189495e-05,
+      "loss": 0.5369,
+      "step": 9269
+    },
+    {
+      "epoch": 0.12392729598635467,
+      "grad_norm": 1.0801435708999634,
+      "learning_rate": 4.164137885110921e-05,
+      "loss": 0.5339,
+      "step": 9300
+    },
+    {
+      "epoch": 0.12434038697297585,
+      "grad_norm": 1.0672699213027954,
+      "learning_rate": 4.157861239462495e-05,
+      "loss": 0.5384,
+      "step": 9331
+    },
+    {
+      "epoch": 0.12475347795959704,
+      "grad_norm": 1.1884500980377197,
+      "learning_rate": 4.1515658850753114e-05,
+      "loss": 0.5382,
+      "step": 9362
+    },
+    {
+      "epoch": 0.12516656894621822,
+      "grad_norm": 0.972347617149353,
+      "learning_rate": 4.145251892991588e-05,
+      "loss": 0.5404,
+      "step": 9393
+    },
+    {
+      "epoch": 0.1255796599328394,
+      "grad_norm": 1.0719892978668213,
+      "learning_rate": 4.138919334463868e-05,
+      "loss": 0.5336,
+      "step": 9424
+    },
+    {
+      "epoch": 0.12599275091946058,
+      "grad_norm": 1.1035733222961426,
+      "learning_rate": 4.1325682809542124e-05,
+      "loss": 0.5389,
+      "step": 9455
+    },
+    {
+      "epoch": 0.12640584190608176,
+      "grad_norm": 0.984730064868927,
+      "learning_rate": 4.126198804133398e-05,
+      "loss": 0.5431,
+      "step": 9486
+    },
+    {
+      "epoch": 0.12681893289270293,
+      "grad_norm": 0.9801311492919922,
+      "learning_rate": 4.1198109758801055e-05,
+      "loss": 0.5359,
+      "step": 9517
+    },
+    {
+      "epoch": 0.12723202387932414,
+      "grad_norm": 1.0692319869995117,
+      "learning_rate": 4.113404868280107e-05,
+      "loss": 0.5438,
+      "step": 9548
+    },
+    {
+      "epoch": 0.12764511486594532,
+      "grad_norm": 1.055517554283142,
+      "learning_rate": 4.106980553625457e-05,
+      "loss": 0.5386,
+      "step": 9579
+    },
+    {
+      "epoch": 0.1280582058525665,
+      "grad_norm": 1.0372021198272705,
+      "learning_rate": 4.100538104413674e-05,
+      "loss": 0.5304,
+      "step": 9610
+    },
+    {
+      "epoch": 0.12847129683918768,
+      "grad_norm": 0.9788863062858582,
+      "learning_rate": 4.09407759334692e-05,
+      "loss": 0.5336,
+      "step": 9641
+    },
+    {
+      "epoch": 0.12888438782580885,
+      "grad_norm": 0.9786235094070435,
+      "learning_rate": 4.087599093331186e-05,
+      "loss": 0.537,
+      "step": 9672
+    },
+    {
+      "epoch": 0.12929747881243003,
+      "grad_norm": 1.1998993158340454,
+      "learning_rate": 4.081102677475462e-05,
+      "loss": 0.5398,
+      "step": 9703
+    },
+    {
+      "epoch": 0.1297105697990512,
+      "grad_norm": 1.1321024894714355,
+      "learning_rate": 4.0745884190909194e-05,
+      "loss": 0.5388,
+      "step": 9734
+    },
+    {
+      "epoch": 0.13012366078567242,
+      "grad_norm": 1.0685521364212036,
+      "learning_rate": 4.0680563916900796e-05,
+      "loss": 0.5408,
+      "step": 9765
+    },
+    {
+      "epoch": 0.1305367517722936,
+      "grad_norm": 1.0620450973510742,
+      "learning_rate": 4.0615066689859815e-05,
+      "loss": 0.5315,
+      "step": 9796
+    },
+    {
+      "epoch": 0.13094984275891478,
+      "grad_norm": 0.9215843081474304,
+      "learning_rate": 4.0549393248913584e-05,
+      "loss": 0.5304,
+      "step": 9827
+    },
+    {
+      "epoch": 0.13136293374553595,
+      "grad_norm": 0.9986915588378906,
+      "learning_rate": 4.048354433517794e-05,
+      "loss": 0.5385,
+      "step": 9858
+    },
+    {
+      "epoch": 0.13177602473215713,
+      "grad_norm": 1.0409338474273682,
+      "learning_rate": 4.0417520691748916e-05,
+      "loss": 0.535,
+      "step": 9889
+    },
+    {
+      "epoch": 0.1321891157187783,
+      "grad_norm": 1.0327932834625244,
+      "learning_rate": 4.035132306369438e-05,
+      "loss": 0.5306,
+      "step": 9920
+    },
+    {
+      "epoch": 0.1326022067053995,
+      "grad_norm": 0.936696469783783,
+      "learning_rate": 4.028495219804555e-05,
+      "loss": 0.5246,
+      "step": 9951
+    },
+    {
+      "epoch": 0.13301529769202067,
+      "grad_norm": 0.9742713570594788,
+      "learning_rate": 4.021840884378864e-05,
+      "loss": 0.534,
+      "step": 9982
+    },
+    {
+      "epoch": 0.13342838867864187,
+      "grad_norm": 1.065794825553894,
+      "learning_rate": 4.015169375185633e-05,
+      "loss": 0.5332,
+      "step": 10013
+    },
+    {
+      "epoch": 0.13384147966526305,
+      "grad_norm": 0.9426642060279846,
+      "learning_rate": 4.0084807675119396e-05,
+      "loss": 0.5347,
+      "step": 10044
+    },
+    {
+      "epoch": 0.13425457065188423,
+      "grad_norm": 1.0314723253250122,
+      "learning_rate": 4.0017751368378106e-05,
+      "loss": 0.5248,
+      "step": 10075
+    },
+    {
+      "epoch": 0.1346676616385054,
+      "grad_norm": 1.100043773651123,
+      "learning_rate": 3.995052558835377e-05,
+      "loss": 0.5313,
+      "step": 10106
+    },
+    {
+      "epoch": 0.1350807526251266,
+      "grad_norm": 0.9944085478782654,
+      "learning_rate": 3.988313109368017e-05,
+      "loss": 0.5334,
+      "step": 10137
+    },
+    {
+      "epoch": 0.13549384361174777,
+      "grad_norm": 0.990739107131958,
+      "learning_rate": 3.981556864489504e-05,
+      "loss": 0.5333,
+      "step": 10168
+    },
+    {
+      "epoch": 0.13590693459836894,
+      "grad_norm": 0.9905965328216553,
+      "learning_rate": 3.974783900443142e-05,
+      "loss": 0.5335,
+      "step": 10199
+    },
+    {
+      "epoch": 0.13632002558499015,
+      "grad_norm": 1.0360978841781616,
+      "learning_rate": 3.9679942936609095e-05,
+      "loss": 0.5356,
+      "step": 10230
+    },
+    {
+      "epoch": 0.13673311657161133,
+      "grad_norm": 0.9169942140579224,
+      "learning_rate": 3.961188120762596e-05,
+      "loss": 0.5378,
+      "step": 10261
+    },
+    {
+      "epoch": 0.1371462075582325,
+      "grad_norm": 0.9882243275642395,
+      "learning_rate": 3.954365458554938e-05,
+      "loss": 0.5335,
+      "step": 10292
+    },
+    {
+      "epoch": 0.1375592985448537,
+      "grad_norm": 1.0667204856872559,
+      "learning_rate": 3.947526384030751e-05,
+      "loss": 0.5356,
+      "step": 10323
+    },
+    {
+      "epoch": 0.13797238953147486,
+      "grad_norm": 1.0076349973678589,
+      "learning_rate": 3.9406709743680624e-05,
+      "loss": 0.5394,
+      "step": 10354
+    },
+    {
+      "epoch": 0.13838548051809604,
+      "grad_norm": 0.9394388794898987,
+      "learning_rate": 3.9337993069292366e-05,
+      "loss": 0.5302,
+      "step": 10385
+    },
+    {
+      "epoch": 0.13879857150471722,
+      "grad_norm": 0.9442508220672607,
+      "learning_rate": 3.926911459260109e-05,
+      "loss": 0.5314,
+      "step": 10416
+    },
+    {
+      "epoch": 0.13921166249133843,
+      "grad_norm": 1.1007356643676758,
+      "learning_rate": 3.920007509089102e-05,
+      "loss": 0.5266,
+      "step": 10447
+    },
+    {
+      "epoch": 0.1396247534779596,
+      "grad_norm": 0.9338844418525696,
+      "learning_rate": 3.913087534326357e-05,
+      "loss": 0.5377,
+      "step": 10478
+    },
+    {
+      "epoch": 0.14003784446458079,
+      "grad_norm": 0.9138127565383911,
+      "learning_rate": 3.9061516130628475e-05,
+      "loss": 0.5336,
+      "step": 10509
+    },
+    {
+      "epoch": 0.14045093545120196,
+      "grad_norm": 1.0976738929748535,
+      "learning_rate": 3.8991998235695025e-05,
+      "loss": 0.5273,
+      "step": 10540
+    },
+    {
+      "epoch": 0.14086402643782314,
+      "grad_norm": 0.9631217122077942,
+      "learning_rate": 3.8922322442963224e-05,
+      "loss": 0.5258,
+      "step": 10571
+    },
+    {
+      "epoch": 0.14127711742444432,
+      "grad_norm": 1.0172392129898071,
+      "learning_rate": 3.885248953871491e-05,
+      "loss": 0.5354,
+      "step": 10602
+    },
+    {
+      "epoch": 0.1416902084110655,
+      "grad_norm": 1.0365219116210938,
+      "learning_rate": 3.8782500311004915e-05,
+      "loss": 0.5294,
+      "step": 10633
+    },
+    {
+      "epoch": 0.14210329939768668,
+      "grad_norm": 1.049512267112732,
+      "learning_rate": 3.871235554965218e-05,
+      "loss": 0.5298,
+      "step": 10664
+    },
+    {
+      "epoch": 0.14251639038430788,
+      "grad_norm": 0.9561368823051453,
+      "learning_rate": 3.864205604623078e-05,
+      "loss": 0.5278,
+      "step": 10695
+    },
+    {
+      "epoch": 0.14292948137092906,
+      "grad_norm": 0.9618587493896484,
+      "learning_rate": 3.857160259406107e-05,
+      "loss": 0.5264,
+      "step": 10726
+    },
+    {
+      "epoch": 0.14334257235755024,
+      "grad_norm": 0.953888475894928,
+      "learning_rate": 3.8500995988200674e-05,
+      "loss": 0.5305,
+      "step": 10757
+    },
+    {
+      "epoch": 0.14375566334417142,
+      "grad_norm": 1.0138510465621948,
+      "learning_rate": 3.843023702543556e-05,
+      "loss": 0.5319,
+      "step": 10788
+    },
+    {
+      "epoch": 0.1441687543307926,
+      "grad_norm": 0.9504718780517578,
+      "learning_rate": 3.8359326504270984e-05,
+      "loss": 0.521,
+      "step": 10819
+    },
+    {
+      "epoch": 0.14458184531741378,
+      "grad_norm": 0.8887687921524048,
+      "learning_rate": 3.828826522492255e-05,
+      "loss": 0.531,
+      "step": 10850
+    },
+    {
+      "epoch": 0.14499493630403495,
+      "grad_norm": 1.0202603340148926,
+      "learning_rate": 3.821705398930713e-05,
+      "loss": 0.5285,
+      "step": 10881
+    },
+    {
+      "epoch": 0.14540802729065616,
+      "grad_norm": 0.9483754634857178,
+      "learning_rate": 3.814569360103385e-05,
+      "loss": 0.5297,
+      "step": 10912
+    },
+    {
+      "epoch": 0.14582111827727734,
+      "grad_norm": 1.0187376737594604,
+      "learning_rate": 3.807418486539499e-05,
+      "loss": 0.5211,
+      "step": 10943
+    },
+    {
+      "epoch": 0.14623420926389852,
+      "grad_norm": 0.9178217053413391,
+      "learning_rate": 3.80025285893569e-05,
+      "loss": 0.5313,
+      "step": 10974
+    },
+    {
+      "epoch": 0.1466473002505197,
+      "grad_norm": 0.9808756709098816,
+      "learning_rate": 3.793072558155093e-05,
+      "loss": 0.5309,
+      "step": 11005
+    },
+    {
+      "epoch": 0.14706039123714088,
+      "grad_norm": 1.000494122505188,
+      "learning_rate": 3.785877665226426e-05,
+      "loss": 0.5264,
+      "step": 11036
+    },
+    {
+      "epoch": 0.14747348222376205,
+      "grad_norm": 0.9617404937744141,
+      "learning_rate": 3.778668261343079e-05,
+      "loss": 0.5173,
+      "step": 11067
+    },
+    {
+      "epoch": 0.14788657321038323,
+      "grad_norm": 0.9080610871315002,
+      "learning_rate": 3.771444427862192e-05,
+      "loss": 0.5365,
+      "step": 11098
+    },
+    {
+      "epoch": 0.1482996641970044,
+      "grad_norm": 0.9060406684875488,
+      "learning_rate": 3.7642062463037465e-05,
+      "loss": 0.5312,
+      "step": 11129
+    },
+    {
+      "epoch": 0.14871275518362562,
+      "grad_norm": 0.9788529276847839,
+      "learning_rate": 3.7569537983496373e-05,
+      "loss": 0.5307,
+      "step": 11160
+    },
+    {
+      "epoch": 0.1491258461702468,
+      "grad_norm": 1.0558335781097412,
+      "learning_rate": 3.749687165842753e-05,
+      "loss": 0.5322,
+      "step": 11191
+    },
+    {
+      "epoch": 0.14953893715686797,
+      "grad_norm": 0.9077445864677429,
+      "learning_rate": 3.7424064307860536e-05,
+      "loss": 0.5329,
+      "step": 11222
+    },
+    {
+      "epoch": 0.14995202814348915,
+      "grad_norm": 0.982660174369812,
+      "learning_rate": 3.735111675341645e-05,
+      "loss": 0.5272,
+      "step": 11253
+    },
+    {
+      "epoch": 0.15036511913011033,
+      "grad_norm": 0.9390408396720886,
+      "learning_rate": 3.7278029818298524e-05,
+      "loss": 0.5247,
+      "step": 11284
+    },
+    {
+      "epoch": 0.1507782101167315,
+      "grad_norm": 0.9156445860862732,
+      "learning_rate": 3.720480432728287e-05,
+      "loss": 0.5289,
+      "step": 11315
+    },
+    {
+      "epoch": 0.1511913011033527,
+      "grad_norm": 0.8932639956474304,
+      "learning_rate": 3.71314411067092e-05,
+      "loss": 0.5318,
+      "step": 11346
+    },
+    {
+      "epoch": 0.1516043920899739,
+      "grad_norm": 0.9631366729736328,
+      "learning_rate": 3.70579409844715e-05,
+      "loss": 0.5273,
+      "step": 11377
+    },
+    {
+      "epoch": 0.15201748307659507,
+      "grad_norm": 0.8670406341552734,
+      "learning_rate": 3.698430479000865e-05,
+      "loss": 0.5269,
+      "step": 11408
+    },
+    {
+      "epoch": 0.15243057406321625,
+      "grad_norm": 1.0833216905593872,
+      "learning_rate": 3.691053335429509e-05,
+      "loss": 0.5321,
+      "step": 11439
+    },
+    {
+      "epoch": 0.15284366504983743,
+      "grad_norm": 1.005530595779419,
+      "learning_rate": 3.683662750983147e-05,
+      "loss": 0.531,
+      "step": 11470
+    },
+    {
+      "epoch": 0.1532567560364586,
+      "grad_norm": 0.9262946844100952,
+      "learning_rate": 3.676258809063518e-05,
+      "loss": 0.5208,
+      "step": 11501
+    },
+    {
+      "epoch": 0.1536698470230798,
+      "grad_norm": 1.003950595855713,
+      "learning_rate": 3.6688415932231004e-05,
+      "loss": 0.5301,
+      "step": 11532
+    },
+    {
+      "epoch": 0.15408293800970096,
+      "grad_norm": 0.9618176221847534,
+      "learning_rate": 3.661411187164166e-05,
+      "loss": 0.5325,
+      "step": 11563
+    },
+    {
+      "epoch": 0.15449602899632217,
+      "grad_norm": 0.927218496799469,
+      "learning_rate": 3.65396767473784e-05,
+      "loss": 0.5256,
+      "step": 11594
+    },
+    {
+      "epoch": 0.15490911998294335,
+      "grad_norm": 2.2212576866149902,
+      "learning_rate": 3.6465111399431465e-05,
+      "loss": 0.5246,
+      "step": 11625
+    },
+    {
+      "epoch": 0.15532221096956453,
+      "grad_norm": 0.9497107863426208,
+      "learning_rate": 3.6390416669260674e-05,
+      "loss": 0.5265,
+      "step": 11656
+    },
+    {
+      "epoch": 0.1557353019561857,
+      "grad_norm": 0.9918030500411987,
+      "learning_rate": 3.63155933997859e-05,
+      "loss": 0.5259,
+      "step": 11687
+    },
+    {
+      "epoch": 0.15614839294280689,
+      "grad_norm": 0.8972867727279663,
+      "learning_rate": 3.624064243537758e-05,
+      "loss": 0.5212,
+      "step": 11718
+    },
+    {
+      "epoch": 0.15656148392942806,
+      "grad_norm": 0.9101853370666504,
+      "learning_rate": 3.616556462184716e-05,
+      "loss": 0.5312,
+      "step": 11749
+    },
+    {
+      "epoch": 0.15697457491604924,
+      "grad_norm": 0.9557723999023438,
+      "learning_rate": 3.609036080643755e-05,
+      "loss": 0.5205,
+      "step": 11780
+    },
+    {
+      "epoch": 0.15738766590267042,
+      "grad_norm": 0.9494668841362,
+      "learning_rate": 3.60150318378136e-05,
+      "loss": 0.5319,
+      "step": 11811
+    },
+    {
+      "epoch": 0.15780075688929163,
+      "grad_norm": 0.9334656596183777,
+      "learning_rate": 3.5939578566052465e-05,
+      "loss": 0.5195,
+      "step": 11842
+    },
+    {
+      "epoch": 0.1582138478759128,
+      "grad_norm": 0.8615605235099792,
+      "learning_rate": 3.586400184263408e-05,
+      "loss": 0.5188,
+      "step": 11873
+    },
+    {
+      "epoch": 0.15862693886253398,
+      "grad_norm": 0.9325541853904724,
+      "learning_rate": 3.578830252043148e-05,
+      "loss": 0.5176,
+      "step": 11904
+    },
+    {
+      "epoch": 0.15904002984915516,
+      "grad_norm": 0.9173383712768555,
+      "learning_rate": 3.571248145370125e-05,
+      "loss": 0.5259,
+      "step": 11935
+    },
+    {
+      "epoch": 0.15945312083577634,
+      "grad_norm": 1.119055151939392,
+      "learning_rate": 3.5636539498073794e-05,
+      "loss": 0.5333,
+      "step": 11966
+    },
+    {
+      "epoch": 0.15986621182239752,
+      "grad_norm": 0.9429402947425842,
+      "learning_rate": 3.556047751054378e-05,
+      "loss": 0.5251,
+      "step": 11997
+    },
+    {
+      "epoch": 0.1602793028090187,
+      "grad_norm": 0.9553139805793762,
+      "learning_rate": 3.548429634946039e-05,
+      "loss": 0.5237,
+      "step": 12028
+    },
+    {
+      "epoch": 0.1606923937956399,
+      "grad_norm": 1.031934380531311,
+      "learning_rate": 3.540799687451768e-05,
+      "loss": 0.5233,
+      "step": 12059
+    },
+    {
+      "epoch": 0.16110548478226108,
+      "grad_norm": 1.0917397737503052,
+      "learning_rate": 3.533157994674485e-05,
+      "loss": 0.523,
+      "step": 12090
+    },
+    {
+      "epoch": 0.16151857576888226,
+      "grad_norm": 1.0041927099227905,
+      "learning_rate": 3.5255046428496546e-05,
+      "loss": 0.5343,
+      "step": 12121
+    },
+    {
+      "epoch": 0.16193166675550344,
+      "grad_norm": 0.9607194066047668,
+      "learning_rate": 3.517839718344311e-05,
+      "loss": 0.5222,
+      "step": 12152
+    },
+    {
+      "epoch": 0.16234475774212462,
+      "grad_norm": 0.9220179915428162,
+      "learning_rate": 3.510163307656086e-05,
+      "loss": 0.5258,
+      "step": 12183
+    }
+  ],
+  "logging_steps": 31,
+  "max_steps": 30517,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 1,
+  "save_steps": 3052,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 9.006624315386364e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}

checkpoint-12208/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1fc4955674ee351f186535908caf566dac206e79a7c64f521967d1c769a6cc33
+size 5304

checkpoint-15260/config.json

@@ -0,0 +1,39 @@
+{
+  "_name_or_path": "meta-llama/Llama-3.1-8B-Instruct",
+  "architectures": [
+    "LlamaForCausalLM"
+  ],
+  "attention_bias": false,
+  "attention_dropout": 0.0,
+  "bos_token_id": 128000,
+  "eos_token_id": [
+    128001,
+    128008,
+    128009
+  ],
+  "hidden_act": "silu",
+  "hidden_size": 4096,
+  "initializer_range": 0.02,
+  "intermediate_size": 14336,
+  "max_position_embeddings": 131072,
+  "mlp_bias": false,
+  "model_type": "llama",
+  "num_attention_heads": 32,
+  "num_hidden_layers": 32,
+  "num_key_value_heads": 8,
+  "pretraining_tp": 1,
+  "rms_norm_eps": 1e-05,
+  "rope_scaling": {
+    "factor": 8.0,
+    "high_freq_factor": 4.0,
+    "low_freq_factor": 1.0,
+    "original_max_position_embeddings": 8192,
+    "rope_type": "llama3"
+  },
+  "rope_theta": 500000.0,
+  "tie_word_embeddings": false,
+  "torch_dtype": "float32",
+  "transformers_version": "4.44.0.dev0",
+  "use_cache": true,
+  "vocab_size": 128256
+}

checkpoint-15260/generation_config.json

@@ -0,0 +1,12 @@
+{
+  "bos_token_id": 128000,
+  "do_sample": true,
+  "eos_token_id": [
+    128001,
+    128008,
+    128009
+  ],
+  "temperature": 0.6,
+  "top_p": 0.9,
+  "transformers_version": "4.44.0.dev0"
+}

checkpoint-15260/model-00001-of-00007.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:83ce496f307d6e00e798870f5c8b22ad7db933460a9799f6c2d7784f7389ccee
+size 4886466168

checkpoint-15260/model-00002-of-00007.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:23fa5aaf8ac56c17c821980c9aee44fcbc2a617fc570e875e6fdb93d92886c72
+size 4832007448

checkpoint-15260/model-00003-of-00007.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1a223e87240e50dfb05634ea909cf1ec715d37a12b681b14a3a6e97f28688f2b
+size 4999813112

checkpoint-15260/model-00004-of-00007.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3047d199c611c6d12b1cca780f033ad89df9c076f742674f7ac4a892ca9573b3
+size 4999813128

checkpoint-15260/model-00005-of-00007.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0a2f91d749d664ecd22c71f3d54488966fc6899a03924b8fea3613c8f83398a8
+size 4832007496

checkpoint-15260/model-00006-of-00007.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:10bdeff7dcaf3aab7bf5274c91a09bd04a5a1b70f3b4a3c630b2e1001f4aef83
+size 4999813120

checkpoint-15260/model-00007-of-00007.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a9f360fb23dc537f1af6831c7f36dbe033869bff14da5553891691d1c25faf16
+size 2571158184

checkpoint-15260/model.safetensors.index.json

@@ -0,0 +1,298 @@
+{
+  "metadata": {
+    "total_size": 32121044992
+  },
+  "weight_map": {
+    "lm_head.weight": "model-00007-of-00007.safetensors",
+    "model.embed_tokens.weight": "model-00001-of-00007.safetensors",
+    "model.layers.0.input_layernorm.weight": "model-00001-of-00007.safetensors",
+    "model.layers.0.mlp.down_proj.weight": "model-00001-of-00007.safetensors",
+    "model.layers.0.mlp.gate_proj.weight": "model-00001-of-00007.safetensors",
+    "model.layers.0.mlp.up_proj.weight": "model-00001-of-00007.safetensors",
+    "model.layers.0.post_attention_layernorm.weight": "model-00001-of-00007.safetensors",
+    "model.layers.0.self_attn.k_proj.weight": "model-00001-of-00007.safetensors",
+    "model.layers.0.self_attn.o_proj.weight": "model-00001-of-00007.safetensors",
+    "model.layers.0.self_attn.q_proj.weight": "model-00001-of-00007.safetensors",
+    "model.layers.0.self_attn.v_proj.weight": "model-00001-of-00007.safetensors",
+    "model.layers.1.input_layernorm.weight": "model-00001-of-00007.safetensors",
+    "model.layers.1.mlp.down_proj.weight": "model-00001-of-00007.safetensors",
+    "model.layers.1.mlp.gate_proj.weight": "model-00001-of-00007.safetensors",
+    "model.layers.1.mlp.up_proj.weight": "model-00001-of-00007.safetensors",
+    "model.layers.1.post_attention_layernorm.weight": "model-00001-of-00007.safetensors",
+    "model.layers.1.self_attn.k_proj.weight": "model-00001-of-00007.safetensors",
+    "model.layers.1.self_attn.o_proj.weight": "model-00001-of-00007.safetensors",
+    "model.layers.1.self_attn.q_proj.weight": "model-00001-of-00007.safetensors",
+    "model.layers.1.self_attn.v_proj.weight": "model-00001-of-00007.safetensors",
+    "model.layers.10.input_layernorm.weight": "model-00003-of-00007.safetensors",
+    "model.layers.10.mlp.down_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.10.mlp.gate_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.10.mlp.up_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.10.post_attention_layernorm.weight": "model-00003-of-00007.safetensors",
+    "model.layers.10.self_attn.k_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.10.self_attn.o_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.10.self_attn.q_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.10.self_attn.v_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.11.input_layernorm.weight": "model-00003-of-00007.safetensors",
+    "model.layers.11.mlp.down_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.11.mlp.gate_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.11.mlp.up_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.11.post_attention_layernorm.weight": "model-00003-of-00007.safetensors",
+    "model.layers.11.self_attn.k_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.11.self_attn.o_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.11.self_attn.q_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.11.self_attn.v_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.12.input_layernorm.weight": "model-00003-of-00007.safetensors",
+    "model.layers.12.mlp.down_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.12.mlp.gate_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.12.mlp.up_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.12.post_attention_layernorm.weight": "model-00003-of-00007.safetensors",
+    "model.layers.12.self_attn.k_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.12.self_attn.o_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.12.self_attn.q_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.12.self_attn.v_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.13.input_layernorm.weight": "model-00003-of-00007.safetensors",
+    "model.layers.13.mlp.down_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.13.mlp.gate_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.13.mlp.up_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.13.post_attention_layernorm.weight": "model-00003-of-00007.safetensors",
+    "model.layers.13.self_attn.k_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.13.self_attn.o_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.13.self_attn.q_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.13.self_attn.v_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.14.input_layernorm.weight": "model-00004-of-00007.safetensors",
+    "model.layers.14.mlp.down_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.14.mlp.gate_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.14.mlp.up_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.14.post_attention_layernorm.weight": "model-00004-of-00007.safetensors",
+    "model.layers.14.self_attn.k_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.14.self_attn.o_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.14.self_attn.q_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.14.self_attn.v_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.15.input_layernorm.weight": "model-00004-of-00007.safetensors",
+    "model.layers.15.mlp.down_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.15.mlp.gate_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.15.mlp.up_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.15.post_attention_layernorm.weight": "model-00004-of-00007.safetensors",
+    "model.layers.15.self_attn.k_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.15.self_attn.o_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.15.self_attn.q_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.15.self_attn.v_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.16.input_layernorm.weight": "model-00004-of-00007.safetensors",
+    "model.layers.16.mlp.down_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.16.mlp.gate_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.16.mlp.up_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.16.post_attention_layernorm.weight": "model-00004-of-00007.safetensors",
+    "model.layers.16.self_attn.k_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.16.self_attn.o_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.16.self_attn.q_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.16.self_attn.v_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.17.input_layernorm.weight": "model-00004-of-00007.safetensors",
+    "model.layers.17.mlp.down_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.17.mlp.gate_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.17.mlp.up_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.17.post_attention_layernorm.weight": "model-00004-of-00007.safetensors",
+    "model.layers.17.self_attn.k_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.17.self_attn.o_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.17.self_attn.q_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.17.self_attn.v_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.18.input_layernorm.weight": "model-00004-of-00007.safetensors",
+    "model.layers.18.mlp.down_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.18.mlp.gate_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.18.mlp.up_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.18.post_attention_layernorm.weight": "model-00004-of-00007.safetensors",
+    "model.layers.18.self_attn.k_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.18.self_attn.o_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.18.self_attn.q_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.18.self_attn.v_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.19.input_layernorm.weight": "model-00004-of-00007.safetensors",
+    "model.layers.19.mlp.down_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.19.mlp.gate_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.19.mlp.up_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.19.post_attention_layernorm.weight": "model-00004-of-00007.safetensors",
+    "model.layers.19.self_attn.k_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.19.self_attn.o_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.19.self_attn.q_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.19.self_attn.v_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.2.input_layernorm.weight": "model-00001-of-00007.safetensors",
+    "model.layers.2.mlp.down_proj.weight": "model-00001-of-00007.safetensors",
+    "model.layers.2.mlp.gate_proj.weight": "model-00001-of-00007.safetensors",
+    "model.layers.2.mlp.up_proj.weight": "model-00001-of-00007.safetensors",
+    "model.layers.2.post_attention_layernorm.weight": "model-00001-of-00007.safetensors",
+    "model.layers.2.self_attn.k_proj.weight": "model-00001-of-00007.safetensors",
+    "model.layers.2.self_attn.o_proj.weight": "model-00001-of-00007.safetensors",
+    "model.layers.2.self_attn.q_proj.weight": "model-00001-of-00007.safetensors",
+    "model.layers.2.self_attn.v_proj.weight": "model-00001-of-00007.safetensors",
+    "model.layers.20.input_layernorm.weight": "model-00005-of-00007.safetensors",
+    "model.layers.20.mlp.down_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.20.mlp.gate_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.20.mlp.up_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.20.post_attention_layernorm.weight": "model-00005-of-00007.safetensors",
+    "model.layers.20.self_attn.k_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.20.self_attn.o_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.20.self_attn.q_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.20.self_attn.v_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.21.input_layernorm.weight": "model-00005-of-00007.safetensors",
+    "model.layers.21.mlp.down_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.21.mlp.gate_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.21.mlp.up_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.21.post_attention_layernorm.weight": "model-00005-of-00007.safetensors",
+    "model.layers.21.self_attn.k_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.21.self_attn.o_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.21.self_attn.q_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.21.self_attn.v_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.22.input_layernorm.weight": "model-00005-of-00007.safetensors",
+    "model.layers.22.mlp.down_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.22.mlp.gate_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.22.mlp.up_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.22.post_attention_layernorm.weight": "model-00005-of-00007.safetensors",
+    "model.layers.22.self_attn.k_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.22.self_attn.o_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.22.self_attn.q_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.22.self_attn.v_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.23.input_layernorm.weight": "model-00005-of-00007.safetensors",
+    "model.layers.23.mlp.down_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.23.mlp.gate_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.23.mlp.up_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.23.post_attention_layernorm.weight": "model-00005-of-00007.safetensors",
+    "model.layers.23.self_attn.k_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.23.self_attn.o_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.23.self_attn.q_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.23.self_attn.v_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.24.input_layernorm.weight": "model-00005-of-00007.safetensors",
+    "model.layers.24.mlp.down_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.24.mlp.gate_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.24.mlp.up_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.24.post_attention_layernorm.weight": "model-00005-of-00007.safetensors",
+    "model.layers.24.self_attn.k_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.24.self_attn.o_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.24.self_attn.q_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.24.self_attn.v_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.25.input_layernorm.weight": "model-00006-of-00007.safetensors",
+    "model.layers.25.mlp.down_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.25.mlp.gate_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.25.mlp.up_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.25.post_attention_layernorm.weight": "model-00006-of-00007.safetensors",
+    "model.layers.25.self_attn.k_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.25.self_attn.o_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.25.self_attn.q_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.25.self_attn.v_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.26.input_layernorm.weight": "model-00006-of-00007.safetensors",
+    "model.layers.26.mlp.down_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.26.mlp.gate_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.26.mlp.up_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.26.post_attention_layernorm.weight": "model-00006-of-00007.safetensors",
+    "model.layers.26.self_attn.k_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.26.self_attn.o_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.26.self_attn.q_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.26.self_attn.v_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.27.input_layernorm.weight": "model-00006-of-00007.safetensors",
+    "model.layers.27.mlp.down_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.27.mlp.gate_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.27.mlp.up_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.27.post_attention_layernorm.weight": "model-00006-of-00007.safetensors",
+    "model.layers.27.self_attn.k_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.27.self_attn.o_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.27.self_attn.q_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.27.self_attn.v_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.28.input_layernorm.weight": "model-00006-of-00007.safetensors",
+    "model.layers.28.mlp.down_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.28.mlp.gate_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.28.mlp.up_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.28.post_attention_layernorm.weight": "model-00006-of-00007.safetensors",
+    "model.layers.28.self_attn.k_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.28.self_attn.o_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.28.self_attn.q_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.28.self_attn.v_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.29.input_layernorm.weight": "model-00006-of-00007.safetensors",
+    "model.layers.29.mlp.down_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.29.mlp.gate_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.29.mlp.up_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.29.post_attention_layernorm.weight": "model-00006-of-00007.safetensors",
+    "model.layers.29.self_attn.k_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.29.self_attn.o_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.29.self_attn.q_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.29.self_attn.v_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.3.input_layernorm.weight": "model-00002-of-00007.safetensors",
+    "model.layers.3.mlp.down_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.3.mlp.gate_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.3.mlp.up_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.3.post_attention_layernorm.weight": "model-00002-of-00007.safetensors",
+    "model.layers.3.self_attn.k_proj.weight": "model-00001-of-00007.safetensors",
+    "model.layers.3.self_attn.o_proj.weight": "model-00001-of-00007.safetensors",
+    "model.layers.3.self_attn.q_proj.weight": "model-00001-of-00007.safetensors",
+    "model.layers.3.self_attn.v_proj.weight": "model-00001-of-00007.safetensors",
+    "model.layers.30.input_layernorm.weight": "model-00006-of-00007.safetensors",
+    "model.layers.30.mlp.down_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.30.mlp.gate_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.30.mlp.up_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.30.post_attention_layernorm.weight": "model-00006-of-00007.safetensors",
+    "model.layers.30.self_attn.k_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.30.self_attn.o_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.30.self_attn.q_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.30.self_attn.v_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.31.input_layernorm.weight": "model-00007-of-00007.safetensors",
+    "model.layers.31.mlp.down_proj.weight": "model-00007-of-00007.safetensors",
+    "model.layers.31.mlp.gate_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.31.mlp.up_proj.weight": "model-00007-of-00007.safetensors",
+    "model.layers.31.post_attention_layernorm.weight": "model-00007-of-00007.safetensors",
+    "model.layers.31.self_attn.k_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.31.self_attn.o_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.31.self_attn.q_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.31.self_attn.v_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.4.input_layernorm.weight": "model-00002-of-00007.safetensors",
+    "model.layers.4.mlp.down_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.4.mlp.gate_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.4.mlp.up_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.4.post_attention_layernorm.weight": "model-00002-of-00007.safetensors",
+    "model.layers.4.self_attn.k_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.4.self_attn.o_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.4.self_attn.q_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.4.self_attn.v_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.5.input_layernorm.weight": "model-00002-of-00007.safetensors",
+    "model.layers.5.mlp.down_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.5.mlp.gate_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.5.mlp.up_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.5.post_attention_layernorm.weight": "model-00002-of-00007.safetensors",
+    "model.layers.5.self_attn.k_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.5.self_attn.o_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.5.self_attn.q_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.5.self_attn.v_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.6.input_layernorm.weight": "model-00002-of-00007.safetensors",
+    "model.layers.6.mlp.down_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.6.mlp.gate_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.6.mlp.up_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.6.post_attention_layernorm.weight": "model-00002-of-00007.safetensors",
+    "model.layers.6.self_attn.k_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.6.self_attn.o_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.6.self_attn.q_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.6.self_attn.v_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.7.input_layernorm.weight": "model-00002-of-00007.safetensors",
+    "model.layers.7.mlp.down_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.7.mlp.gate_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.7.mlp.up_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.7.post_attention_layernorm.weight": "model-00002-of-00007.safetensors",
+    "model.layers.7.self_attn.k_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.7.self_attn.o_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.7.self_attn.q_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.7.self_attn.v_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.8.input_layernorm.weight": "model-00003-of-00007.safetensors",
+    "model.layers.8.mlp.down_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.8.mlp.gate_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.8.mlp.up_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.8.post_attention_layernorm.weight": "model-00003-of-00007.safetensors",
+    "model.layers.8.self_attn.k_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.8.self_attn.o_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.8.self_attn.q_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.8.self_attn.v_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.9.input_layernorm.weight": "model-00003-of-00007.safetensors",
+    "model.layers.9.mlp.down_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.9.mlp.gate_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.9.mlp.up_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.9.post_attention_layernorm.weight": "model-00003-of-00007.safetensors",
+    "model.layers.9.self_attn.k_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.9.self_attn.o_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.9.self_attn.q_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.9.self_attn.v_proj.weight": "model-00003-of-00007.safetensors",
+    "model.norm.weight": "model-00007-of-00007.safetensors"
+  }
+}

checkpoint-15260/optimizer.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3f770cd65c8e10cb6d65fedd63b2f54a8b7861be1eb493fa858968f707359ea2
+size 15385036334

checkpoint-15260/rng_state.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9d9cd6a0487226e5bd30d1846894c82af483733ab4381b75bae9c0745e05d405
+size 14244

checkpoint-15260/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2ce335347efaebf038b8915bc0a8d2d587a4a1aa08ad0e015b4bc7cc4fba634e
+size 1064

checkpoint-15260/trainer_state.json

@@ -0,0 +1,3477 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.20334736954320132,
+  "eval_steps": 500,
+  "global_step": 15260,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.00041309098662118223,
+      "grad_norm": 3.6193394660949707,
+      "learning_rate": 1.0157273918741808e-06,
+      "loss": 0.7111,
+      "step": 31
+    },
+    {
+      "epoch": 0.0008261819732423645,
+      "grad_norm": 3.099701404571533,
+      "learning_rate": 2.0314547837483616e-06,
+      "loss": 0.6721,
+      "step": 62
+    },
+    {
+      "epoch": 0.0012392729598635467,
+      "grad_norm": 2.8040711879730225,
+      "learning_rate": 3.0471821756225426e-06,
+      "loss": 0.653,
+      "step": 93
+    },
+    {
+      "epoch": 0.001652363946484729,
+      "grad_norm": 2.99450945854187,
+      "learning_rate": 4.062909567496723e-06,
+      "loss": 0.647,
+      "step": 124
+    },
+    {
+      "epoch": 0.002065454933105911,
+      "grad_norm": 2.6550698280334473,
+      "learning_rate": 5.078636959370905e-06,
+      "loss": 0.6351,
+      "step": 155
+    },
+    {
+      "epoch": 0.0024785459197270935,
+      "grad_norm": 2.857128143310547,
+      "learning_rate": 6.094364351245085e-06,
+      "loss": 0.629,
+      "step": 186
+    },
+    {
+      "epoch": 0.002891636906348276,
+      "grad_norm": 2.6330196857452393,
+      "learning_rate": 7.110091743119267e-06,
+      "loss": 0.6349,
+      "step": 217
+    },
+    {
+      "epoch": 0.003304727892969458,
+      "grad_norm": 2.5196075439453125,
+      "learning_rate": 8.125819134993446e-06,
+      "loss": 0.6322,
+      "step": 248
+    },
+    {
+      "epoch": 0.0037178188795906402,
+      "grad_norm": 2.8110358715057373,
+      "learning_rate": 9.141546526867629e-06,
+      "loss": 0.6335,
+      "step": 279
+    },
+    {
+      "epoch": 0.004130909866211822,
+      "grad_norm": 2.4525060653686523,
+      "learning_rate": 1.015727391874181e-05,
+      "loss": 0.6385,
+      "step": 310
+    },
+    {
+      "epoch": 0.004544000852833005,
+      "grad_norm": 2.6030209064483643,
+      "learning_rate": 1.117300131061599e-05,
+      "loss": 0.6388,
+      "step": 341
+    },
+    {
+      "epoch": 0.004957091839454187,
+      "grad_norm": 2.976867914199829,
+      "learning_rate": 1.218872870249017e-05,
+      "loss": 0.632,
+      "step": 372
+    },
+    {
+      "epoch": 0.005370182826075369,
+      "grad_norm": 2.6810641288757324,
+      "learning_rate": 1.3204456094364351e-05,
+      "loss": 0.6304,
+      "step": 403
+    },
+    {
+      "epoch": 0.005783273812696552,
+      "grad_norm": 2.627316474914551,
+      "learning_rate": 1.4220183486238533e-05,
+      "loss": 0.6231,
+      "step": 434
+    },
+    {
+      "epoch": 0.006196364799317734,
+      "grad_norm": 3.0429155826568604,
+      "learning_rate": 1.5235910878112714e-05,
+      "loss": 0.6376,
+      "step": 465
+    },
+    {
+      "epoch": 0.006609455785938916,
+      "grad_norm": 2.3426945209503174,
+      "learning_rate": 1.6251638269986893e-05,
+      "loss": 0.6366,
+      "step": 496
+    },
+    {
+      "epoch": 0.007022546772560098,
+      "grad_norm": 2.640401840209961,
+      "learning_rate": 1.7267365661861077e-05,
+      "loss": 0.6375,
+      "step": 527
+    },
+    {
+      "epoch": 0.0074356377591812805,
+      "grad_norm": 2.8303098678588867,
+      "learning_rate": 1.8283093053735257e-05,
+      "loss": 0.6371,
+      "step": 558
+    },
+    {
+      "epoch": 0.007848728745802462,
+      "grad_norm": 2.2159361839294434,
+      "learning_rate": 1.9298820445609438e-05,
+      "loss": 0.6265,
+      "step": 589
+    },
+    {
+      "epoch": 0.008261819732423644,
+      "grad_norm": 2.60260272026062,
+      "learning_rate": 2.031454783748362e-05,
+      "loss": 0.6291,
+      "step": 620
+    },
+    {
+      "epoch": 0.008674910719044826,
+      "grad_norm": 2.245054006576538,
+      "learning_rate": 2.13302752293578e-05,
+      "loss": 0.6233,
+      "step": 651
+    },
+    {
+      "epoch": 0.00908800170566601,
+      "grad_norm": 2.550356864929199,
+      "learning_rate": 2.234600262123198e-05,
+      "loss": 0.6296,
+      "step": 682
+    },
+    {
+      "epoch": 0.009501092692287192,
+      "grad_norm": 2.1571903228759766,
+      "learning_rate": 2.336173001310616e-05,
+      "loss": 0.6239,
+      "step": 713
+    },
+    {
+      "epoch": 0.009914183678908374,
+      "grad_norm": 2.3740170001983643,
+      "learning_rate": 2.437745740498034e-05,
+      "loss": 0.6275,
+      "step": 744
+    },
+    {
+      "epoch": 0.010327274665529556,
+      "grad_norm": 2.0983521938323975,
+      "learning_rate": 2.5393184796854525e-05,
+      "loss": 0.6311,
+      "step": 775
+    },
+    {
+      "epoch": 0.010740365652150738,
+      "grad_norm": 2.818451404571533,
+      "learning_rate": 2.6408912188728702e-05,
+      "loss": 0.6383,
+      "step": 806
+    },
+    {
+      "epoch": 0.01115345663877192,
+      "grad_norm": 1.9579570293426514,
+      "learning_rate": 2.7424639580602886e-05,
+      "loss": 0.6297,
+      "step": 837
+    },
+    {
+      "epoch": 0.011566547625393104,
+      "grad_norm": 2.228142261505127,
+      "learning_rate": 2.8440366972477066e-05,
+      "loss": 0.6203,
+      "step": 868
+    },
+    {
+      "epoch": 0.011979638612014286,
+      "grad_norm": 2.132103681564331,
+      "learning_rate": 2.9456094364351244e-05,
+      "loss": 0.6248,
+      "step": 899
+    },
+    {
+      "epoch": 0.012392729598635467,
+      "grad_norm": 1.9388266801834106,
+      "learning_rate": 3.0471821756225428e-05,
+      "loss": 0.6271,
+      "step": 930
+    },
+    {
+      "epoch": 0.01280582058525665,
+      "grad_norm": 2.023773431777954,
+      "learning_rate": 3.148754914809961e-05,
+      "loss": 0.6291,
+      "step": 961
+    },
+    {
+      "epoch": 0.013218911571877831,
+      "grad_norm": 2.125810146331787,
+      "learning_rate": 3.2503276539973785e-05,
+      "loss": 0.6332,
+      "step": 992
+    },
+    {
+      "epoch": 0.013632002558499013,
+      "grad_norm": 2.299865484237671,
+      "learning_rate": 3.351900393184797e-05,
+      "loss": 0.6276,
+      "step": 1023
+    },
+    {
+      "epoch": 0.014045093545120195,
+      "grad_norm": 2.183851957321167,
+      "learning_rate": 3.453473132372215e-05,
+      "loss": 0.628,
+      "step": 1054
+    },
+    {
+      "epoch": 0.014458184531741379,
+      "grad_norm": 2.0226590633392334,
+      "learning_rate": 3.555045871559633e-05,
+      "loss": 0.6237,
+      "step": 1085
+    },
+    {
+      "epoch": 0.014871275518362561,
+      "grad_norm": 1.9166854619979858,
+      "learning_rate": 3.6566186107470514e-05,
+      "loss": 0.6282,
+      "step": 1116
+    },
+    {
+      "epoch": 0.015284366504983743,
+      "grad_norm": 1.8313781023025513,
+      "learning_rate": 3.7581913499344695e-05,
+      "loss": 0.6227,
+      "step": 1147
+    },
+    {
+      "epoch": 0.015697457491604925,
+      "grad_norm": 1.7466970682144165,
+      "learning_rate": 3.8597640891218876e-05,
+      "loss": 0.6219,
+      "step": 1178
+    },
+    {
+      "epoch": 0.016110548478226107,
+      "grad_norm": 1.744147777557373,
+      "learning_rate": 3.9613368283093056e-05,
+      "loss": 0.6264,
+      "step": 1209
+    },
+    {
+      "epoch": 0.01652363946484729,
+      "grad_norm": 2.1382932662963867,
+      "learning_rate": 4.062909567496724e-05,
+      "loss": 0.6186,
+      "step": 1240
+    },
+    {
+      "epoch": 0.01693673045146847,
+      "grad_norm": 1.9084594249725342,
+      "learning_rate": 4.164482306684142e-05,
+      "loss": 0.6295,
+      "step": 1271
+    },
+    {
+      "epoch": 0.017349821438089653,
+      "grad_norm": 1.9866138696670532,
+      "learning_rate": 4.26605504587156e-05,
+      "loss": 0.6284,
+      "step": 1302
+    },
+    {
+      "epoch": 0.017762912424710835,
+      "grad_norm": 1.7818348407745361,
+      "learning_rate": 4.367627785058978e-05,
+      "loss": 0.6217,
+      "step": 1333
+    },
+    {
+      "epoch": 0.01817600341133202,
+      "grad_norm": 1.9320199489593506,
+      "learning_rate": 4.469200524246396e-05,
+      "loss": 0.6263,
+      "step": 1364
+    },
+    {
+      "epoch": 0.018589094397953202,
+      "grad_norm": 1.9324169158935547,
+      "learning_rate": 4.570773263433814e-05,
+      "loss": 0.6185,
+      "step": 1395
+    },
+    {
+      "epoch": 0.019002185384574384,
+      "grad_norm": 1.887138843536377,
+      "learning_rate": 4.672346002621232e-05,
+      "loss": 0.6239,
+      "step": 1426
+    },
+    {
+      "epoch": 0.019415276371195566,
+      "grad_norm": 1.8720016479492188,
+      "learning_rate": 4.77391874180865e-05,
+      "loss": 0.6223,
+      "step": 1457
+    },
+    {
+      "epoch": 0.019828367357816748,
+      "grad_norm": 1.784279465675354,
+      "learning_rate": 4.875491480996068e-05,
+      "loss": 0.6267,
+      "step": 1488
+    },
+    {
+      "epoch": 0.02024145834443793,
+      "grad_norm": 2.594247817993164,
+      "learning_rate": 4.977064220183487e-05,
+      "loss": 0.6214,
+      "step": 1519
+    },
+    {
+      "epoch": 0.020654549331059112,
+      "grad_norm": 1.9205982685089111,
+      "learning_rate": 4.9999915451558777e-05,
+      "loss": 0.6167,
+      "step": 1550
+    },
+    {
+      "epoch": 0.021067640317680294,
+      "grad_norm": 1.6222068071365356,
+      "learning_rate": 4.999955597496219e-05,
+      "loss": 0.6197,
+      "step": 1581
+    },
+    {
+      "epoch": 0.021480731304301476,
+      "grad_norm": 1.5908927917480469,
+      "learning_rate": 4.9998914381774255e-05,
+      "loss": 0.6268,
+      "step": 1612
+    },
+    {
+      "epoch": 0.021893822290922658,
+      "grad_norm": 2.100959062576294,
+      "learning_rate": 4.999799067923527e-05,
+      "loss": 0.6224,
+      "step": 1643
+    },
+    {
+      "epoch": 0.02230691327754384,
+      "grad_norm": 1.6441237926483154,
+      "learning_rate": 4.999678487776908e-05,
+      "loss": 0.6189,
+      "step": 1674
+    },
+    {
+      "epoch": 0.02272000426416502,
+      "grad_norm": 1.680490255355835,
+      "learning_rate": 4.9995296990983006e-05,
+      "loss": 0.617,
+      "step": 1705
+    },
+    {
+      "epoch": 0.023133095250786207,
+      "grad_norm": 1.6946313381195068,
+      "learning_rate": 4.999352703566763e-05,
+      "loss": 0.6179,
+      "step": 1736
+    },
+    {
+      "epoch": 0.02354618623740739,
+      "grad_norm": 1.6758601665496826,
+      "learning_rate": 4.999147503179668e-05,
+      "loss": 0.6257,
+      "step": 1767
+    },
+    {
+      "epoch": 0.02395927722402857,
+      "grad_norm": 1.7444614171981812,
+      "learning_rate": 4.998914100252672e-05,
+      "loss": 0.6129,
+      "step": 1798
+    },
+    {
+      "epoch": 0.024372368210649753,
+      "grad_norm": 1.6522071361541748,
+      "learning_rate": 4.998652497419696e-05,
+      "loss": 0.6133,
+      "step": 1829
+    },
+    {
+      "epoch": 0.024785459197270935,
+      "grad_norm": 1.520391821861267,
+      "learning_rate": 4.9983626976328927e-05,
+      "loss": 0.6113,
+      "step": 1860
+    },
+    {
+      "epoch": 0.025198550183892117,
+      "grad_norm": 1.6573857069015503,
+      "learning_rate": 4.998044704162613e-05,
+      "loss": 0.6277,
+      "step": 1891
+    },
+    {
+      "epoch": 0.0256116411705133,
+      "grad_norm": 1.6087454557418823,
+      "learning_rate": 4.9976985205973705e-05,
+      "loss": 0.6121,
+      "step": 1922
+    },
+    {
+      "epoch": 0.02602473215713448,
+      "grad_norm": 1.545267939567566,
+      "learning_rate": 4.997324150843799e-05,
+      "loss": 0.5984,
+      "step": 1953
+    },
+    {
+      "epoch": 0.026437823143755663,
+      "grad_norm": 1.6027580499649048,
+      "learning_rate": 4.99692159912661e-05,
+      "loss": 0.6096,
+      "step": 1984
+    },
+    {
+      "epoch": 0.026850914130376845,
+      "grad_norm": 1.510338544845581,
+      "learning_rate": 4.996490869988546e-05,
+      "loss": 0.6019,
+      "step": 2015
+    },
+    {
+      "epoch": 0.027264005116998027,
+      "grad_norm": 1.6014549732208252,
+      "learning_rate": 4.996031968290326e-05,
+      "loss": 0.6049,
+      "step": 2046
+    },
+    {
+      "epoch": 0.02767709610361921,
+      "grad_norm": 1.8664464950561523,
+      "learning_rate": 4.995544899210594e-05,
+      "loss": 0.5994,
+      "step": 2077
+    },
+    {
+      "epoch": 0.02809018709024039,
+      "grad_norm": 1.6542178392410278,
+      "learning_rate": 4.9950296682458583e-05,
+      "loss": 0.6078,
+      "step": 2108
+    },
+    {
+      "epoch": 0.028503278076861576,
+      "grad_norm": 1.6745108366012573,
+      "learning_rate": 4.994486281210429e-05,
+      "loss": 0.6007,
+      "step": 2139
+    },
+    {
+      "epoch": 0.028916369063482758,
+      "grad_norm": 1.4461243152618408,
+      "learning_rate": 4.9939147442363566e-05,
+      "loss": 0.6057,
+      "step": 2170
+    },
+    {
+      "epoch": 0.02932946005010394,
+      "grad_norm": 1.455697774887085,
+      "learning_rate": 4.9933150637733574e-05,
+      "loss": 0.6015,
+      "step": 2201
+    },
+    {
+      "epoch": 0.029742551036725122,
+      "grad_norm": 1.550209641456604,
+      "learning_rate": 4.992687246588743e-05,
+      "loss": 0.5937,
+      "step": 2232
+    },
+    {
+      "epoch": 0.030155642023346304,
+      "grad_norm": 1.7507678270339966,
+      "learning_rate": 4.992031299767347e-05,
+      "loss": 0.601,
+      "step": 2263
+    },
+    {
+      "epoch": 0.030568733009967486,
+      "grad_norm": 1.5312328338623047,
+      "learning_rate": 4.9913472307114386e-05,
+      "loss": 0.6069,
+      "step": 2294
+    },
+    {
+      "epoch": 0.030981823996588668,
+      "grad_norm": 1.5791345834732056,
+      "learning_rate": 4.9906350471406446e-05,
+      "loss": 0.6011,
+      "step": 2325
+    },
+    {
+      "epoch": 0.03139491498320985,
+      "grad_norm": 1.6071161031723022,
+      "learning_rate": 4.989894757091861e-05,
+      "loss": 0.6043,
+      "step": 2356
+    },
+    {
+      "epoch": 0.031808005969831035,
+      "grad_norm": 1.6286126375198364,
+      "learning_rate": 4.989126368919158e-05,
+      "loss": 0.6074,
+      "step": 2387
+    },
+    {
+      "epoch": 0.032221096956452214,
+      "grad_norm": 1.3929970264434814,
+      "learning_rate": 4.988329891293693e-05,
+      "loss": 0.6037,
+      "step": 2418
+    },
+    {
+      "epoch": 0.0326341879430734,
+      "grad_norm": 1.426689624786377,
+      "learning_rate": 4.987505333203608e-05,
+      "loss": 0.6037,
+      "step": 2449
+    },
+    {
+      "epoch": 0.03304727892969458,
+      "grad_norm": 1.4528875350952148,
+      "learning_rate": 4.9866527039539276e-05,
+      "loss": 0.6027,
+      "step": 2480
+    },
+    {
+      "epoch": 0.03346036991631576,
+      "grad_norm": 1.359504222869873,
+      "learning_rate": 4.9857720131664594e-05,
+      "loss": 0.5901,
+      "step": 2511
+    },
+    {
+      "epoch": 0.03387346090293694,
+      "grad_norm": 1.5734764337539673,
+      "learning_rate": 4.9848632707796773e-05,
+      "loss": 0.5981,
+      "step": 2542
+    },
+    {
+      "epoch": 0.03428655188955813,
+      "grad_norm": 1.7405669689178467,
+      "learning_rate": 4.9839264870486155e-05,
+      "loss": 0.5973,
+      "step": 2573
+    },
+    {
+      "epoch": 0.034699642876179305,
+      "grad_norm": 1.5589804649353027,
+      "learning_rate": 4.9829616725447526e-05,
+      "loss": 0.6041,
+      "step": 2604
+    },
+    {
+      "epoch": 0.03511273386280049,
+      "grad_norm": 1.260091781616211,
+      "learning_rate": 4.981968838155888e-05,
+      "loss": 0.6009,
+      "step": 2635
+    },
+    {
+      "epoch": 0.03552582484942167,
+      "grad_norm": 1.5448715686798096,
+      "learning_rate": 4.980947995086024e-05,
+      "loss": 0.5945,
+      "step": 2666
+    },
+    {
+      "epoch": 0.035938915836042855,
+      "grad_norm": 1.8747330904006958,
+      "learning_rate": 4.979899154855234e-05,
+      "loss": 0.592,
+      "step": 2697
+    },
+    {
+      "epoch": 0.03635200682266404,
+      "grad_norm": 1.553969144821167,
+      "learning_rate": 4.9788223292995386e-05,
+      "loss": 0.5892,
+      "step": 2728
+    },
+    {
+      "epoch": 0.03676509780928522,
+      "grad_norm": 1.351070523262024,
+      "learning_rate": 4.977717530570768e-05,
+      "loss": 0.5964,
+      "step": 2759
+    },
+    {
+      "epoch": 0.037178188795906404,
+      "grad_norm": 1.3044447898864746,
+      "learning_rate": 4.976584771136425e-05,
+      "loss": 0.5948,
+      "step": 2790
+    },
+    {
+      "epoch": 0.03759127978252758,
+      "grad_norm": 1.4545929431915283,
+      "learning_rate": 4.975424063779547e-05,
+      "loss": 0.5873,
+      "step": 2821
+    },
+    {
+      "epoch": 0.03800437076914877,
+      "grad_norm": 1.4722667932510376,
+      "learning_rate": 4.974235421598557e-05,
+      "loss": 0.5886,
+      "step": 2852
+    },
+    {
+      "epoch": 0.03841746175576995,
+      "grad_norm": 1.3582379817962646,
+      "learning_rate": 4.973018858007122e-05,
+      "loss": 0.598,
+      "step": 2883
+    },
+    {
+      "epoch": 0.03883055274239113,
+      "grad_norm": 1.4251989126205444,
+      "learning_rate": 4.9717743867339963e-05,
+      "loss": 0.5839,
+      "step": 2914
+    },
+    {
+      "epoch": 0.03924364372901231,
+      "grad_norm": 1.4309897422790527,
+      "learning_rate": 4.9705020218228695e-05,
+      "loss": 0.5936,
+      "step": 2945
+    },
+    {
+      "epoch": 0.039656734715633496,
+      "grad_norm": 1.5348546504974365,
+      "learning_rate": 4.969201777632205e-05,
+      "loss": 0.5806,
+      "step": 2976
+    },
+    {
+      "epoch": 0.040069825702254674,
+      "grad_norm": 1.3221420049667358,
+      "learning_rate": 4.9678736688350846e-05,
+      "loss": 0.5863,
+      "step": 3007
+    },
+    {
+      "epoch": 0.04048291668887586,
+      "grad_norm": 1.302721619606018,
+      "learning_rate": 4.966517710419033e-05,
+      "loss": 0.5854,
+      "step": 3038
+    },
+    {
+      "epoch": 0.04089600767549704,
+      "grad_norm": 1.354509949684143,
+      "learning_rate": 4.965133917685858e-05,
+      "loss": 0.5896,
+      "step": 3069
+    },
+    {
+      "epoch": 0.041309098662118224,
+      "grad_norm": 1.2658582925796509,
+      "learning_rate": 4.9637223062514714e-05,
+      "loss": 0.5839,
+      "step": 3100
+    },
+    {
+      "epoch": 0.04172218964873941,
+      "grad_norm": 1.5425275564193726,
+      "learning_rate": 4.962282892045718e-05,
+      "loss": 0.5885,
+      "step": 3131
+    },
+    {
+      "epoch": 0.04213528063536059,
+      "grad_norm": 1.4991977214813232,
+      "learning_rate": 4.9608156913121904e-05,
+      "loss": 0.5888,
+      "step": 3162
+    },
+    {
+      "epoch": 0.04254837162198177,
+      "grad_norm": 1.6346509456634521,
+      "learning_rate": 4.959320720608049e-05,
+      "loss": 0.5911,
+      "step": 3193
+    },
+    {
+      "epoch": 0.04296146260860295,
+      "grad_norm": 1.368919849395752,
+      "learning_rate": 4.9577979968038354e-05,
+      "loss": 0.5877,
+      "step": 3224
+    },
+    {
+      "epoch": 0.04337455359522414,
+      "grad_norm": 1.324533224105835,
+      "learning_rate": 4.956247537083282e-05,
+      "loss": 0.5857,
+      "step": 3255
+    },
+    {
+      "epoch": 0.043787644581845316,
+      "grad_norm": 1.315446138381958,
+      "learning_rate": 4.9546693589431145e-05,
+      "loss": 0.5881,
+      "step": 3286
+    },
+    {
+      "epoch": 0.0442007355684665,
+      "grad_norm": 1.3033638000488281,
+      "learning_rate": 4.9530634801928595e-05,
+      "loss": 0.5826,
+      "step": 3317
+    },
+    {
+      "epoch": 0.04461382655508768,
+      "grad_norm": 1.4759029150009155,
+      "learning_rate": 4.9514299189546395e-05,
+      "loss": 0.5815,
+      "step": 3348
+    },
+    {
+      "epoch": 0.045026917541708865,
+      "grad_norm": 1.4898096323013306,
+      "learning_rate": 4.949768693662973e-05,
+      "loss": 0.5755,
+      "step": 3379
+    },
+    {
+      "epoch": 0.04544000852833004,
+      "grad_norm": 1.3367149829864502,
+      "learning_rate": 4.948079823064559e-05,
+      "loss": 0.5841,
+      "step": 3410
+    },
+    {
+      "epoch": 0.04585309951495123,
+      "grad_norm": 1.5252976417541504,
+      "learning_rate": 4.946363326218074e-05,
+      "loss": 0.5882,
+      "step": 3441
+    },
+    {
+      "epoch": 0.046266190501572414,
+      "grad_norm": 1.3493881225585938,
+      "learning_rate": 4.9446192224939525e-05,
+      "loss": 0.5896,
+      "step": 3472
+    },
+    {
+      "epoch": 0.04667928148819359,
+      "grad_norm": 1.275595784187317,
+      "learning_rate": 4.942847531574167e-05,
+      "loss": 0.5852,
+      "step": 3503
+    },
+    {
+      "epoch": 0.04709237247481478,
+      "grad_norm": 1.4550764560699463,
+      "learning_rate": 4.941048273452008e-05,
+      "loss": 0.5702,
+      "step": 3534
+    },
+    {
+      "epoch": 0.04750546346143596,
+      "grad_norm": 1.2401679754257202,
+      "learning_rate": 4.9392214684318605e-05,
+      "loss": 0.5841,
+      "step": 3565
+    },
+    {
+      "epoch": 0.04791855444805714,
+      "grad_norm": 1.3234368562698364,
+      "learning_rate": 4.93736713712897e-05,
+      "loss": 0.5751,
+      "step": 3596
+    },
+    {
+      "epoch": 0.04833164543467832,
+      "grad_norm": 1.4088999032974243,
+      "learning_rate": 4.9354853004692124e-05,
+      "loss": 0.5758,
+      "step": 3627
+    },
+    {
+      "epoch": 0.048744736421299506,
+      "grad_norm": 1.3294583559036255,
+      "learning_rate": 4.93357597968886e-05,
+      "loss": 0.5855,
+      "step": 3658
+    },
+    {
+      "epoch": 0.049157827407920685,
+      "grad_norm": 1.4270315170288086,
+      "learning_rate": 4.931639196334338e-05,
+      "loss": 0.5751,
+      "step": 3689
+    },
+    {
+      "epoch": 0.04957091839454187,
+      "grad_norm": 1.3925172090530396,
+      "learning_rate": 4.9296749722619826e-05,
+      "loss": 0.5724,
+      "step": 3720
+    },
+    {
+      "epoch": 0.04998400938116305,
+      "grad_norm": 1.2499849796295166,
+      "learning_rate": 4.9276833296377966e-05,
+      "loss": 0.5833,
+      "step": 3751
+    },
+    {
+      "epoch": 0.050397100367784234,
+      "grad_norm": 1.2049689292907715,
+      "learning_rate": 4.925664290937196e-05,
+      "loss": 0.5808,
+      "step": 3782
+    },
+    {
+      "epoch": 0.05081019135440541,
+      "grad_norm": 1.604710578918457,
+      "learning_rate": 4.9236178789447576e-05,
+      "loss": 0.5829,
+      "step": 3813
+    },
+    {
+      "epoch": 0.0512232823410266,
+      "grad_norm": 1.3707163333892822,
+      "learning_rate": 4.921544116753962e-05,
+      "loss": 0.5813,
+      "step": 3844
+    },
+    {
+      "epoch": 0.05163637332764778,
+      "grad_norm": 1.2728792428970337,
+      "learning_rate": 4.919443027766935e-05,
+      "loss": 0.5764,
+      "step": 3875
+    },
+    {
+      "epoch": 0.05204946431426896,
+      "grad_norm": 1.197830319404602,
+      "learning_rate": 4.91731463569418e-05,
+      "loss": 0.5713,
+      "step": 3906
+    },
+    {
+      "epoch": 0.05246255530089015,
+      "grad_norm": 1.3592743873596191,
+      "learning_rate": 4.915158964554312e-05,
+      "loss": 0.5877,
+      "step": 3937
+    },
+    {
+      "epoch": 0.052875646287511326,
+      "grad_norm": 1.2711323499679565,
+      "learning_rate": 4.912976038673786e-05,
+      "loss": 0.5737,
+      "step": 3968
+    },
+    {
+      "epoch": 0.05328873727413251,
+      "grad_norm": 1.2760255336761475,
+      "learning_rate": 4.9107658826866254e-05,
+      "loss": 0.5806,
+      "step": 3999
+    },
+    {
+      "epoch": 0.05370182826075369,
+      "grad_norm": 1.5267741680145264,
+      "learning_rate": 4.908528521534139e-05,
+      "loss": 0.5794,
+      "step": 4030
+    },
+    {
+      "epoch": 0.054114919247374875,
+      "grad_norm": 1.318003535270691,
+      "learning_rate": 4.906263980464644e-05,
+      "loss": 0.5683,
+      "step": 4061
+    },
+    {
+      "epoch": 0.05452801023399605,
+      "grad_norm": 1.4435062408447266,
+      "learning_rate": 4.903972285033178e-05,
+      "loss": 0.5772,
+      "step": 4092
+    },
+    {
+      "epoch": 0.05494110122061724,
+      "grad_norm": 1.4204825162887573,
+      "learning_rate": 4.901653461101213e-05,
+      "loss": 0.5794,
+      "step": 4123
+    },
+    {
+      "epoch": 0.05535419220723842,
+      "grad_norm": 1.222264289855957,
+      "learning_rate": 4.8993075348363626e-05,
+      "loss": 0.5681,
+      "step": 4154
+    },
+    {
+      "epoch": 0.0557672831938596,
+      "grad_norm": 1.3331806659698486,
+      "learning_rate": 4.896934532712084e-05,
+      "loss": 0.5761,
+      "step": 4185
+    },
+    {
+      "epoch": 0.05618037418048078,
+      "grad_norm": 1.2175304889678955,
+      "learning_rate": 4.8945344815073846e-05,
+      "loss": 0.5718,
+      "step": 4216
+    },
+    {
+      "epoch": 0.05659346516710197,
+      "grad_norm": 1.2743812799453735,
+      "learning_rate": 4.892107408306516e-05,
+      "loss": 0.5801,
+      "step": 4247
+    },
+    {
+      "epoch": 0.05700655615372315,
+      "grad_norm": 1.2461763620376587,
+      "learning_rate": 4.889653340498669e-05,
+      "loss": 0.5756,
+      "step": 4278
+    },
+    {
+      "epoch": 0.05741964714034433,
+      "grad_norm": 1.3406884670257568,
+      "learning_rate": 4.8871723057776664e-05,
+      "loss": 0.5787,
+      "step": 4309
+    },
+    {
+      "epoch": 0.057832738126965516,
+      "grad_norm": 1.3336981534957886,
+      "learning_rate": 4.8846643321416476e-05,
+      "loss": 0.568,
+      "step": 4340
+    },
+    {
+      "epoch": 0.058245829113586695,
+      "grad_norm": 1.194069743156433,
+      "learning_rate": 4.882129447892753e-05,
+      "loss": 0.5764,
+      "step": 4371
+    },
+    {
+      "epoch": 0.05865892010020788,
+      "grad_norm": 1.3039817810058594,
+      "learning_rate": 4.8795676816368076e-05,
+      "loss": 0.5712,
+      "step": 4402
+    },
+    {
+      "epoch": 0.05907201108682906,
+      "grad_norm": 1.2447880506515503,
+      "learning_rate": 4.876979062282995e-05,
+      "loss": 0.5722,
+      "step": 4433
+    },
+    {
+      "epoch": 0.059485102073450244,
+      "grad_norm": 1.3227744102478027,
+      "learning_rate": 4.8743636190435325e-05,
+      "loss": 0.5756,
+      "step": 4464
+    },
+    {
+      "epoch": 0.05989819306007142,
+      "grad_norm": 1.2006003856658936,
+      "learning_rate": 4.871721381433344e-05,
+      "loss": 0.5582,
+      "step": 4495
+    },
+    {
+      "epoch": 0.06031128404669261,
+      "grad_norm": 1.5150734186172485,
+      "learning_rate": 4.869052379269719e-05,
+      "loss": 0.5773,
+      "step": 4526
+    },
+    {
+      "epoch": 0.060724375033313786,
+      "grad_norm": 1.1430795192718506,
+      "learning_rate": 4.866356642671985e-05,
+      "loss": 0.5712,
+      "step": 4557
+    },
+    {
+      "epoch": 0.06113746601993497,
+      "grad_norm": 1.181968331336975,
+      "learning_rate": 4.8636342020611634e-05,
+      "loss": 0.5778,
+      "step": 4588
+    },
+    {
+      "epoch": 0.06155055700655616,
+      "grad_norm": 1.0638952255249023,
+      "learning_rate": 4.860885088159626e-05,
+      "loss": 0.5646,
+      "step": 4619
+    },
+    {
+      "epoch": 0.061963647993177336,
+      "grad_norm": 1.2407903671264648,
+      "learning_rate": 4.858109331990751e-05,
+      "loss": 0.5694,
+      "step": 4650
+    },
+    {
+      "epoch": 0.06237673897979852,
+      "grad_norm": 1.1231423616409302,
+      "learning_rate": 4.855306964878567e-05,
+      "loss": 0.5732,
+      "step": 4681
+    },
+    {
+      "epoch": 0.0627898299664197,
+      "grad_norm": 1.324874997138977,
+      "learning_rate": 4.8524780184474084e-05,
+      "loss": 0.5677,
+      "step": 4712
+    },
+    {
+      "epoch": 0.06320292095304088,
+      "grad_norm": 1.2522512674331665,
+      "learning_rate": 4.8496225246215496e-05,
+      "loss": 0.5692,
+      "step": 4743
+    },
+    {
+      "epoch": 0.06361601193966207,
+      "grad_norm": 1.3129041194915771,
+      "learning_rate": 4.8467405156248505e-05,
+      "loss": 0.5718,
+      "step": 4774
+    },
+    {
+      "epoch": 0.06402910292628325,
+      "grad_norm": 1.2068370580673218,
+      "learning_rate": 4.843832023980392e-05,
+      "loss": 0.5701,
+      "step": 4805
+    },
+    {
+      "epoch": 0.06444219391290443,
+      "grad_norm": 1.2644529342651367,
+      "learning_rate": 4.840897082510106e-05,
+      "loss": 0.5711,
+      "step": 4836
+    },
+    {
+      "epoch": 0.0648552848995256,
+      "grad_norm": 1.3398981094360352,
+      "learning_rate": 4.8379357243344084e-05,
+      "loss": 0.5588,
+      "step": 4867
+    },
+    {
+      "epoch": 0.0652683758861468,
+      "grad_norm": 1.208992600440979,
+      "learning_rate": 4.8349479828718236e-05,
+      "loss": 0.5634,
+      "step": 4898
+    },
+    {
+      "epoch": 0.06568146687276798,
+      "grad_norm": 1.1965893507003784,
+      "learning_rate": 4.8319338918386075e-05,
+      "loss": 0.5606,
+      "step": 4929
+    },
+    {
+      "epoch": 0.06609455785938916,
+      "grad_norm": 1.0229936838150024,
+      "learning_rate": 4.828893485248369e-05,
+      "loss": 0.5708,
+      "step": 4960
+    },
+    {
+      "epoch": 0.06650764884601033,
+      "grad_norm": 1.3772759437561035,
+      "learning_rate": 4.825826797411682e-05,
+      "loss": 0.5671,
+      "step": 4991
+    },
+    {
+      "epoch": 0.06692073983263153,
+      "grad_norm": 1.2997850179672241,
+      "learning_rate": 4.822733862935702e-05,
+      "loss": 0.5615,
+      "step": 5022
+    },
+    {
+      "epoch": 0.0673338308192527,
+      "grad_norm": 1.2622663974761963,
+      "learning_rate": 4.819614716723775e-05,
+      "loss": 0.5628,
+      "step": 5053
+    },
+    {
+      "epoch": 0.06774692180587388,
+      "grad_norm": 0.981414794921875,
+      "learning_rate": 4.8164693939750425e-05,
+      "loss": 0.5573,
+      "step": 5084
+    },
+    {
+      "epoch": 0.06816001279249508,
+      "grad_norm": 1.3478119373321533,
+      "learning_rate": 4.813297930184042e-05,
+      "loss": 0.5663,
+      "step": 5115
+    },
+    {
+      "epoch": 0.06857310377911625,
+      "grad_norm": 1.2769936323165894,
+      "learning_rate": 4.810100361140314e-05,
+      "loss": 0.5648,
+      "step": 5146
+    },
+    {
+      "epoch": 0.06898619476573743,
+      "grad_norm": 1.181077003479004,
+      "learning_rate": 4.8068767229279885e-05,
+      "loss": 0.5699,
+      "step": 5177
+    },
+    {
+      "epoch": 0.06939928575235861,
+      "grad_norm": 1.2607613801956177,
+      "learning_rate": 4.8036270519253854e-05,
+      "loss": 0.5609,
+      "step": 5208
+    },
+    {
+      "epoch": 0.0698123767389798,
+      "grad_norm": 1.242676854133606,
+      "learning_rate": 4.8003513848046e-05,
+      "loss": 0.5609,
+      "step": 5239
+    },
+    {
+      "epoch": 0.07022546772560098,
+      "grad_norm": 1.1886576414108276,
+      "learning_rate": 4.79704975853109e-05,
+      "loss": 0.5665,
+      "step": 5270
+    },
+    {
+      "epoch": 0.07063855871222216,
+      "grad_norm": 1.1143399477005005,
+      "learning_rate": 4.793722210363262e-05,
+      "loss": 0.5573,
+      "step": 5301
+    },
+    {
+      "epoch": 0.07105164969884334,
+      "grad_norm": 1.1033484935760498,
+      "learning_rate": 4.7903687778520414e-05,
+      "loss": 0.5566,
+      "step": 5332
+    },
+    {
+      "epoch": 0.07146474068546453,
+      "grad_norm": 1.2626802921295166,
+      "learning_rate": 4.7869894988404593e-05,
+      "loss": 0.5677,
+      "step": 5363
+    },
+    {
+      "epoch": 0.07187783167208571,
+      "grad_norm": 1.2453737258911133,
+      "learning_rate": 4.783584411463221e-05,
+      "loss": 0.5615,
+      "step": 5394
+    },
+    {
+      "epoch": 0.07229092265870689,
+      "grad_norm": 1.2548526525497437,
+      "learning_rate": 4.780153554146274e-05,
+      "loss": 0.5611,
+      "step": 5425
+    },
+    {
+      "epoch": 0.07270401364532808,
+      "grad_norm": 1.2009012699127197,
+      "learning_rate": 4.7766969656063766e-05,
+      "loss": 0.5648,
+      "step": 5456
+    },
+    {
+      "epoch": 0.07311710463194926,
+      "grad_norm": 1.1282159090042114,
+      "learning_rate": 4.773214684850662e-05,
+      "loss": 0.5584,
+      "step": 5487
+    },
+    {
+      "epoch": 0.07353019561857044,
+      "grad_norm": 1.1493905782699585,
+      "learning_rate": 4.769706751176193e-05,
+      "loss": 0.5655,
+      "step": 5518
+    },
+    {
+      "epoch": 0.07394328660519162,
+      "grad_norm": 1.1352925300598145,
+      "learning_rate": 4.7661732041695264e-05,
+      "loss": 0.5641,
+      "step": 5549
+    },
+    {
+      "epoch": 0.07435637759181281,
+      "grad_norm": 1.1498860120773315,
+      "learning_rate": 4.762614083706258e-05,
+      "loss": 0.5609,
+      "step": 5580
+    },
+    {
+      "epoch": 0.07476946857843399,
+      "grad_norm": 1.3136110305786133,
+      "learning_rate": 4.759029429950581e-05,
+      "loss": 0.5565,
+      "step": 5611
+    },
+    {
+      "epoch": 0.07518255956505517,
+      "grad_norm": 1.1240262985229492,
+      "learning_rate": 4.7554192833548235e-05,
+      "loss": 0.5605,
+      "step": 5642
+    },
+    {
+      "epoch": 0.07559565055167634,
+      "grad_norm": 1.4061548709869385,
+      "learning_rate": 4.751783684659e-05,
+      "loss": 0.5637,
+      "step": 5673
+    },
+    {
+      "epoch": 0.07600874153829754,
+      "grad_norm": 1.2112833261489868,
+      "learning_rate": 4.748122674890348e-05,
+      "loss": 0.564,
+      "step": 5704
+    },
+    {
+      "epoch": 0.07642183252491871,
+      "grad_norm": 1.1065545082092285,
+      "learning_rate": 4.7444362953628654e-05,
+      "loss": 0.553,
+      "step": 5735
+    },
+    {
+      "epoch": 0.0768349235115399,
+      "grad_norm": 1.3150484561920166,
+      "learning_rate": 4.7407245876768424e-05,
+      "loss": 0.56,
+      "step": 5766
+    },
+    {
+      "epoch": 0.07724801449816109,
+      "grad_norm": 1.0889962911605835,
+      "learning_rate": 4.736987593718397e-05,
+      "loss": 0.557,
+      "step": 5797
+    },
+    {
+      "epoch": 0.07766110548478226,
+      "grad_norm": 1.0983312129974365,
+      "learning_rate": 4.733225355658999e-05,
+      "loss": 0.5517,
+      "step": 5828
+    },
+    {
+      "epoch": 0.07807419647140344,
+      "grad_norm": 1.1129499673843384,
+      "learning_rate": 4.7294379159549926e-05,
+      "loss": 0.5572,
+      "step": 5859
+    },
+    {
+      "epoch": 0.07848728745802462,
+      "grad_norm": 1.1074233055114746,
+      "learning_rate": 4.725625317347119e-05,
+      "loss": 0.5548,
+      "step": 5890
+    },
+    {
+      "epoch": 0.07890037844464581,
+      "grad_norm": 1.3234531879425049,
+      "learning_rate": 4.7217876028600374e-05,
+      "loss": 0.5536,
+      "step": 5921
+    },
+    {
+      "epoch": 0.07931346943126699,
+      "grad_norm": 1.2597479820251465,
+      "learning_rate": 4.717924815801832e-05,
+      "loss": 0.5647,
+      "step": 5952
+    },
+    {
+      "epoch": 0.07972656041788817,
+      "grad_norm": 1.2188916206359863,
+      "learning_rate": 4.714036999763532e-05,
+      "loss": 0.5588,
+      "step": 5983
+    },
+    {
+      "epoch": 0.08013965140450935,
+      "grad_norm": 1.2447947263717651,
+      "learning_rate": 4.7101241986186116e-05,
+      "loss": 0.566,
+      "step": 6014
+    },
+    {
+      "epoch": 0.08055274239113054,
+      "grad_norm": 1.1489174365997314,
+      "learning_rate": 4.7061864565225e-05,
+      "loss": 0.5536,
+      "step": 6045
+    },
+    {
+      "epoch": 0.08096583337775172,
+      "grad_norm": 1.119452953338623,
+      "learning_rate": 4.702223817912081e-05,
+      "loss": 0.5496,
+      "step": 6076
+    },
+    {
+      "epoch": 0.0813789243643729,
+      "grad_norm": 1.109541416168213,
+      "learning_rate": 4.698236327505195e-05,
+      "loss": 0.5463,
+      "step": 6107
+    },
+    {
+      "epoch": 0.08179201535099408,
+      "grad_norm": 1.1216788291931152,
+      "learning_rate": 4.694224030300127e-05,
+      "loss": 0.5602,
+      "step": 6138
+    },
+    {
+      "epoch": 0.08220510633761527,
+      "grad_norm": 1.1995866298675537,
+      "learning_rate": 4.690186971575107e-05,
+      "loss": 0.5576,
+      "step": 6169
+    },
+    {
+      "epoch": 0.08261819732423645,
+      "grad_norm": 1.0980093479156494,
+      "learning_rate": 4.6861251968877916e-05,
+      "loss": 0.5546,
+      "step": 6200
+    },
+    {
+      "epoch": 0.08303128831085763,
+      "grad_norm": 1.0906014442443848,
+      "learning_rate": 4.68203875207476e-05,
+      "loss": 0.5523,
+      "step": 6231
+    },
+    {
+      "epoch": 0.08344437929747882,
+      "grad_norm": 1.1709624528884888,
+      "learning_rate": 4.677927683250983e-05,
+      "loss": 0.5475,
+      "step": 6262
+    },
+    {
+      "epoch": 0.0838574702841,
+      "grad_norm": 0.9864281415939331,
+      "learning_rate": 4.6737920368093156e-05,
+      "loss": 0.5671,
+      "step": 6293
+    },
+    {
+      "epoch": 0.08427056127072118,
+      "grad_norm": 1.1931488513946533,
+      "learning_rate": 4.669631859419965e-05,
+      "loss": 0.5607,
+      "step": 6324
+    },
+    {
+      "epoch": 0.08468365225734235,
+      "grad_norm": 1.1696871519088745,
+      "learning_rate": 4.6654471980299676e-05,
+      "loss": 0.5632,
+      "step": 6355
+    },
+    {
+      "epoch": 0.08509674324396355,
+      "grad_norm": 1.1524802446365356,
+      "learning_rate": 4.661238099862658e-05,
+      "loss": 0.5446,
+      "step": 6386
+    },
+    {
+      "epoch": 0.08550983423058472,
+      "grad_norm": 1.0331172943115234,
+      "learning_rate": 4.657004612417138e-05,
+      "loss": 0.5496,
+      "step": 6417
+    },
+    {
+      "epoch": 0.0859229252172059,
+      "grad_norm": 1.132674217224121,
+      "learning_rate": 4.6527467834677374e-05,
+      "loss": 0.5583,
+      "step": 6448
+    },
+    {
+      "epoch": 0.08633601620382708,
+      "grad_norm": 1.140329360961914,
+      "learning_rate": 4.648464661063478e-05,
+      "loss": 0.549,
+      "step": 6479
+    },
+    {
+      "epoch": 0.08674910719044827,
+      "grad_norm": 1.0936356782913208,
+      "learning_rate": 4.6441582935275264e-05,
+      "loss": 0.5504,
+      "step": 6510
+    },
+    {
+      "epoch": 0.08716219817706945,
+      "grad_norm": 1.0445035696029663,
+      "learning_rate": 4.6398277294566586e-05,
+      "loss": 0.5524,
+      "step": 6541
+    },
+    {
+      "epoch": 0.08757528916369063,
+      "grad_norm": 1.013694167137146,
+      "learning_rate": 4.6354730177207e-05,
+      "loss": 0.551,
+      "step": 6572
+    },
+    {
+      "epoch": 0.08798838015031182,
+      "grad_norm": 1.2154170274734497,
+      "learning_rate": 4.6310942074619787e-05,
+      "loss": 0.5563,
+      "step": 6603
+    },
+    {
+      "epoch": 0.088401471136933,
+      "grad_norm": 1.0196465253829956,
+      "learning_rate": 4.626691348094777e-05,
+      "loss": 0.5604,
+      "step": 6634
+    },
+    {
+      "epoch": 0.08881456212355418,
+      "grad_norm": 1.2061710357666016,
+      "learning_rate": 4.622264489304762e-05,
+      "loss": 0.5515,
+      "step": 6665
+    },
+    {
+      "epoch": 0.08922765311017536,
+      "grad_norm": 1.3378483057022095,
+      "learning_rate": 4.617813681048434e-05,
+      "loss": 0.5466,
+      "step": 6696
+    },
+    {
+      "epoch": 0.08964074409679655,
+      "grad_norm": 1.1906929016113281,
+      "learning_rate": 4.61333897355256e-05,
+      "loss": 0.5484,
+      "step": 6727
+    },
+    {
+      "epoch": 0.09005383508341773,
+      "grad_norm": 1.1115672588348389,
+      "learning_rate": 4.608840417313604e-05,
+      "loss": 0.5534,
+      "step": 6758
+    },
+    {
+      "epoch": 0.09046692607003891,
+      "grad_norm": 1.0681543350219727,
+      "learning_rate": 4.6043180630971646e-05,
+      "loss": 0.549,
+      "step": 6789
+    },
+    {
+      "epoch": 0.09088001705666009,
+      "grad_norm": 1.111663579940796,
+      "learning_rate": 4.599771961937391e-05,
+      "loss": 0.5529,
+      "step": 6820
+    },
+    {
+      "epoch": 0.09129310804328128,
+      "grad_norm": 1.111009120941162,
+      "learning_rate": 4.5952021651364204e-05,
+      "loss": 0.549,
+      "step": 6851
+    },
+    {
+      "epoch": 0.09170619902990246,
+      "grad_norm": 1.075073480606079,
+      "learning_rate": 4.590608724263786e-05,
+      "loss": 0.5502,
+      "step": 6882
+    },
+    {
+      "epoch": 0.09211929001652364,
+      "grad_norm": 1.2174952030181885,
+      "learning_rate": 4.585991691155845e-05,
+      "loss": 0.5499,
+      "step": 6913
+    },
+    {
+      "epoch": 0.09253238100314483,
+      "grad_norm": 1.0371384620666504,
+      "learning_rate": 4.581351117915188e-05,
+      "loss": 0.5481,
+      "step": 6944
+    },
+    {
+      "epoch": 0.09294547198976601,
+      "grad_norm": 1.1694340705871582,
+      "learning_rate": 4.5766870569100534e-05,
+      "loss": 0.5543,
+      "step": 6975
+    },
+    {
+      "epoch": 0.09335856297638719,
+      "grad_norm": 1.0203725099563599,
+      "learning_rate": 4.571999560773736e-05,
+      "loss": 0.5537,
+      "step": 7006
+    },
+    {
+      "epoch": 0.09377165396300836,
+      "grad_norm": 1.1278784275054932,
+      "learning_rate": 4.5672886824039915e-05,
+      "loss": 0.5451,
+      "step": 7037
+    },
+    {
+      "epoch": 0.09418474494962956,
+      "grad_norm": 1.2053276300430298,
+      "learning_rate": 4.5625544749624435e-05,
+      "loss": 0.5526,
+      "step": 7068
+    },
+    {
+      "epoch": 0.09459783593625073,
+      "grad_norm": 1.021291971206665,
+      "learning_rate": 4.5577969918739794e-05,
+      "loss": 0.5419,
+      "step": 7099
+    },
+    {
+      "epoch": 0.09501092692287191,
+      "grad_norm": 1.146509051322937,
+      "learning_rate": 4.5530162868261486e-05,
+      "loss": 0.5514,
+      "step": 7130
+    },
+    {
+      "epoch": 0.09542401790949309,
+      "grad_norm": 1.055192232131958,
+      "learning_rate": 4.548212413768558e-05,
+      "loss": 0.5572,
+      "step": 7161
+    },
+    {
+      "epoch": 0.09583710889611428,
+      "grad_norm": 1.0573087930679321,
+      "learning_rate": 4.543385426912261e-05,
+      "loss": 0.555,
+      "step": 7192
+    },
+    {
+      "epoch": 0.09625019988273546,
+      "grad_norm": 1.189032793045044,
+      "learning_rate": 4.53853538072915e-05,
+      "loss": 0.5421,
+      "step": 7223
+    },
+    {
+      "epoch": 0.09666329086935664,
+      "grad_norm": 1.0390995740890503,
+      "learning_rate": 4.533662329951336e-05,
+      "loss": 0.5442,
+      "step": 7254
+    },
+    {
+      "epoch": 0.09707638185597782,
+      "grad_norm": 1.0945408344268799,
+      "learning_rate": 4.528766329570536e-05,
+      "loss": 0.5529,
+      "step": 7285
+    },
+    {
+      "epoch": 0.09748947284259901,
+      "grad_norm": 1.2713547945022583,
+      "learning_rate": 4.523847434837447e-05,
+      "loss": 0.5452,
+      "step": 7316
+    },
+    {
+      "epoch": 0.09790256382922019,
+      "grad_norm": 1.123363733291626,
+      "learning_rate": 4.518905701261128e-05,
+      "loss": 0.5534,
+      "step": 7347
+    },
+    {
+      "epoch": 0.09831565481584137,
+      "grad_norm": 1.0784560441970825,
+      "learning_rate": 4.5139411846083715e-05,
+      "loss": 0.5471,
+      "step": 7378
+    },
+    {
+      "epoch": 0.09872874580246256,
+      "grad_norm": 1.1390719413757324,
+      "learning_rate": 4.508953940903073e-05,
+      "loss": 0.5526,
+      "step": 7409
+    },
+    {
+      "epoch": 0.09914183678908374,
+      "grad_norm": 0.9794678688049316,
+      "learning_rate": 4.5039440264255994e-05,
+      "loss": 0.542,
+      "step": 7440
+    },
+    {
+      "epoch": 0.09955492777570492,
+      "grad_norm": 1.1383626461029053,
+      "learning_rate": 4.498911497712155e-05,
+      "loss": 0.5396,
+      "step": 7471
+    },
+    {
+      "epoch": 0.0999680187623261,
+      "grad_norm": 1.0440157651901245,
+      "learning_rate": 4.493856411554142e-05,
+      "loss": 0.5461,
+      "step": 7502
+    },
+    {
+      "epoch": 0.10038110974894729,
+      "grad_norm": 1.1624081134796143,
+      "learning_rate": 4.4887788249975206e-05,
+      "loss": 0.5439,
+      "step": 7533
+    },
+    {
+      "epoch": 0.10079420073556847,
+      "grad_norm": 1.0614362955093384,
+      "learning_rate": 4.4836787953421656e-05,
+      "loss": 0.55,
+      "step": 7564
+    },
+    {
+      "epoch": 0.10120729172218965,
+      "grad_norm": 1.2120076417922974,
+      "learning_rate": 4.478556380141218e-05,
+      "loss": 0.5517,
+      "step": 7595
+    },
+    {
+      "epoch": 0.10162038270881082,
+      "grad_norm": 0.9873495697975159,
+      "learning_rate": 4.4734116372004375e-05,
+      "loss": 0.5436,
+      "step": 7626
+    },
+    {
+      "epoch": 0.10203347369543202,
+      "grad_norm": 0.9908550381660461,
+      "learning_rate": 4.4682446245775477e-05,
+      "loss": 0.5412,
+      "step": 7657
+    },
+    {
+      "epoch": 0.1024465646820532,
+      "grad_norm": 1.065071940422058,
+      "learning_rate": 4.463055400581586e-05,
+      "loss": 0.5433,
+      "step": 7688
+    },
+    {
+      "epoch": 0.10285965566867437,
+      "grad_norm": 1.1011327505111694,
+      "learning_rate": 4.4578440237722374e-05,
+      "loss": 0.5469,
+      "step": 7719
+    },
+    {
+      "epoch": 0.10327274665529557,
+      "grad_norm": 1.1491224765777588,
+      "learning_rate": 4.452610552959183e-05,
+      "loss": 0.545,
+      "step": 7750
+    },
+    {
+      "epoch": 0.10368583764191674,
+      "grad_norm": 1.0363606214523315,
+      "learning_rate": 4.447355047201428e-05,
+      "loss": 0.5458,
+      "step": 7781
+    },
+    {
+      "epoch": 0.10409892862853792,
+      "grad_norm": 1.137918472290039,
+      "learning_rate": 4.4420775658066414e-05,
+      "loss": 0.5358,
+      "step": 7812
+    },
+    {
+      "epoch": 0.1045120196151591,
+      "grad_norm": 1.1191500425338745,
+      "learning_rate": 4.436778168330484e-05,
+      "loss": 0.5414,
+      "step": 7843
+    },
+    {
+      "epoch": 0.1049251106017803,
+      "grad_norm": 1.10321044921875,
+      "learning_rate": 4.4314569145759353e-05,
+      "loss": 0.5439,
+      "step": 7874
+    },
+    {
+      "epoch": 0.10533820158840147,
+      "grad_norm": 1.0266538858413696,
+      "learning_rate": 4.42611386459262e-05,
+      "loss": 0.5418,
+      "step": 7905
+    },
+    {
+      "epoch": 0.10575129257502265,
+      "grad_norm": 1.050444483757019,
+      "learning_rate": 4.420749078676133e-05,
+      "loss": 0.5462,
+      "step": 7936
+    },
+    {
+      "epoch": 0.10616438356164383,
+      "grad_norm": 1.0405640602111816,
+      "learning_rate": 4.4153626173673516e-05,
+      "loss": 0.5464,
+      "step": 7967
+    },
+    {
+      "epoch": 0.10657747454826502,
+      "grad_norm": 1.1248273849487305,
+      "learning_rate": 4.409954541451762e-05,
+      "loss": 0.5436,
+      "step": 7998
+    },
+    {
+      "epoch": 0.1069905655348862,
+      "grad_norm": 1.1844524145126343,
+      "learning_rate": 4.404524911958764e-05,
+      "loss": 0.5506,
+      "step": 8029
+    },
+    {
+      "epoch": 0.10740365652150738,
+      "grad_norm": 0.9842180013656616,
+      "learning_rate": 4.399073790160989e-05,
+      "loss": 0.5408,
+      "step": 8060
+    },
+    {
+      "epoch": 0.10781674750812857,
+      "grad_norm": 1.0924715995788574,
+      "learning_rate": 4.393601237573607e-05,
+      "loss": 0.539,
+      "step": 8091
+    },
+    {
+      "epoch": 0.10822983849474975,
+      "grad_norm": 1.0117294788360596,
+      "learning_rate": 4.388107315953628e-05,
+      "loss": 0.5474,
+      "step": 8122
+    },
+    {
+      "epoch": 0.10864292948137093,
+      "grad_norm": 0.9410596489906311,
+      "learning_rate": 4.382592087299212e-05,
+      "loss": 0.5413,
+      "step": 8153
+    },
+    {
+      "epoch": 0.1090560204679921,
+      "grad_norm": 1.1258411407470703,
+      "learning_rate": 4.377055613848964e-05,
+      "loss": 0.5378,
+      "step": 8184
+    },
+    {
+      "epoch": 0.1094691114546133,
+      "grad_norm": 1.0934258699417114,
+      "learning_rate": 4.3714979580812355e-05,
+      "loss": 0.5429,
+      "step": 8215
+    },
+    {
+      "epoch": 0.10988220244123448,
+      "grad_norm": 1.0256693363189697,
+      "learning_rate": 4.365919182713416e-05,
+      "loss": 0.5521,
+      "step": 8246
+    },
+    {
+      "epoch": 0.11029529342785566,
+      "grad_norm": 1.1147851943969727,
+      "learning_rate": 4.360319350701226e-05,
+      "loss": 0.5367,
+      "step": 8277
+    },
+    {
+      "epoch": 0.11070838441447683,
+      "grad_norm": 1.065674066543579,
+      "learning_rate": 4.3546985252380115e-05,
+      "loss": 0.5399,
+      "step": 8308
+    },
+    {
+      "epoch": 0.11112147540109803,
+      "grad_norm": 1.096153736114502,
+      "learning_rate": 4.349056769754021e-05,
+      "loss": 0.5358,
+      "step": 8339
+    },
+    {
+      "epoch": 0.1115345663877192,
+      "grad_norm": 0.9637765884399414,
+      "learning_rate": 4.3433941479156994e-05,
+      "loss": 0.5411,
+      "step": 8370
+    },
+    {
+      "epoch": 0.11194765737434038,
+      "grad_norm": 1.0165079832077026,
+      "learning_rate": 4.3377107236249647e-05,
+      "loss": 0.5404,
+      "step": 8401
+    },
+    {
+      "epoch": 0.11236074836096156,
+      "grad_norm": 1.0558301210403442,
+      "learning_rate": 4.332006561018488e-05,
+      "loss": 0.5421,
+      "step": 8432
+    },
+    {
+      "epoch": 0.11277383934758275,
+      "grad_norm": 0.9278895854949951,
+      "learning_rate": 4.3262817244669683e-05,
+      "loss": 0.5309,
+      "step": 8463
+    },
+    {
+      "epoch": 0.11318693033420393,
+      "grad_norm": 1.0022321939468384,
+      "learning_rate": 4.3205362785744083e-05,
+      "loss": 0.5432,
+      "step": 8494
+    },
+    {
+      "epoch": 0.11360002132082511,
+      "grad_norm": 1.0881727933883667,
+      "learning_rate": 4.314770288177384e-05,
+      "loss": 0.5324,
+      "step": 8525
+    },
+    {
+      "epoch": 0.1140131123074463,
+      "grad_norm": 1.1254336833953857,
+      "learning_rate": 4.308983818344313e-05,
+      "loss": 0.5485,
+      "step": 8556
+    },
+    {
+      "epoch": 0.11442620329406748,
+      "grad_norm": 0.9436982274055481,
+      "learning_rate": 4.3031769343747206e-05,
+      "loss": 0.5402,
+      "step": 8587
+    },
+    {
+      "epoch": 0.11483929428068866,
+      "grad_norm": 1.0071033239364624,
+      "learning_rate": 4.297349701798505e-05,
+      "loss": 0.5493,
+      "step": 8618
+    },
+    {
+      "epoch": 0.11525238526730984,
+      "grad_norm": 1.1261751651763916,
+      "learning_rate": 4.2915021863751916e-05,
+      "loss": 0.5327,
+      "step": 8649
+    },
+    {
+      "epoch": 0.11566547625393103,
+      "grad_norm": 1.0863945484161377,
+      "learning_rate": 4.285634454093198e-05,
+      "loss": 0.5429,
+      "step": 8680
+    },
+    {
+      "epoch": 0.11607856724055221,
+      "grad_norm": 1.0068254470825195,
+      "learning_rate": 4.279746571169086e-05,
+      "loss": 0.5387,
+      "step": 8711
+    },
+    {
+      "epoch": 0.11649165822717339,
+      "grad_norm": 1.219416618347168,
+      "learning_rate": 4.2738386040468136e-05,
+      "loss": 0.536,
+      "step": 8742
+    },
+    {
+      "epoch": 0.11690474921379457,
+      "grad_norm": 1.0604240894317627,
+      "learning_rate": 4.2679106193969866e-05,
+      "loss": 0.5326,
+      "step": 8773
+    },
+    {
+      "epoch": 0.11731784020041576,
+      "grad_norm": 1.0701854228973389,
+      "learning_rate": 4.261962684116106e-05,
+      "loss": 0.5386,
+      "step": 8804
+    },
+    {
+      "epoch": 0.11773093118703694,
+      "grad_norm": 1.18992018699646,
+      "learning_rate": 4.2559948653258145e-05,
+      "loss": 0.5442,
+      "step": 8835
+    },
+    {
+      "epoch": 0.11814402217365812,
+      "grad_norm": 1.0788195133209229,
+      "learning_rate": 4.250007230372134e-05,
+      "loss": 0.5448,
+      "step": 8866
+    },
+    {
+      "epoch": 0.11855711316027931,
+      "grad_norm": 1.2016948461532593,
+      "learning_rate": 4.2439998468247126e-05,
+      "loss": 0.5407,
+      "step": 8897
+    },
+    {
+      "epoch": 0.11897020414690049,
+      "grad_norm": 0.9844837188720703,
+      "learning_rate": 4.2379727824760566e-05,
+      "loss": 0.541,
+      "step": 8928
+    },
+    {
+      "epoch": 0.11938329513352167,
+      "grad_norm": 0.9805975556373596,
+      "learning_rate": 4.231926105340768e-05,
+      "loss": 0.5437,
+      "step": 8959
+    },
+    {
+      "epoch": 0.11979638612014284,
+      "grad_norm": 0.9650508761405945,
+      "learning_rate": 4.225859883654776e-05,
+      "loss": 0.5234,
+      "step": 8990
+    },
+    {
+      "epoch": 0.12020947710676404,
+      "grad_norm": 1.111871600151062,
+      "learning_rate": 4.219774185874569e-05,
+      "loss": 0.5354,
+      "step": 9021
+    },
+    {
+      "epoch": 0.12062256809338522,
+      "grad_norm": 1.1532142162322998,
+      "learning_rate": 4.213669080676418e-05,
+      "loss": 0.5464,
+      "step": 9052
+    },
+    {
+      "epoch": 0.1210356590800064,
+      "grad_norm": 0.9762258529663086,
+      "learning_rate": 4.2075446369556056e-05,
+      "loss": 0.5416,
+      "step": 9083
+    },
+    {
+      "epoch": 0.12144875006662757,
+      "grad_norm": 1.082337498664856,
+      "learning_rate": 4.201400923825648e-05,
+      "loss": 0.5298,
+      "step": 9114
+    },
+    {
+      "epoch": 0.12186184105324877,
+      "grad_norm": 0.9814473390579224,
+      "learning_rate": 4.195238010617511e-05,
+      "loss": 0.5427,
+      "step": 9145
+    },
+    {
+      "epoch": 0.12227493203986994,
+      "grad_norm": 0.9980778694152832,
+      "learning_rate": 4.1890559668788344e-05,
+      "loss": 0.5408,
+      "step": 9176
+    },
+    {
+      "epoch": 0.12268802302649112,
+      "grad_norm": 0.9563232660293579,
+      "learning_rate": 4.1828548623731405e-05,
+      "loss": 0.5407,
+      "step": 9207
+    },
+    {
+      "epoch": 0.12310111401311231,
+      "grad_norm": 1.090119481086731,
+      "learning_rate": 4.1766347670790506e-05,
+      "loss": 0.5377,
+      "step": 9238
+    },
+    {
+      "epoch": 0.12351420499973349,
+      "grad_norm": 1.140044093132019,
+      "learning_rate": 4.170395751189495e-05,
+      "loss": 0.5369,
+      "step": 9269
+    },
+    {
+      "epoch": 0.12392729598635467,
+      "grad_norm": 1.0801435708999634,
+      "learning_rate": 4.164137885110921e-05,
+      "loss": 0.5339,
+      "step": 9300
+    },
+    {
+      "epoch": 0.12434038697297585,
+      "grad_norm": 1.0672699213027954,
+      "learning_rate": 4.157861239462495e-05,
+      "loss": 0.5384,
+      "step": 9331
+    },
+    {
+      "epoch": 0.12475347795959704,
+      "grad_norm": 1.1884500980377197,
+      "learning_rate": 4.1515658850753114e-05,
+      "loss": 0.5382,
+      "step": 9362
+    },
+    {
+      "epoch": 0.12516656894621822,
+      "grad_norm": 0.972347617149353,
+      "learning_rate": 4.145251892991588e-05,
+      "loss": 0.5404,
+      "step": 9393
+    },
+    {
+      "epoch": 0.1255796599328394,
+      "grad_norm": 1.0719892978668213,
+      "learning_rate": 4.138919334463868e-05,
+      "loss": 0.5336,
+      "step": 9424
+    },
+    {
+      "epoch": 0.12599275091946058,
+      "grad_norm": 1.1035733222961426,
+      "learning_rate": 4.1325682809542124e-05,
+      "loss": 0.5389,
+      "step": 9455
+    },
+    {
+      "epoch": 0.12640584190608176,
+      "grad_norm": 0.984730064868927,
+      "learning_rate": 4.126198804133398e-05,
+      "loss": 0.5431,
+      "step": 9486
+    },
+    {
+      "epoch": 0.12681893289270293,
+      "grad_norm": 0.9801311492919922,
+      "learning_rate": 4.1198109758801055e-05,
+      "loss": 0.5359,
+      "step": 9517
+    },
+    {
+      "epoch": 0.12723202387932414,
+      "grad_norm": 1.0692319869995117,
+      "learning_rate": 4.113404868280107e-05,
+      "loss": 0.5438,
+      "step": 9548
+    },
+    {
+      "epoch": 0.12764511486594532,
+      "grad_norm": 1.055517554283142,
+      "learning_rate": 4.106980553625457e-05,
+      "loss": 0.5386,
+      "step": 9579
+    },
+    {
+      "epoch": 0.1280582058525665,
+      "grad_norm": 1.0372021198272705,
+      "learning_rate": 4.100538104413674e-05,
+      "loss": 0.5304,
+      "step": 9610
+    },
+    {
+      "epoch": 0.12847129683918768,
+      "grad_norm": 0.9788863062858582,
+      "learning_rate": 4.09407759334692e-05,
+      "loss": 0.5336,
+      "step": 9641
+    },
+    {
+      "epoch": 0.12888438782580885,
+      "grad_norm": 0.9786235094070435,
+      "learning_rate": 4.087599093331186e-05,
+      "loss": 0.537,
+      "step": 9672
+    },
+    {
+      "epoch": 0.12929747881243003,
+      "grad_norm": 1.1998993158340454,
+      "learning_rate": 4.081102677475462e-05,
+      "loss": 0.5398,
+      "step": 9703
+    },
+    {
+      "epoch": 0.1297105697990512,
+      "grad_norm": 1.1321024894714355,
+      "learning_rate": 4.0745884190909194e-05,
+      "loss": 0.5388,
+      "step": 9734
+    },
+    {
+      "epoch": 0.13012366078567242,
+      "grad_norm": 1.0685521364212036,
+      "learning_rate": 4.0680563916900796e-05,
+      "loss": 0.5408,
+      "step": 9765
+    },
+    {
+      "epoch": 0.1305367517722936,
+      "grad_norm": 1.0620450973510742,
+      "learning_rate": 4.0615066689859815e-05,
+      "loss": 0.5315,
+      "step": 9796
+    },
+    {
+      "epoch": 0.13094984275891478,
+      "grad_norm": 0.9215843081474304,
+      "learning_rate": 4.0549393248913584e-05,
+      "loss": 0.5304,
+      "step": 9827
+    },
+    {
+      "epoch": 0.13136293374553595,
+      "grad_norm": 0.9986915588378906,
+      "learning_rate": 4.048354433517794e-05,
+      "loss": 0.5385,
+      "step": 9858
+    },
+    {
+      "epoch": 0.13177602473215713,
+      "grad_norm": 1.0409338474273682,
+      "learning_rate": 4.0417520691748916e-05,
+      "loss": 0.535,
+      "step": 9889
+    },
+    {
+      "epoch": 0.1321891157187783,
+      "grad_norm": 1.0327932834625244,
+      "learning_rate": 4.035132306369438e-05,
+      "loss": 0.5306,
+      "step": 9920
+    },
+    {
+      "epoch": 0.1326022067053995,
+      "grad_norm": 0.936696469783783,
+      "learning_rate": 4.028495219804555e-05,
+      "loss": 0.5246,
+      "step": 9951
+    },
+    {
+      "epoch": 0.13301529769202067,
+      "grad_norm": 0.9742713570594788,
+      "learning_rate": 4.021840884378864e-05,
+      "loss": 0.534,
+      "step": 9982
+    },
+    {
+      "epoch": 0.13342838867864187,
+      "grad_norm": 1.065794825553894,
+      "learning_rate": 4.015169375185633e-05,
+      "loss": 0.5332,
+      "step": 10013
+    },
+    {
+      "epoch": 0.13384147966526305,
+      "grad_norm": 0.9426642060279846,
+      "learning_rate": 4.0084807675119396e-05,
+      "loss": 0.5347,
+      "step": 10044
+    },
+    {
+      "epoch": 0.13425457065188423,
+      "grad_norm": 1.0314723253250122,
+      "learning_rate": 4.0017751368378106e-05,
+      "loss": 0.5248,
+      "step": 10075
+    },
+    {
+      "epoch": 0.1346676616385054,
+      "grad_norm": 1.100043773651123,
+      "learning_rate": 3.995052558835377e-05,
+      "loss": 0.5313,
+      "step": 10106
+    },
+    {
+      "epoch": 0.1350807526251266,
+      "grad_norm": 0.9944085478782654,
+      "learning_rate": 3.988313109368017e-05,
+      "loss": 0.5334,
+      "step": 10137
+    },
+    {
+      "epoch": 0.13549384361174777,
+      "grad_norm": 0.990739107131958,
+      "learning_rate": 3.981556864489504e-05,
+      "loss": 0.5333,
+      "step": 10168
+    },
+    {
+      "epoch": 0.13590693459836894,
+      "grad_norm": 0.9905965328216553,
+      "learning_rate": 3.974783900443142e-05,
+      "loss": 0.5335,
+      "step": 10199
+    },
+    {
+      "epoch": 0.13632002558499015,
+      "grad_norm": 1.0360978841781616,
+      "learning_rate": 3.9679942936609095e-05,
+      "loss": 0.5356,
+      "step": 10230
+    },
+    {
+      "epoch": 0.13673311657161133,
+      "grad_norm": 0.9169942140579224,
+      "learning_rate": 3.961188120762596e-05,
+      "loss": 0.5378,
+      "step": 10261
+    },
+    {
+      "epoch": 0.1371462075582325,
+      "grad_norm": 0.9882243275642395,
+      "learning_rate": 3.954365458554938e-05,
+      "loss": 0.5335,
+      "step": 10292
+    },
+    {
+      "epoch": 0.1375592985448537,
+      "grad_norm": 1.0667204856872559,
+      "learning_rate": 3.947526384030751e-05,
+      "loss": 0.5356,
+      "step": 10323
+    },
+    {
+      "epoch": 0.13797238953147486,
+      "grad_norm": 1.0076349973678589,
+      "learning_rate": 3.9406709743680624e-05,
+      "loss": 0.5394,
+      "step": 10354
+    },
+    {
+      "epoch": 0.13838548051809604,
+      "grad_norm": 0.9394388794898987,
+      "learning_rate": 3.9337993069292366e-05,
+      "loss": 0.5302,
+      "step": 10385
+    },
+    {
+      "epoch": 0.13879857150471722,
+      "grad_norm": 0.9442508220672607,
+      "learning_rate": 3.926911459260109e-05,
+      "loss": 0.5314,
+      "step": 10416
+    },
+    {
+      "epoch": 0.13921166249133843,
+      "grad_norm": 1.1007356643676758,
+      "learning_rate": 3.920007509089102e-05,
+      "loss": 0.5266,
+      "step": 10447
+    },
+    {
+      "epoch": 0.1396247534779596,
+      "grad_norm": 0.9338844418525696,
+      "learning_rate": 3.913087534326357e-05,
+      "loss": 0.5377,
+      "step": 10478
+    },
+    {
+      "epoch": 0.14003784446458079,
+      "grad_norm": 0.9138127565383911,
+      "learning_rate": 3.9061516130628475e-05,
+      "loss": 0.5336,
+      "step": 10509
+    },
+    {
+      "epoch": 0.14045093545120196,
+      "grad_norm": 1.0976738929748535,
+      "learning_rate": 3.8991998235695025e-05,
+      "loss": 0.5273,
+      "step": 10540
+    },
+    {
+      "epoch": 0.14086402643782314,
+      "grad_norm": 0.9631217122077942,
+      "learning_rate": 3.8922322442963224e-05,
+      "loss": 0.5258,
+      "step": 10571
+    },
+    {
+      "epoch": 0.14127711742444432,
+      "grad_norm": 1.0172392129898071,
+      "learning_rate": 3.885248953871491e-05,
+      "loss": 0.5354,
+      "step": 10602
+    },
+    {
+      "epoch": 0.1416902084110655,
+      "grad_norm": 1.0365219116210938,
+      "learning_rate": 3.8782500311004915e-05,
+      "loss": 0.5294,
+      "step": 10633
+    },
+    {
+      "epoch": 0.14210329939768668,
+      "grad_norm": 1.049512267112732,
+      "learning_rate": 3.871235554965218e-05,
+      "loss": 0.5298,
+      "step": 10664
+    },
+    {
+      "epoch": 0.14251639038430788,
+      "grad_norm": 0.9561368823051453,
+      "learning_rate": 3.864205604623078e-05,
+      "loss": 0.5278,
+      "step": 10695
+    },
+    {
+      "epoch": 0.14292948137092906,
+      "grad_norm": 0.9618587493896484,
+      "learning_rate": 3.857160259406107e-05,
+      "loss": 0.5264,
+      "step": 10726
+    },
+    {
+      "epoch": 0.14334257235755024,
+      "grad_norm": 0.953888475894928,
+      "learning_rate": 3.8500995988200674e-05,
+      "loss": 0.5305,
+      "step": 10757
+    },
+    {
+      "epoch": 0.14375566334417142,
+      "grad_norm": 1.0138510465621948,
+      "learning_rate": 3.843023702543556e-05,
+      "loss": 0.5319,
+      "step": 10788
+    },
+    {
+      "epoch": 0.1441687543307926,
+      "grad_norm": 0.9504718780517578,
+      "learning_rate": 3.8359326504270984e-05,
+      "loss": 0.521,
+      "step": 10819
+    },
+    {
+      "epoch": 0.14458184531741378,
+      "grad_norm": 0.8887687921524048,
+      "learning_rate": 3.828826522492255e-05,
+      "loss": 0.531,
+      "step": 10850
+    },
+    {
+      "epoch": 0.14499493630403495,
+      "grad_norm": 1.0202603340148926,
+      "learning_rate": 3.821705398930713e-05,
+      "loss": 0.5285,
+      "step": 10881
+    },
+    {
+      "epoch": 0.14540802729065616,
+      "grad_norm": 0.9483754634857178,
+      "learning_rate": 3.814569360103385e-05,
+      "loss": 0.5297,
+      "step": 10912
+    },
+    {
+      "epoch": 0.14582111827727734,
+      "grad_norm": 1.0187376737594604,
+      "learning_rate": 3.807418486539499e-05,
+      "loss": 0.5211,
+      "step": 10943
+    },
+    {
+      "epoch": 0.14623420926389852,
+      "grad_norm": 0.9178217053413391,
+      "learning_rate": 3.80025285893569e-05,
+      "loss": 0.5313,
+      "step": 10974
+    },
+    {
+      "epoch": 0.1466473002505197,
+      "grad_norm": 0.9808756709098816,
+      "learning_rate": 3.793072558155093e-05,
+      "loss": 0.5309,
+      "step": 11005
+    },
+    {
+      "epoch": 0.14706039123714088,
+      "grad_norm": 1.000494122505188,
+      "learning_rate": 3.785877665226426e-05,
+      "loss": 0.5264,
+      "step": 11036
+    },
+    {
+      "epoch": 0.14747348222376205,
+      "grad_norm": 0.9617404937744141,
+      "learning_rate": 3.778668261343079e-05,
+      "loss": 0.5173,
+      "step": 11067
+    },
+    {
+      "epoch": 0.14788657321038323,
+      "grad_norm": 0.9080610871315002,
+      "learning_rate": 3.771444427862192e-05,
+      "loss": 0.5365,
+      "step": 11098
+    },
+    {
+      "epoch": 0.1482996641970044,
+      "grad_norm": 0.9060406684875488,
+      "learning_rate": 3.7642062463037465e-05,
+      "loss": 0.5312,
+      "step": 11129
+    },
+    {
+      "epoch": 0.14871275518362562,
+      "grad_norm": 0.9788529276847839,
+      "learning_rate": 3.7569537983496373e-05,
+      "loss": 0.5307,
+      "step": 11160
+    },
+    {
+      "epoch": 0.1491258461702468,
+      "grad_norm": 1.0558335781097412,
+      "learning_rate": 3.749687165842753e-05,
+      "loss": 0.5322,
+      "step": 11191
+    },
+    {
+      "epoch": 0.14953893715686797,
+      "grad_norm": 0.9077445864677429,
+      "learning_rate": 3.7424064307860536e-05,
+      "loss": 0.5329,
+      "step": 11222
+    },
+    {
+      "epoch": 0.14995202814348915,
+      "grad_norm": 0.982660174369812,
+      "learning_rate": 3.735111675341645e-05,
+      "loss": 0.5272,
+      "step": 11253
+    },
+    {
+      "epoch": 0.15036511913011033,
+      "grad_norm": 0.9390408396720886,
+      "learning_rate": 3.7278029818298524e-05,
+      "loss": 0.5247,
+      "step": 11284
+    },
+    {
+      "epoch": 0.1507782101167315,
+      "grad_norm": 0.9156445860862732,
+      "learning_rate": 3.720480432728287e-05,
+      "loss": 0.5289,
+      "step": 11315
+    },
+    {
+      "epoch": 0.1511913011033527,
+      "grad_norm": 0.8932639956474304,
+      "learning_rate": 3.71314411067092e-05,
+      "loss": 0.5318,
+      "step": 11346
+    },
+    {
+      "epoch": 0.1516043920899739,
+      "grad_norm": 0.9631366729736328,
+      "learning_rate": 3.70579409844715e-05,
+      "loss": 0.5273,
+      "step": 11377
+    },
+    {
+      "epoch": 0.15201748307659507,
+      "grad_norm": 0.8670406341552734,
+      "learning_rate": 3.698430479000865e-05,
+      "loss": 0.5269,
+      "step": 11408
+    },
+    {
+      "epoch": 0.15243057406321625,
+      "grad_norm": 1.0833216905593872,
+      "learning_rate": 3.691053335429509e-05,
+      "loss": 0.5321,
+      "step": 11439
+    },
+    {
+      "epoch": 0.15284366504983743,
+      "grad_norm": 1.005530595779419,
+      "learning_rate": 3.683662750983147e-05,
+      "loss": 0.531,
+      "step": 11470
+    },
+    {
+      "epoch": 0.1532567560364586,
+      "grad_norm": 0.9262946844100952,
+      "learning_rate": 3.676258809063518e-05,
+      "loss": 0.5208,
+      "step": 11501
+    },
+    {
+      "epoch": 0.1536698470230798,
+      "grad_norm": 1.003950595855713,
+      "learning_rate": 3.6688415932231004e-05,
+      "loss": 0.5301,
+      "step": 11532
+    },
+    {
+      "epoch": 0.15408293800970096,
+      "grad_norm": 0.9618176221847534,
+      "learning_rate": 3.661411187164166e-05,
+      "loss": 0.5325,
+      "step": 11563
+    },
+    {
+      "epoch": 0.15449602899632217,
+      "grad_norm": 0.927218496799469,
+      "learning_rate": 3.65396767473784e-05,
+      "loss": 0.5256,
+      "step": 11594
+    },
+    {
+      "epoch": 0.15490911998294335,
+      "grad_norm": 2.2212576866149902,
+      "learning_rate": 3.6465111399431465e-05,
+      "loss": 0.5246,
+      "step": 11625
+    },
+    {
+      "epoch": 0.15532221096956453,
+      "grad_norm": 0.9497107863426208,
+      "learning_rate": 3.6390416669260674e-05,
+      "loss": 0.5265,
+      "step": 11656
+    },
+    {
+      "epoch": 0.1557353019561857,
+      "grad_norm": 0.9918030500411987,
+      "learning_rate": 3.63155933997859e-05,
+      "loss": 0.5259,
+      "step": 11687
+    },
+    {
+      "epoch": 0.15614839294280689,
+      "grad_norm": 0.8972867727279663,
+      "learning_rate": 3.624064243537758e-05,
+      "loss": 0.5212,
+      "step": 11718
+    },
+    {
+      "epoch": 0.15656148392942806,
+      "grad_norm": 0.9101853370666504,
+      "learning_rate": 3.616556462184716e-05,
+      "loss": 0.5312,
+      "step": 11749
+    },
+    {
+      "epoch": 0.15697457491604924,
+      "grad_norm": 0.9557723999023438,
+      "learning_rate": 3.609036080643755e-05,
+      "loss": 0.5205,
+      "step": 11780
+    },
+    {
+      "epoch": 0.15738766590267042,
+      "grad_norm": 0.9494668841362,
+      "learning_rate": 3.60150318378136e-05,
+      "loss": 0.5319,
+      "step": 11811
+    },
+    {
+      "epoch": 0.15780075688929163,
+      "grad_norm": 0.9334656596183777,
+      "learning_rate": 3.5939578566052465e-05,
+      "loss": 0.5195,
+      "step": 11842
+    },
+    {
+      "epoch": 0.1582138478759128,
+      "grad_norm": 0.8615605235099792,
+      "learning_rate": 3.586400184263408e-05,
+      "loss": 0.5188,
+      "step": 11873
+    },
+    {
+      "epoch": 0.15862693886253398,
+      "grad_norm": 0.9325541853904724,
+      "learning_rate": 3.578830252043148e-05,
+      "loss": 0.5176,
+      "step": 11904
+    },
+    {
+      "epoch": 0.15904002984915516,
+      "grad_norm": 0.9173383712768555,
+      "learning_rate": 3.571248145370125e-05,
+      "loss": 0.5259,
+      "step": 11935
+    },
+    {
+      "epoch": 0.15945312083577634,
+      "grad_norm": 1.119055151939392,
+      "learning_rate": 3.5636539498073794e-05,
+      "loss": 0.5333,
+      "step": 11966
+    },
+    {
+      "epoch": 0.15986621182239752,
+      "grad_norm": 0.9429402947425842,
+      "learning_rate": 3.556047751054378e-05,
+      "loss": 0.5251,
+      "step": 11997
+    },
+    {
+      "epoch": 0.1602793028090187,
+      "grad_norm": 0.9553139805793762,
+      "learning_rate": 3.548429634946039e-05,
+      "loss": 0.5237,
+      "step": 12028
+    },
+    {
+      "epoch": 0.1606923937956399,
+      "grad_norm": 1.031934380531311,
+      "learning_rate": 3.540799687451768e-05,
+      "loss": 0.5233,
+      "step": 12059
+    },
+    {
+      "epoch": 0.16110548478226108,
+      "grad_norm": 1.0917397737503052,
+      "learning_rate": 3.533157994674485e-05,
+      "loss": 0.523,
+      "step": 12090
+    },
+    {
+      "epoch": 0.16151857576888226,
+      "grad_norm": 1.0041927099227905,
+      "learning_rate": 3.5255046428496546e-05,
+      "loss": 0.5343,
+      "step": 12121
+    },
+    {
+      "epoch": 0.16193166675550344,
+      "grad_norm": 0.9607194066047668,
+      "learning_rate": 3.517839718344311e-05,
+      "loss": 0.5222,
+      "step": 12152
+    },
+    {
+      "epoch": 0.16234475774212462,
+      "grad_norm": 0.9220179915428162,
+      "learning_rate": 3.510163307656086e-05,
+      "loss": 0.5258,
+      "step": 12183
+    },
+    {
+      "epoch": 0.1627578487287458,
+      "grad_norm": 0.9989302754402161,
+      "learning_rate": 3.5024754974122324e-05,
+      "loss": 0.5312,
+      "step": 12214
+    },
+    {
+      "epoch": 0.16317093971536697,
+      "grad_norm": 1.092763900756836,
+      "learning_rate": 3.494776374368643e-05,
+      "loss": 0.5326,
+      "step": 12245
+    },
+    {
+      "epoch": 0.16358403070198815,
+      "grad_norm": 1.0502803325653076,
+      "learning_rate": 3.4870660254088724e-05,
+      "loss": 0.5249,
+      "step": 12276
+    },
+    {
+      "epoch": 0.16399712168860936,
+      "grad_norm": 0.9208263158798218,
+      "learning_rate": 3.479344537543164e-05,
+      "loss": 0.5238,
+      "step": 12307
+    },
+    {
+      "epoch": 0.16441021267523054,
+      "grad_norm": 1.0266058444976807,
+      "learning_rate": 3.4716119979074565e-05,
+      "loss": 0.5279,
+      "step": 12338
+    },
+    {
+      "epoch": 0.16482330366185172,
+      "grad_norm": 0.9023658633232117,
+      "learning_rate": 3.463868493762412e-05,
+      "loss": 0.5222,
+      "step": 12369
+    },
+    {
+      "epoch": 0.1652363946484729,
+      "grad_norm": 0.9667336344718933,
+      "learning_rate": 3.456114112492418e-05,
+      "loss": 0.5289,
+      "step": 12400
+    },
+    {
+      "epoch": 0.16564948563509407,
+      "grad_norm": 1.06847083568573,
+      "learning_rate": 3.4483489416046164e-05,
+      "loss": 0.5269,
+      "step": 12431
+    },
+    {
+      "epoch": 0.16606257662171525,
+      "grad_norm": 0.9444907903671265,
+      "learning_rate": 3.440573068727905e-05,
+      "loss": 0.5299,
+      "step": 12462
+    },
+    {
+      "epoch": 0.16647566760833643,
+      "grad_norm": 1.0396606922149658,
+      "learning_rate": 3.4327865816119495e-05,
+      "loss": 0.5201,
+      "step": 12493
+    },
+    {
+      "epoch": 0.16688875859495764,
+      "grad_norm": 1.003663182258606,
+      "learning_rate": 3.4249895681262025e-05,
+      "loss": 0.5228,
+      "step": 12524
+    },
+    {
+      "epoch": 0.16730184958157882,
+      "grad_norm": 0.9666677713394165,
+      "learning_rate": 3.417182116258899e-05,
+      "loss": 0.5196,
+      "step": 12555
+    },
+    {
+      "epoch": 0.1677149405682,
+      "grad_norm": 0.9915894269943237,
+      "learning_rate": 3.409364314116074e-05,
+      "loss": 0.5243,
+      "step": 12586
+    },
+    {
+      "epoch": 0.16812803155482117,
+      "grad_norm": 1.0233943462371826,
+      "learning_rate": 3.401536249920559e-05,
+      "loss": 0.5115,
+      "step": 12617
+    },
+    {
+      "epoch": 0.16854112254144235,
+      "grad_norm": 0.8652256727218628,
+      "learning_rate": 3.393698012010998e-05,
+      "loss": 0.5267,
+      "step": 12648
+    },
+    {
+      "epoch": 0.16895421352806353,
+      "grad_norm": 0.9623583555221558,
+      "learning_rate": 3.385849688840839e-05,
+      "loss": 0.5184,
+      "step": 12679
+    },
+    {
+      "epoch": 0.1693673045146847,
+      "grad_norm": 1.0796388387680054,
+      "learning_rate": 3.3779913689773414e-05,
+      "loss": 0.5243,
+      "step": 12710
+    },
+    {
+      "epoch": 0.16978039550130591,
+      "grad_norm": 0.8938823938369751,
+      "learning_rate": 3.370123141100578e-05,
+      "loss": 0.5282,
+      "step": 12741
+    },
+    {
+      "epoch": 0.1701934864879271,
+      "grad_norm": 0.9748397469520569,
+      "learning_rate": 3.3622450940024305e-05,
+      "loss": 0.5213,
+      "step": 12772
+    },
+    {
+      "epoch": 0.17060657747454827,
+      "grad_norm": 0.8365583419799805,
+      "learning_rate": 3.35435731658559e-05,
+      "loss": 0.529,
+      "step": 12803
+    },
+    {
+      "epoch": 0.17101966846116945,
+      "grad_norm": 0.9087313413619995,
+      "learning_rate": 3.346459897862552e-05,
+      "loss": 0.5322,
+      "step": 12834
+    },
+    {
+      "epoch": 0.17143275944779063,
+      "grad_norm": 0.9148082733154297,
+      "learning_rate": 3.338552926954613e-05,
+      "loss": 0.5179,
+      "step": 12865
+    },
+    {
+      "epoch": 0.1718458504344118,
+      "grad_norm": 0.9637525677680969,
+      "learning_rate": 3.330636493090868e-05,
+      "loss": 0.5219,
+      "step": 12896
+    },
+    {
+      "epoch": 0.17225894142103299,
+      "grad_norm": 0.9411911964416504,
+      "learning_rate": 3.322710685607193e-05,
+      "loss": 0.5197,
+      "step": 12927
+    },
+    {
+      "epoch": 0.17267203240765416,
+      "grad_norm": 0.9024232029914856,
+      "learning_rate": 3.314775593945251e-05,
+      "loss": 0.5163,
+      "step": 12958
+    },
+    {
+      "epoch": 0.17308512339427537,
+      "grad_norm": 0.8858060240745544,
+      "learning_rate": 3.3068313076514714e-05,
+      "loss": 0.5227,
+      "step": 12989
+    },
+    {
+      "epoch": 0.17349821438089655,
+      "grad_norm": 0.9769457578659058,
+      "learning_rate": 3.298877916376047e-05,
+      "loss": 0.5142,
+      "step": 13020
+    },
+    {
+      "epoch": 0.17391130536751773,
+      "grad_norm": 1.009259581565857,
+      "learning_rate": 3.290915509871915e-05,
+      "loss": 0.5188,
+      "step": 13051
+    },
+    {
+      "epoch": 0.1743243963541389,
+      "grad_norm": 0.9282262325286865,
+      "learning_rate": 3.282944177993753e-05,
+      "loss": 0.5202,
+      "step": 13082
+    },
+    {
+      "epoch": 0.17473748734076008,
+      "grad_norm": 1.1408578157424927,
+      "learning_rate": 3.274964010696957e-05,
+      "loss": 0.5279,
+      "step": 13113
+    },
+    {
+      "epoch": 0.17515057832738126,
+      "grad_norm": 0.8920156359672546,
+      "learning_rate": 3.266975098036629e-05,
+      "loss": 0.5203,
+      "step": 13144
+    },
+    {
+      "epoch": 0.17556366931400244,
+      "grad_norm": 0.9983100295066833,
+      "learning_rate": 3.258977530166562e-05,
+      "loss": 0.5275,
+      "step": 13175
+    },
+    {
+      "epoch": 0.17597676030062365,
+      "grad_norm": 1.017619013786316,
+      "learning_rate": 3.250971397338227e-05,
+      "loss": 0.5191,
+      "step": 13206
+    },
+    {
+      "epoch": 0.17638985128724483,
+      "grad_norm": 1.072635531425476,
+      "learning_rate": 3.2429567898997404e-05,
+      "loss": 0.5206,
+      "step": 13237
+    },
+    {
+      "epoch": 0.176802942273866,
+      "grad_norm": 0.9399890303611755,
+      "learning_rate": 3.234933798294859e-05,
+      "loss": 0.5257,
+      "step": 13268
+    },
+    {
+      "epoch": 0.17721603326048718,
+      "grad_norm": 0.9694241285324097,
+      "learning_rate": 3.2269025130619535e-05,
+      "loss": 0.5248,
+      "step": 13299
+    },
+    {
+      "epoch": 0.17762912424710836,
+      "grad_norm": 0.8554127812385559,
+      "learning_rate": 3.218863024832985e-05,
+      "loss": 0.5139,
+      "step": 13330
+    },
+    {
+      "epoch": 0.17804221523372954,
+      "grad_norm": 0.8834357857704163,
+      "learning_rate": 3.2108154243324864e-05,
+      "loss": 0.5208,
+      "step": 13361
+    },
+    {
+      "epoch": 0.17845530622035072,
+      "grad_norm": 0.8996387720108032,
+      "learning_rate": 3.2027598023765345e-05,
+      "loss": 0.52,
+      "step": 13392
+    },
+    {
+      "epoch": 0.1788683972069719,
+      "grad_norm": 0.9528855085372925,
+      "learning_rate": 3.194696249871729e-05,
+      "loss": 0.5188,
+      "step": 13423
+    },
+    {
+      "epoch": 0.1792814881935931,
+      "grad_norm": 0.8824790120124817,
+      "learning_rate": 3.186624857814164e-05,
+      "loss": 0.5242,
+      "step": 13454
+    },
+    {
+      "epoch": 0.17969457918021428,
+      "grad_norm": 0.881786584854126,
+      "learning_rate": 3.178545717288401e-05,
+      "loss": 0.5306,
+      "step": 13485
+    },
+    {
+      "epoch": 0.18010767016683546,
+      "grad_norm": 0.989400327205658,
+      "learning_rate": 3.170458919466444e-05,
+      "loss": 0.5213,
+      "step": 13516
+    },
+    {
+      "epoch": 0.18052076115345664,
+      "grad_norm": 0.9115697145462036,
+      "learning_rate": 3.1623645556067063e-05,
+      "loss": 0.5204,
+      "step": 13547
+    },
+    {
+      "epoch": 0.18093385214007782,
+      "grad_norm": 0.8908875584602356,
+      "learning_rate": 3.154262717052985e-05,
+      "loss": 0.5138,
+      "step": 13578
+    },
+    {
+      "epoch": 0.181346943126699,
+      "grad_norm": 0.9119309782981873,
+      "learning_rate": 3.146153495233426e-05,
+      "loss": 0.5209,
+      "step": 13609
+    },
+    {
+      "epoch": 0.18176003411332017,
+      "grad_norm": 0.8793107867240906,
+      "learning_rate": 3.1380369816594944e-05,
+      "loss": 0.5143,
+      "step": 13640
+    },
+    {
+      "epoch": 0.18217312509994138,
+      "grad_norm": 0.9101561307907104,
+      "learning_rate": 3.129913267924946e-05,
+      "loss": 0.5125,
+      "step": 13671
+    },
+    {
+      "epoch": 0.18258621608656256,
+      "grad_norm": 0.9199904203414917,
+      "learning_rate": 3.121782445704782e-05,
+      "loss": 0.5192,
+      "step": 13702
+    },
+    {
+      "epoch": 0.18299930707318374,
+      "grad_norm": 0.9675397276878357,
+      "learning_rate": 3.11364460675423e-05,
+      "loss": 0.5162,
+      "step": 13733
+    },
+    {
+      "epoch": 0.18341239805980492,
+      "grad_norm": 0.9324703216552734,
+      "learning_rate": 3.1054998429076934e-05,
+      "loss": 0.5227,
+      "step": 13764
+    },
+    {
+      "epoch": 0.1838254890464261,
+      "grad_norm": 1.0210453271865845,
+      "learning_rate": 3.097348246077728e-05,
+      "loss": 0.5162,
+      "step": 13795
+    },
+    {
+      "epoch": 0.18423858003304727,
+      "grad_norm": 0.91178959608078,
+      "learning_rate": 3.0891899082539924e-05,
+      "loss": 0.5121,
+      "step": 13826
+    },
+    {
+      "epoch": 0.18465167101966845,
+      "grad_norm": 0.9580845832824707,
+      "learning_rate": 3.0810249215022233e-05,
+      "loss": 0.5201,
+      "step": 13857
+    },
+    {
+      "epoch": 0.18506476200628966,
+      "grad_norm": 0.9808583855628967,
+      "learning_rate": 3.0728533779631865e-05,
+      "loss": 0.5218,
+      "step": 13888
+    },
+    {
+      "epoch": 0.18547785299291084,
+      "grad_norm": 1.2176625728607178,
+      "learning_rate": 3.064675369851637e-05,
+      "loss": 0.5187,
+      "step": 13919
+    },
+    {
+      "epoch": 0.18589094397953201,
+      "grad_norm": 0.9023287296295166,
+      "learning_rate": 3.056490989455289e-05,
+      "loss": 0.5296,
+      "step": 13950
+    },
+    {
+      "epoch": 0.1863040349661532,
+      "grad_norm": 0.9504340291023254,
+      "learning_rate": 3.0483003291337596e-05,
+      "loss": 0.5198,
+      "step": 13981
+    },
+    {
+      "epoch": 0.18671712595277437,
+      "grad_norm": 0.942531168460846,
+      "learning_rate": 3.040103481317539e-05,
+      "loss": 0.5181,
+      "step": 14012
+    },
+    {
+      "epoch": 0.18713021693939555,
+      "grad_norm": 0.975398600101471,
+      "learning_rate": 3.03190053850694e-05,
+      "loss": 0.5206,
+      "step": 14043
+    },
+    {
+      "epoch": 0.18754330792601673,
+      "grad_norm": 0.9098790884017944,
+      "learning_rate": 3.0236915932710573e-05,
+      "loss": 0.5092,
+      "step": 14074
+    },
+    {
+      "epoch": 0.1879563989126379,
+      "grad_norm": 0.9327256083488464,
+      "learning_rate": 3.0154767382467232e-05,
+      "loss": 0.5243,
+      "step": 14105
+    },
+    {
+      "epoch": 0.1883694898992591,
+      "grad_norm": 0.9311143159866333,
+      "learning_rate": 3.0072560661374582e-05,
+      "loss": 0.5115,
+      "step": 14136
+    },
+    {
+      "epoch": 0.1887825808858803,
+      "grad_norm": 0.9428762793540955,
+      "learning_rate": 2.999029669712431e-05,
+      "loss": 0.5212,
+      "step": 14167
+    },
+    {
+      "epoch": 0.18919567187250147,
+      "grad_norm": 0.866916835308075,
+      "learning_rate": 2.990797641805408e-05,
+      "loss": 0.5067,
+      "step": 14198
+    },
+    {
+      "epoch": 0.18960876285912265,
+      "grad_norm": 0.9511030316352844,
+      "learning_rate": 2.982560075313704e-05,
+      "loss": 0.5181,
+      "step": 14229
+    },
+    {
+      "epoch": 0.19002185384574383,
+      "grad_norm": 1.0162817239761353,
+      "learning_rate": 2.9743170631971368e-05,
+      "loss": 0.5236,
+      "step": 14260
+    },
+    {
+      "epoch": 0.190434944832365,
+      "grad_norm": 0.9300915002822876,
+      "learning_rate": 2.9660686984769792e-05,
+      "loss": 0.5079,
+      "step": 14291
+    },
+    {
+      "epoch": 0.19084803581898618,
+      "grad_norm": 0.9040572047233582,
+      "learning_rate": 2.9578150742349047e-05,
+      "loss": 0.5221,
+      "step": 14322
+    },
+    {
+      "epoch": 0.1912611268056074,
+      "grad_norm": 0.8688935041427612,
+      "learning_rate": 2.949556283611942e-05,
+      "loss": 0.5169,
+      "step": 14353
+    },
+    {
+      "epoch": 0.19167421779222857,
+      "grad_norm": 1.0615109205245972,
+      "learning_rate": 2.9412924198074206e-05,
+      "loss": 0.5124,
+      "step": 14384
+    },
+    {
+      "epoch": 0.19208730877884975,
+      "grad_norm": 0.8621201515197754,
+      "learning_rate": 2.9330235760779208e-05,
+      "loss": 0.5104,
+      "step": 14415
+    },
+    {
+      "epoch": 0.19250039976547093,
+      "grad_norm": 0.938222348690033,
+      "learning_rate": 2.9247498457362188e-05,
+      "loss": 0.5172,
+      "step": 14446
+    },
+    {
+      "epoch": 0.1929134907520921,
+      "grad_norm": 0.9711362719535828,
+      "learning_rate": 2.9164713221502373e-05,
+      "loss": 0.5205,
+      "step": 14477
+    },
+    {
+      "epoch": 0.19332658173871328,
+      "grad_norm": 0.9693037271499634,
+      "learning_rate": 2.9081880987419912e-05,
+      "loss": 0.5185,
+      "step": 14508
+    },
+    {
+      "epoch": 0.19373967272533446,
+      "grad_norm": 0.9850576519966125,
+      "learning_rate": 2.8999002689865296e-05,
+      "loss": 0.5213,
+      "step": 14539
+    },
+    {
+      "epoch": 0.19415276371195564,
+      "grad_norm": 0.9740920662879944,
+      "learning_rate": 2.8916079264108852e-05,
+      "loss": 0.5211,
+      "step": 14570
+    },
+    {
+      "epoch": 0.19456585469857685,
+      "grad_norm": 0.8955512642860413,
+      "learning_rate": 2.883311164593017e-05,
+      "loss": 0.518,
+      "step": 14601
+    },
+    {
+      "epoch": 0.19497894568519802,
+      "grad_norm": 0.9833427667617798,
+      "learning_rate": 2.875010077160754e-05,
+      "loss": 0.5093,
+      "step": 14632
+    },
+    {
+      "epoch": 0.1953920366718192,
+      "grad_norm": 0.7993410229682922,
+      "learning_rate": 2.866704757790741e-05,
+      "loss": 0.5212,
+      "step": 14663
+    },
+    {
+      "epoch": 0.19580512765844038,
+      "grad_norm": 0.8917197585105896,
+      "learning_rate": 2.858395300207376e-05,
+      "loss": 0.5097,
+      "step": 14694
+    },
+    {
+      "epoch": 0.19621821864506156,
+      "grad_norm": 0.9120102524757385,
+      "learning_rate": 2.8500817981817607e-05,
+      "loss": 0.513,
+      "step": 14725
+    },
+    {
+      "epoch": 0.19663130963168274,
+      "grad_norm": 0.8676596283912659,
+      "learning_rate": 2.8417643455306336e-05,
+      "loss": 0.5176,
+      "step": 14756
+    },
+    {
+      "epoch": 0.19704440061830392,
+      "grad_norm": 1.0420931577682495,
+      "learning_rate": 2.8334430361153185e-05,
+      "loss": 0.5152,
+      "step": 14787
+    },
+    {
+      "epoch": 0.19745749160492512,
+      "grad_norm": 0.8551843762397766,
+      "learning_rate": 2.8251179638406612e-05,
+      "loss": 0.5129,
+      "step": 14818
+    },
+    {
+      "epoch": 0.1978705825915463,
+      "grad_norm": 0.925971508026123,
+      "learning_rate": 2.8167892226539704e-05,
+      "loss": 0.5167,
+      "step": 14849
+    },
+    {
+      "epoch": 0.19828367357816748,
+      "grad_norm": 0.9412753582000732,
+      "learning_rate": 2.8084569065439588e-05,
+      "loss": 0.5206,
+      "step": 14880
+    },
+    {
+      "epoch": 0.19869676456478866,
+      "grad_norm": 0.8888112306594849,
+      "learning_rate": 2.8001211095396807e-05,
+      "loss": 0.513,
+      "step": 14911
+    },
+    {
+      "epoch": 0.19910985555140984,
+      "grad_norm": 0.9865559339523315,
+      "learning_rate": 2.791781925709473e-05,
+      "loss": 0.5219,
+      "step": 14942
+    },
+    {
+      "epoch": 0.19952294653803102,
+      "grad_norm": 1.011573314666748,
+      "learning_rate": 2.7834394491598908e-05,
+      "loss": 0.5174,
+      "step": 14973
+    },
+    {
+      "epoch": 0.1999360375246522,
+      "grad_norm": 1.0012012720108032,
+      "learning_rate": 2.7750937740346485e-05,
+      "loss": 0.5174,
+      "step": 15004
+    },
+    {
+      "epoch": 0.2003491285112734,
+      "grad_norm": 0.930383026599884,
+      "learning_rate": 2.7667449945135564e-05,
+      "loss": 0.5131,
+      "step": 15035
+    },
+    {
+      "epoch": 0.20076221949789458,
+      "grad_norm": 0.94016033411026,
+      "learning_rate": 2.7583932048114557e-05,
+      "loss": 0.5227,
+      "step": 15066
+    },
+    {
+      "epoch": 0.20117531048451576,
+      "grad_norm": 0.9186452031135559,
+      "learning_rate": 2.7500384991771587e-05,
+      "loss": 0.5147,
+      "step": 15097
+    },
+    {
+      "epoch": 0.20158840147113694,
+      "grad_norm": 0.9641884565353394,
+      "learning_rate": 2.7416809718923825e-05,
+      "loss": 0.5139,
+      "step": 15128
+    },
+    {
+      "epoch": 0.20200149245775811,
+      "grad_norm": 0.929161012172699,
+      "learning_rate": 2.7333207172706864e-05,
+      "loss": 0.5141,
+      "step": 15159
+    },
+    {
+      "epoch": 0.2024145834443793,
+      "grad_norm": 0.9425679445266724,
+      "learning_rate": 2.7249578296564088e-05,
+      "loss": 0.52,
+      "step": 15190
+    },
+    {
+      "epoch": 0.20282767443100047,
+      "grad_norm": 0.9315940737724304,
+      "learning_rate": 2.7165924034235973e-05,
+      "loss": 0.5177,
+      "step": 15221
+    },
+    {
+      "epoch": 0.20324076541762165,
+      "grad_norm": 0.8574975728988647,
+      "learning_rate": 2.708224532974953e-05,
+      "loss": 0.5089,
+      "step": 15252
+    }
+  ],
+  "logging_steps": 31,
+  "max_steps": 30517,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 1,
+  "save_steps": 3052,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.1258280394232955e+19,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}

checkpoint-15260/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1fc4955674ee351f186535908caf566dac206e79a7c64f521967d1c769a6cc33
+size 5304

checkpoint-18312/config.json

@@ -0,0 +1,39 @@
+{
+  "_name_or_path": "meta-llama/Llama-3.1-8B-Instruct",
+  "architectures": [
+    "LlamaForCausalLM"
+  ],
+  "attention_bias": false,
+  "attention_dropout": 0.0,
+  "bos_token_id": 128000,
+  "eos_token_id": [
+    128001,
+    128008,
+    128009
+  ],
+  "hidden_act": "silu",
+  "hidden_size": 4096,
+  "initializer_range": 0.02,
+  "intermediate_size": 14336,
+  "max_position_embeddings": 131072,
+  "mlp_bias": false,
+  "model_type": "llama",
+  "num_attention_heads": 32,
+  "num_hidden_layers": 32,
+  "num_key_value_heads": 8,
+  "pretraining_tp": 1,
+  "rms_norm_eps": 1e-05,
+  "rope_scaling": {
+    "factor": 8.0,
+    "high_freq_factor": 4.0,
+    "low_freq_factor": 1.0,
+    "original_max_position_embeddings": 8192,
+    "rope_type": "llama3"
+  },
+  "rope_theta": 500000.0,
+  "tie_word_embeddings": false,
+  "torch_dtype": "float32",
+  "transformers_version": "4.44.0.dev0",
+  "use_cache": true,
+  "vocab_size": 128256
+}

checkpoint-18312/generation_config.json

@@ -0,0 +1,12 @@
+{
+  "bos_token_id": 128000,
+  "do_sample": true,
+  "eos_token_id": [
+    128001,
+    128008,
+    128009
+  ],
+  "temperature": 0.6,
+  "top_p": 0.9,
+  "transformers_version": "4.44.0.dev0"
+}

checkpoint-18312/model-00001-of-00007.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:57804fc6cb0fb2165a1a170c1b611ca542b4890ab2d41d6d2e5f4ade814473f4
+size 4886466168

checkpoint-18312/model-00002-of-00007.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:23fa5aaf8ac56c17c821980c9aee44fcbc2a617fc570e875e6fdb93d92886c72
+size 4832007448

checkpoint-18312/model-00003-of-00007.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1a223e87240e50dfb05634ea909cf1ec715d37a12b681b14a3a6e97f28688f2b
+size 4999813112

checkpoint-18312/model-00004-of-00007.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3047d199c611c6d12b1cca780f033ad89df9c076f742674f7ac4a892ca9573b3
+size 4999813128

checkpoint-18312/model-00005-of-00007.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0a2f91d749d664ecd22c71f3d54488966fc6899a03924b8fea3613c8f83398a8
+size 4832007496

checkpoint-18312/model-00006-of-00007.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c48110ace1e314a877b491fa1f8795fbe816ef54ce8dc4da3d229e27d86ee7dd
+size 4999813120

checkpoint-18312/model-00007-of-00007.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2a4c84c2a739527ab7e81b62873ad8bf2ab254b795897450e550730dea1d6142
+size 2571158184

checkpoint-18312/model.safetensors.index.json

@@ -0,0 +1,298 @@
+{
+  "metadata": {
+    "total_size": 32121044992
+  },
+  "weight_map": {
+    "lm_head.weight": "model-00007-of-00007.safetensors",
+    "model.embed_tokens.weight": "model-00001-of-00007.safetensors",
+    "model.layers.0.input_layernorm.weight": "model-00001-of-00007.safetensors",
+    "model.layers.0.mlp.down_proj.weight": "model-00001-of-00007.safetensors",
+    "model.layers.0.mlp.gate_proj.weight": "model-00001-of-00007.safetensors",
+    "model.layers.0.mlp.up_proj.weight": "model-00001-of-00007.safetensors",
+    "model.layers.0.post_attention_layernorm.weight": "model-00001-of-00007.safetensors",
+    "model.layers.0.self_attn.k_proj.weight": "model-00001-of-00007.safetensors",
+    "model.layers.0.self_attn.o_proj.weight": "model-00001-of-00007.safetensors",
+    "model.layers.0.self_attn.q_proj.weight": "model-00001-of-00007.safetensors",
+    "model.layers.0.self_attn.v_proj.weight": "model-00001-of-00007.safetensors",
+    "model.layers.1.input_layernorm.weight": "model-00001-of-00007.safetensors",
+    "model.layers.1.mlp.down_proj.weight": "model-00001-of-00007.safetensors",
+    "model.layers.1.mlp.gate_proj.weight": "model-00001-of-00007.safetensors",
+    "model.layers.1.mlp.up_proj.weight": "model-00001-of-00007.safetensors",
+    "model.layers.1.post_attention_layernorm.weight": "model-00001-of-00007.safetensors",
+    "model.layers.1.self_attn.k_proj.weight": "model-00001-of-00007.safetensors",
+    "model.layers.1.self_attn.o_proj.weight": "model-00001-of-00007.safetensors",
+    "model.layers.1.self_attn.q_proj.weight": "model-00001-of-00007.safetensors",
+    "model.layers.1.self_attn.v_proj.weight": "model-00001-of-00007.safetensors",
+    "model.layers.10.input_layernorm.weight": "model-00003-of-00007.safetensors",
+    "model.layers.10.mlp.down_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.10.mlp.gate_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.10.mlp.up_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.10.post_attention_layernorm.weight": "model-00003-of-00007.safetensors",
+    "model.layers.10.self_attn.k_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.10.self_attn.o_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.10.self_attn.q_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.10.self_attn.v_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.11.input_layernorm.weight": "model-00003-of-00007.safetensors",
+    "model.layers.11.mlp.down_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.11.mlp.gate_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.11.mlp.up_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.11.post_attention_layernorm.weight": "model-00003-of-00007.safetensors",
+    "model.layers.11.self_attn.k_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.11.self_attn.o_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.11.self_attn.q_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.11.self_attn.v_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.12.input_layernorm.weight": "model-00003-of-00007.safetensors",
+    "model.layers.12.mlp.down_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.12.mlp.gate_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.12.mlp.up_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.12.post_attention_layernorm.weight": "model-00003-of-00007.safetensors",
+    "model.layers.12.self_attn.k_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.12.self_attn.o_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.12.self_attn.q_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.12.self_attn.v_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.13.input_layernorm.weight": "model-00003-of-00007.safetensors",
+    "model.layers.13.mlp.down_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.13.mlp.gate_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.13.mlp.up_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.13.post_attention_layernorm.weight": "model-00003-of-00007.safetensors",
+    "model.layers.13.self_attn.k_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.13.self_attn.o_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.13.self_attn.q_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.13.self_attn.v_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.14.input_layernorm.weight": "model-00004-of-00007.safetensors",
+    "model.layers.14.mlp.down_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.14.mlp.gate_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.14.mlp.up_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.14.post_attention_layernorm.weight": "model-00004-of-00007.safetensors",
+    "model.layers.14.self_attn.k_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.14.self_attn.o_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.14.self_attn.q_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.14.self_attn.v_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.15.input_layernorm.weight": "model-00004-of-00007.safetensors",
+    "model.layers.15.mlp.down_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.15.mlp.gate_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.15.mlp.up_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.15.post_attention_layernorm.weight": "model-00004-of-00007.safetensors",
+    "model.layers.15.self_attn.k_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.15.self_attn.o_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.15.self_attn.q_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.15.self_attn.v_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.16.input_layernorm.weight": "model-00004-of-00007.safetensors",
+    "model.layers.16.mlp.down_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.16.mlp.gate_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.16.mlp.up_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.16.post_attention_layernorm.weight": "model-00004-of-00007.safetensors",
+    "model.layers.16.self_attn.k_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.16.self_attn.o_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.16.self_attn.q_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.16.self_attn.v_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.17.input_layernorm.weight": "model-00004-of-00007.safetensors",
+    "model.layers.17.mlp.down_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.17.mlp.gate_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.17.mlp.up_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.17.post_attention_layernorm.weight": "model-00004-of-00007.safetensors",
+    "model.layers.17.self_attn.k_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.17.self_attn.o_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.17.self_attn.q_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.17.self_attn.v_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.18.input_layernorm.weight": "model-00004-of-00007.safetensors",
+    "model.layers.18.mlp.down_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.18.mlp.gate_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.18.mlp.up_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.18.post_attention_layernorm.weight": "model-00004-of-00007.safetensors",
+    "model.layers.18.self_attn.k_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.18.self_attn.o_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.18.self_attn.q_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.18.self_attn.v_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.19.input_layernorm.weight": "model-00004-of-00007.safetensors",
+    "model.layers.19.mlp.down_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.19.mlp.gate_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.19.mlp.up_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.19.post_attention_layernorm.weight": "model-00004-of-00007.safetensors",
+    "model.layers.19.self_attn.k_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.19.self_attn.o_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.19.self_attn.q_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.19.self_attn.v_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.2.input_layernorm.weight": "model-00001-of-00007.safetensors",
+    "model.layers.2.mlp.down_proj.weight": "model-00001-of-00007.safetensors",
+    "model.layers.2.mlp.gate_proj.weight": "model-00001-of-00007.safetensors",
+    "model.layers.2.mlp.up_proj.weight": "model-00001-of-00007.safetensors",
+    "model.layers.2.post_attention_layernorm.weight": "model-00001-of-00007.safetensors",
+    "model.layers.2.self_attn.k_proj.weight": "model-00001-of-00007.safetensors",
+    "model.layers.2.self_attn.o_proj.weight": "model-00001-of-00007.safetensors",
+    "model.layers.2.self_attn.q_proj.weight": "model-00001-of-00007.safetensors",
+    "model.layers.2.self_attn.v_proj.weight": "model-00001-of-00007.safetensors",
+    "model.layers.20.input_layernorm.weight": "model-00005-of-00007.safetensors",
+    "model.layers.20.mlp.down_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.20.mlp.gate_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.20.mlp.up_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.20.post_attention_layernorm.weight": "model-00005-of-00007.safetensors",
+    "model.layers.20.self_attn.k_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.20.self_attn.o_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.20.self_attn.q_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.20.self_attn.v_proj.weight": "model-00004-of-00007.safetensors",
+    "model.layers.21.input_layernorm.weight": "model-00005-of-00007.safetensors",
+    "model.layers.21.mlp.down_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.21.mlp.gate_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.21.mlp.up_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.21.post_attention_layernorm.weight": "model-00005-of-00007.safetensors",
+    "model.layers.21.self_attn.k_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.21.self_attn.o_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.21.self_attn.q_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.21.self_attn.v_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.22.input_layernorm.weight": "model-00005-of-00007.safetensors",
+    "model.layers.22.mlp.down_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.22.mlp.gate_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.22.mlp.up_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.22.post_attention_layernorm.weight": "model-00005-of-00007.safetensors",
+    "model.layers.22.self_attn.k_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.22.self_attn.o_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.22.self_attn.q_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.22.self_attn.v_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.23.input_layernorm.weight": "model-00005-of-00007.safetensors",
+    "model.layers.23.mlp.down_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.23.mlp.gate_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.23.mlp.up_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.23.post_attention_layernorm.weight": "model-00005-of-00007.safetensors",
+    "model.layers.23.self_attn.k_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.23.self_attn.o_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.23.self_attn.q_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.23.self_attn.v_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.24.input_layernorm.weight": "model-00005-of-00007.safetensors",
+    "model.layers.24.mlp.down_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.24.mlp.gate_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.24.mlp.up_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.24.post_attention_layernorm.weight": "model-00005-of-00007.safetensors",
+    "model.layers.24.self_attn.k_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.24.self_attn.o_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.24.self_attn.q_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.24.self_attn.v_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.25.input_layernorm.weight": "model-00006-of-00007.safetensors",
+    "model.layers.25.mlp.down_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.25.mlp.gate_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.25.mlp.up_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.25.post_attention_layernorm.weight": "model-00006-of-00007.safetensors",
+    "model.layers.25.self_attn.k_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.25.self_attn.o_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.25.self_attn.q_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.25.self_attn.v_proj.weight": "model-00005-of-00007.safetensors",
+    "model.layers.26.input_layernorm.weight": "model-00006-of-00007.safetensors",
+    "model.layers.26.mlp.down_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.26.mlp.gate_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.26.mlp.up_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.26.post_attention_layernorm.weight": "model-00006-of-00007.safetensors",
+    "model.layers.26.self_attn.k_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.26.self_attn.o_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.26.self_attn.q_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.26.self_attn.v_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.27.input_layernorm.weight": "model-00006-of-00007.safetensors",
+    "model.layers.27.mlp.down_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.27.mlp.gate_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.27.mlp.up_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.27.post_attention_layernorm.weight": "model-00006-of-00007.safetensors",
+    "model.layers.27.self_attn.k_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.27.self_attn.o_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.27.self_attn.q_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.27.self_attn.v_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.28.input_layernorm.weight": "model-00006-of-00007.safetensors",
+    "model.layers.28.mlp.down_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.28.mlp.gate_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.28.mlp.up_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.28.post_attention_layernorm.weight": "model-00006-of-00007.safetensors",
+    "model.layers.28.self_attn.k_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.28.self_attn.o_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.28.self_attn.q_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.28.self_attn.v_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.29.input_layernorm.weight": "model-00006-of-00007.safetensors",
+    "model.layers.29.mlp.down_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.29.mlp.gate_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.29.mlp.up_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.29.post_attention_layernorm.weight": "model-00006-of-00007.safetensors",
+    "model.layers.29.self_attn.k_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.29.self_attn.o_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.29.self_attn.q_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.29.self_attn.v_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.3.input_layernorm.weight": "model-00002-of-00007.safetensors",
+    "model.layers.3.mlp.down_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.3.mlp.gate_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.3.mlp.up_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.3.post_attention_layernorm.weight": "model-00002-of-00007.safetensors",
+    "model.layers.3.self_attn.k_proj.weight": "model-00001-of-00007.safetensors",
+    "model.layers.3.self_attn.o_proj.weight": "model-00001-of-00007.safetensors",
+    "model.layers.3.self_attn.q_proj.weight": "model-00001-of-00007.safetensors",
+    "model.layers.3.self_attn.v_proj.weight": "model-00001-of-00007.safetensors",
+    "model.layers.30.input_layernorm.weight": "model-00006-of-00007.safetensors",
+    "model.layers.30.mlp.down_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.30.mlp.gate_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.30.mlp.up_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.30.post_attention_layernorm.weight": "model-00006-of-00007.safetensors",
+    "model.layers.30.self_attn.k_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.30.self_attn.o_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.30.self_attn.q_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.30.self_attn.v_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.31.input_layernorm.weight": "model-00007-of-00007.safetensors",
+    "model.layers.31.mlp.down_proj.weight": "model-00007-of-00007.safetensors",
+    "model.layers.31.mlp.gate_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.31.mlp.up_proj.weight": "model-00007-of-00007.safetensors",
+    "model.layers.31.post_attention_layernorm.weight": "model-00007-of-00007.safetensors",
+    "model.layers.31.self_attn.k_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.31.self_attn.o_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.31.self_attn.q_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.31.self_attn.v_proj.weight": "model-00006-of-00007.safetensors",
+    "model.layers.4.input_layernorm.weight": "model-00002-of-00007.safetensors",
+    "model.layers.4.mlp.down_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.4.mlp.gate_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.4.mlp.up_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.4.post_attention_layernorm.weight": "model-00002-of-00007.safetensors",
+    "model.layers.4.self_attn.k_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.4.self_attn.o_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.4.self_attn.q_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.4.self_attn.v_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.5.input_layernorm.weight": "model-00002-of-00007.safetensors",
+    "model.layers.5.mlp.down_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.5.mlp.gate_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.5.mlp.up_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.5.post_attention_layernorm.weight": "model-00002-of-00007.safetensors",
+    "model.layers.5.self_attn.k_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.5.self_attn.o_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.5.self_attn.q_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.5.self_attn.v_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.6.input_layernorm.weight": "model-00002-of-00007.safetensors",
+    "model.layers.6.mlp.down_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.6.mlp.gate_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.6.mlp.up_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.6.post_attention_layernorm.weight": "model-00002-of-00007.safetensors",
+    "model.layers.6.self_attn.k_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.6.self_attn.o_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.6.self_attn.q_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.6.self_attn.v_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.7.input_layernorm.weight": "model-00002-of-00007.safetensors",
+    "model.layers.7.mlp.down_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.7.mlp.gate_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.7.mlp.up_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.7.post_attention_layernorm.weight": "model-00002-of-00007.safetensors",
+    "model.layers.7.self_attn.k_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.7.self_attn.o_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.7.self_attn.q_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.7.self_attn.v_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.8.input_layernorm.weight": "model-00003-of-00007.safetensors",
+    "model.layers.8.mlp.down_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.8.mlp.gate_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.8.mlp.up_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.8.post_attention_layernorm.weight": "model-00003-of-00007.safetensors",
+    "model.layers.8.self_attn.k_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.8.self_attn.o_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.8.self_attn.q_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.8.self_attn.v_proj.weight": "model-00002-of-00007.safetensors",
+    "model.layers.9.input_layernorm.weight": "model-00003-of-00007.safetensors",
+    "model.layers.9.mlp.down_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.9.mlp.gate_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.9.mlp.up_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.9.post_attention_layernorm.weight": "model-00003-of-00007.safetensors",
+    "model.layers.9.self_attn.k_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.9.self_attn.o_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.9.self_attn.q_proj.weight": "model-00003-of-00007.safetensors",
+    "model.layers.9.self_attn.v_proj.weight": "model-00003-of-00007.safetensors",
+    "model.norm.weight": "model-00007-of-00007.safetensors"
+  }
+}

checkpoint-18312/optimizer.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:80f3cc02b34106550db9d9b377c020d4dc55884eb1000e932bb097c8285e1a0b
+size 15385036334

checkpoint-18312/rng_state.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9d9cd6a0487226e5bd30d1846894c82af483733ab4381b75bae9c0745e05d405
+size 14244

checkpoint-18312/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a0e65c3d6f29e706fd941a38280ce5628189a6998eac6d29abbeab00ad838d00
+size 1064

checkpoint-18312/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1fc4955674ee351f186535908caf566dac206e79a7c64f521967d1c769a6cc33
+size 5304

checkpoint-21364/config.json

@@ -0,0 +1,39 @@
+{
+  "_name_or_path": "meta-llama/Llama-3.1-8B-Instruct",
+  "architectures": [
+    "LlamaForCausalLM"
+  ],
+  "attention_bias": false,
+  "attention_dropout": 0.0,
+  "bos_token_id": 128000,
+  "eos_token_id": [
+    128001,
+    128008,
+    128009
+  ],
+  "hidden_act": "silu",
+  "hidden_size": 4096,
+  "initializer_range": 0.02,
+  "intermediate_size": 14336,
+  "max_position_embeddings": 131072,
+  "mlp_bias": false,
+  "model_type": "llama",
+  "num_attention_heads": 32,
+  "num_hidden_layers": 32,
+  "num_key_value_heads": 8,
+  "pretraining_tp": 1,
+  "rms_norm_eps": 1e-05,
+  "rope_scaling": {
+    "factor": 8.0,
+    "high_freq_factor": 4.0,
+    "low_freq_factor": 1.0,
+    "original_max_position_embeddings": 8192,
+    "rope_type": "llama3"
+  },
+  "rope_theta": 500000.0,
+  "tie_word_embeddings": false,
+  "torch_dtype": "float32",
+  "transformers_version": "4.44.0.dev0",
+  "use_cache": true,
+  "vocab_size": 128256
+}

checkpoint-21364/generation_config.json

@@ -0,0 +1,12 @@
+{
+  "bos_token_id": 128000,
+  "do_sample": true,
+  "eos_token_id": [
+    128001,
+    128008,
+    128009
+  ],
+  "temperature": 0.6,
+  "top_p": 0.9,
+  "transformers_version": "4.44.0.dev0"
+}

checkpoint-21364/model-00001-of-00007.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2edcb4b9497154f995a13756e00da18711a5c5e4a04bc6826220075b9b3db8f0
+size 4886466168

checkpoint-21364/model-00002-of-00007.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:23fa5aaf8ac56c17c821980c9aee44fcbc2a617fc570e875e6fdb93d92886c72
+size 4832007448

checkpoint-21364/model-00003-of-00007.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1a223e87240e50dfb05634ea909cf1ec715d37a12b681b14a3a6e97f28688f2b
+size 4999813112