Upload folder using huggingface_hub

checkpoint-12208/config.json

@@ -0,0 +1,39 @@
+{
+  "_name_or_path": "/mnt/parscratch/users/acp23ay/private/models/Llama-3.1-8B-Instruct-gu-madlad-mean/",
+  "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": 138256
+}

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 sha256:72a6dddb27fe56829f663c8c1d729f7b42a4737924c9abce5e3efc06de51083b
+size 4983197184

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

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:168629c67732309d49f60a5ec48a6d160212bc987365e82e183dfbf74ba0c1f3
+size 4899116432

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:f41cb4e8ad68d24cc50e434c9c5f3d17ec7bc1e9102bf9df21c83b21b51a8dd1
+size 4999813120

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

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:06ec8c367476f42f18cc746d3b7e54f5eceb1c1a820b505d99df1abb57161b0f
+size 2734998184

checkpoint-12208/model.safetensors.index.json

@@ -0,0 +1,298 @@
+{
+  "metadata": {
+    "total_size": 32448724992
+  },
+  "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-00002-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 sha256:2b3da7833ce8444c4142310e5c980b7533ef3d8f332db1d799290607f63377ce
+size 16040396334

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": 0.520275309510111,
+  "eval_steps": 500,
+  "global_step": 12208,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.001321144707963093,
+      "grad_norm": 29.45369529724121,
+      "learning_rate": 1.0157273918741808e-06,
+      "loss": 9.356,
+      "step": 31
+    },
+    {
+      "epoch": 0.002642289415926186,
+      "grad_norm": 17.081205368041992,
+      "learning_rate": 2.0314547837483616e-06,
+      "loss": 7.7447,
+      "step": 62
+    },
+    {
+      "epoch": 0.00396343412388928,
+      "grad_norm": 17.875776290893555,
+      "learning_rate": 3.0471821756225426e-06,
+      "loss": 6.3787,
+      "step": 93
+    },
+    {
+      "epoch": 0.005284578831852372,
+      "grad_norm": 38.05577087402344,
+      "learning_rate": 4.062909567496723e-06,
+      "loss": 5.1973,
+      "step": 124
+    },
+    {
+      "epoch": 0.006605723539815466,
+      "grad_norm": 20.137739181518555,
+      "learning_rate": 5.078636959370905e-06,
+      "loss": 4.6404,
+      "step": 155
+    },
+    {
+      "epoch": 0.00792686824777856,
+      "grad_norm": 15.796778678894043,
+      "learning_rate": 6.094364351245085e-06,
+      "loss": 4.3168,
+      "step": 186
+    },
+    {
+      "epoch": 0.009248012955741652,
+      "grad_norm": 18.784709930419922,
+      "learning_rate": 7.110091743119267e-06,
+      "loss": 4.0553,
+      "step": 217
+    },
+    {
+      "epoch": 0.010569157663704745,
+      "grad_norm": 23.772998809814453,
+      "learning_rate": 8.125819134993446e-06,
+      "loss": 3.84,
+      "step": 248
+    },
+    {
+      "epoch": 0.011890302371667839,
+      "grad_norm": 14.564051628112793,
+      "learning_rate": 9.141546526867629e-06,
+      "loss": 3.7012,
+      "step": 279
+    },
+    {
+      "epoch": 0.013211447079630931,
+      "grad_norm": 14.735116004943848,
+      "learning_rate": 1.015727391874181e-05,
+      "loss": 3.5595,
+      "step": 310
+    },
+    {
+      "epoch": 0.014532591787594026,
+      "grad_norm": 13.88759708404541,
+      "learning_rate": 1.117300131061599e-05,
+      "loss": 3.4515,
+      "step": 341
+    },
+    {
+      "epoch": 0.01585373649555712,
+      "grad_norm": 10.884685516357422,
+      "learning_rate": 1.218872870249017e-05,
+      "loss": 3.3299,
+      "step": 372
+    },
+    {
+      "epoch": 0.01717488120352021,
+      "grad_norm": 12.495129585266113,
+      "learning_rate": 1.3204456094364351e-05,
+      "loss": 3.2453,
+      "step": 403
+    },
+    {
+      "epoch": 0.018496025911483305,
+      "grad_norm": 11.930519104003906,
+      "learning_rate": 1.4220183486238533e-05,
+      "loss": 3.1447,
+      "step": 434
+    },
+    {
+      "epoch": 0.0198171706194464,
+      "grad_norm": 7.951854228973389,
+      "learning_rate": 1.5235910878112714e-05,
+      "loss": 3.0765,
+      "step": 465
+    },
+    {
+      "epoch": 0.02113831532740949,
+      "grad_norm": 8.87326717376709,
+      "learning_rate": 1.6251638269986893e-05,
+      "loss": 3.0296,
+      "step": 496
+    },
+    {
+      "epoch": 0.022459460035372584,
+      "grad_norm": 8.348654747009277,
+      "learning_rate": 1.7267365661861077e-05,
+      "loss": 2.9469,
+      "step": 527
+    },
+    {
+      "epoch": 0.023780604743335678,
+      "grad_norm": 7.966938495635986,
+      "learning_rate": 1.8283093053735257e-05,
+      "loss": 2.8941,
+      "step": 558
+    },
+    {
+      "epoch": 0.02510174945129877,
+      "grad_norm": 6.656867027282715,
+      "learning_rate": 1.9298820445609438e-05,
+      "loss": 2.8428,
+      "step": 589
+    },
+    {
+      "epoch": 0.026422894159261863,
+      "grad_norm": 6.973931312561035,
+      "learning_rate": 2.031454783748362e-05,
+      "loss": 2.7875,
+      "step": 620
+    },
+    {
+      "epoch": 0.027744038867224957,
+      "grad_norm": 6.903556823730469,
+      "learning_rate": 2.13302752293578e-05,
+      "loss": 2.7264,
+      "step": 651
+    },
+    {
+      "epoch": 0.02906518357518805,
+      "grad_norm": 7.422015190124512,
+      "learning_rate": 2.234600262123198e-05,
+      "loss": 2.6833,
+      "step": 682
+    },
+    {
+      "epoch": 0.030386328283151142,
+      "grad_norm": 6.663395404815674,
+      "learning_rate": 2.336173001310616e-05,
+      "loss": 2.6303,
+      "step": 713
+    },
+    {
+      "epoch": 0.03170747299111424,
+      "grad_norm": 6.743407249450684,
+      "learning_rate": 2.437745740498034e-05,
+      "loss": 2.578,
+      "step": 744
+    },
+    {
+      "epoch": 0.03302861769907733,
+      "grad_norm": 5.852759838104248,
+      "learning_rate": 2.5393184796854525e-05,
+      "loss": 2.5248,
+      "step": 775
+    },
+    {
+      "epoch": 0.03434976240704042,
+      "grad_norm": 5.186204433441162,
+      "learning_rate": 2.6408912188728702e-05,
+      "loss": 2.5051,
+      "step": 806
+    },
+    {
+      "epoch": 0.035670907115003515,
+      "grad_norm": 5.873666286468506,
+      "learning_rate": 2.7424639580602886e-05,
+      "loss": 2.4878,
+      "step": 837
+    },
+    {
+      "epoch": 0.03699205182296661,
+      "grad_norm": 6.269757270812988,
+      "learning_rate": 2.8440366972477066e-05,
+      "loss": 2.4272,
+      "step": 868
+    },
+    {
+      "epoch": 0.038313196530929704,
+      "grad_norm": 5.196881294250488,
+      "learning_rate": 2.9456094364351244e-05,
+      "loss": 2.3863,
+      "step": 899
+    },
+    {
+      "epoch": 0.0396343412388928,
+      "grad_norm": 4.899730682373047,
+      "learning_rate": 3.0471821756225428e-05,
+      "loss": 2.3714,
+      "step": 930
+    },
+    {
+      "epoch": 0.04095548594685589,
+      "grad_norm": 4.69920015335083,
+      "learning_rate": 3.148754914809961e-05,
+      "loss": 2.3456,
+      "step": 961
+    },
+    {
+      "epoch": 0.04227663065481898,
+      "grad_norm": 5.293018341064453,
+      "learning_rate": 3.2503276539973785e-05,
+      "loss": 2.3162,
+      "step": 992
+    },
+    {
+      "epoch": 0.043597775362782074,
+      "grad_norm": 4.992343902587891,
+      "learning_rate": 3.351900393184797e-05,
+      "loss": 2.2939,
+      "step": 1023
+    },
+    {
+      "epoch": 0.04491892007074517,
+      "grad_norm": 4.305331230163574,
+      "learning_rate": 3.453473132372215e-05,
+      "loss": 2.275,
+      "step": 1054
+    },
+    {
+      "epoch": 0.04624006477870826,
+      "grad_norm": 4.863211631774902,
+      "learning_rate": 3.555045871559633e-05,
+      "loss": 2.2618,
+      "step": 1085
+    },
+    {
+      "epoch": 0.047561209486671356,
+      "grad_norm": 4.6537251472473145,
+      "learning_rate": 3.6566186107470514e-05,
+      "loss": 2.2584,
+      "step": 1116
+    },
+    {
+      "epoch": 0.04888235419463445,
+      "grad_norm": 7.502182483673096,
+      "learning_rate": 3.7581913499344695e-05,
+      "loss": 2.244,
+      "step": 1147
+    },
+    {
+      "epoch": 0.05020349890259754,
+      "grad_norm": 4.560093879699707,
+      "learning_rate": 3.8597640891218876e-05,
+      "loss": 2.2054,
+      "step": 1178
+    },
+    {
+      "epoch": 0.05152464361056063,
+      "grad_norm": 4.21810245513916,
+      "learning_rate": 3.9613368283093056e-05,
+      "loss": 2.1923,
+      "step": 1209
+    },
+    {
+      "epoch": 0.052845788318523726,
+      "grad_norm": 3.815044403076172,
+      "learning_rate": 4.062909567496724e-05,
+      "loss": 2.1746,
+      "step": 1240
+    },
+    {
+      "epoch": 0.05416693302648682,
+      "grad_norm": 3.4870846271514893,
+      "learning_rate": 4.164482306684142e-05,
+      "loss": 2.1751,
+      "step": 1271
+    },
+    {
+      "epoch": 0.055488077734449914,
+      "grad_norm": 3.8318536281585693,
+      "learning_rate": 4.26605504587156e-05,
+      "loss": 2.1773,
+      "step": 1302
+    },
+    {
+      "epoch": 0.05680922244241301,
+      "grad_norm": 3.6395630836486816,
+      "learning_rate": 4.367627785058978e-05,
+      "loss": 2.1534,
+      "step": 1333
+    },
+    {
+      "epoch": 0.0581303671503761,
+      "grad_norm": 3.4870851039886475,
+      "learning_rate": 4.469200524246396e-05,
+      "loss": 2.1373,
+      "step": 1364
+    },
+    {
+      "epoch": 0.05945151185833919,
+      "grad_norm": 3.533464193344116,
+      "learning_rate": 4.570773263433814e-05,
+      "loss": 2.104,
+      "step": 1395
+    },
+    {
+      "epoch": 0.060772656566302284,
+      "grad_norm": 3.7605605125427246,
+      "learning_rate": 4.672346002621232e-05,
+      "loss": 2.1497,
+      "step": 1426
+    },
+    {
+      "epoch": 0.06209380127426538,
+      "grad_norm": 3.331261396408081,
+      "learning_rate": 4.77391874180865e-05,
+      "loss": 2.103,
+      "step": 1457
+    },
+    {
+      "epoch": 0.06341494598222848,
+      "grad_norm": 3.4388535022735596,
+      "learning_rate": 4.875491480996068e-05,
+      "loss": 2.1017,
+      "step": 1488
+    },
+    {
+      "epoch": 0.06473609069019157,
+      "grad_norm": 3.3586745262145996,
+      "learning_rate": 4.977064220183487e-05,
+      "loss": 2.0883,
+      "step": 1519
+    },
+    {
+      "epoch": 0.06605723539815465,
+      "grad_norm": 3.0805885791778564,
+      "learning_rate": 4.9999915451558777e-05,
+      "loss": 2.0741,
+      "step": 1550
+    },
+    {
+      "epoch": 0.06737838010611776,
+      "grad_norm": 3.1708438396453857,
+      "learning_rate": 4.999955597496219e-05,
+      "loss": 2.0718,
+      "step": 1581
+    },
+    {
+      "epoch": 0.06869952481408084,
+      "grad_norm": 3.03056001663208,
+      "learning_rate": 4.9998914381774255e-05,
+      "loss": 2.0673,
+      "step": 1612
+    },
+    {
+      "epoch": 0.07002066952204394,
+      "grad_norm": 2.7995362281799316,
+      "learning_rate": 4.999799067923527e-05,
+      "loss": 2.0683,
+      "step": 1643
+    },
+    {
+      "epoch": 0.07134181423000703,
+      "grad_norm": 3.116344451904297,
+      "learning_rate": 4.999678487776908e-05,
+      "loss": 2.0231,
+      "step": 1674
+    },
+    {
+      "epoch": 0.07266295893797013,
+      "grad_norm": 3.1621696949005127,
+      "learning_rate": 4.9995296990983006e-05,
+      "loss": 2.0391,
+      "step": 1705
+    },
+    {
+      "epoch": 0.07398410364593322,
+      "grad_norm": 2.8406624794006348,
+      "learning_rate": 4.999352703566763e-05,
+      "loss": 2.0132,
+      "step": 1736
+    },
+    {
+      "epoch": 0.0753052483538963,
+      "grad_norm": 2.570624351501465,
+      "learning_rate": 4.999147503179668e-05,
+      "loss": 2.0116,
+      "step": 1767
+    },
+    {
+      "epoch": 0.07662639306185941,
+      "grad_norm": 2.7663609981536865,
+      "learning_rate": 4.998914100252672e-05,
+      "loss": 2.0127,
+      "step": 1798
+    },
+    {
+      "epoch": 0.0779475377698225,
+      "grad_norm": 2.5649290084838867,
+      "learning_rate": 4.998652497419696e-05,
+      "loss": 2.0033,
+      "step": 1829
+    },
+    {
+      "epoch": 0.0792686824777856,
+      "grad_norm": 2.575486183166504,
+      "learning_rate": 4.9983626976328927e-05,
+      "loss": 2.005,
+      "step": 1860
+    },
+    {
+      "epoch": 0.08058982718574868,
+      "grad_norm": 2.509678840637207,
+      "learning_rate": 4.998044704162613e-05,
+      "loss": 1.9824,
+      "step": 1891
+    },
+    {
+      "epoch": 0.08191097189371178,
+      "grad_norm": 2.6347391605377197,
+      "learning_rate": 4.9976985205973705e-05,
+      "loss": 1.9571,
+      "step": 1922
+    },
+    {
+      "epoch": 0.08323211660167487,
+      "grad_norm": 2.5539751052856445,
+      "learning_rate": 4.997324150843799e-05,
+      "loss": 1.9782,
+      "step": 1953
+    },
+    {
+      "epoch": 0.08455326130963796,
+      "grad_norm": 10.463569641113281,
+      "learning_rate": 4.99692159912661e-05,
+      "loss": 1.9713,
+      "step": 1984
+    },
+    {
+      "epoch": 0.08587440601760106,
+      "grad_norm": 2.604071617126465,
+      "learning_rate": 4.996490869988546e-05,
+      "loss": 1.9769,
+      "step": 2015
+    },
+    {
+      "epoch": 0.08719555072556415,
+      "grad_norm": 2.4612362384796143,
+      "learning_rate": 4.996031968290326e-05,
+      "loss": 1.9463,
+      "step": 2046
+    },
+    {
+      "epoch": 0.08851669543352725,
+      "grad_norm": 2.539299488067627,
+      "learning_rate": 4.995544899210594e-05,
+      "loss": 1.9414,
+      "step": 2077
+    },
+    {
+      "epoch": 0.08983784014149034,
+      "grad_norm": 2.6650774478912354,
+      "learning_rate": 4.9950296682458583e-05,
+      "loss": 1.9371,
+      "step": 2108
+    },
+    {
+      "epoch": 0.09115898484945342,
+      "grad_norm": 2.5480291843414307,
+      "learning_rate": 4.994486281210429e-05,
+      "loss": 1.9581,
+      "step": 2139
+    },
+    {
+      "epoch": 0.09248012955741652,
+      "grad_norm": 2.4799551963806152,
+      "learning_rate": 4.9939147442363566e-05,
+      "loss": 1.9524,
+      "step": 2170
+    },
+    {
+      "epoch": 0.09380127426537961,
+      "grad_norm": 2.3326504230499268,
+      "learning_rate": 4.9933150637733574e-05,
+      "loss": 1.9319,
+      "step": 2201
+    },
+    {
+      "epoch": 0.09512241897334271,
+      "grad_norm": 2.3550055027008057,
+      "learning_rate": 4.992687246588743e-05,
+      "loss": 1.9015,
+      "step": 2232
+    },
+    {
+      "epoch": 0.0964435636813058,
+      "grad_norm": 2.422797918319702,
+      "learning_rate": 4.992031299767347e-05,
+      "loss": 1.9364,
+      "step": 2263
+    },
+    {
+      "epoch": 0.0977647083892689,
+      "grad_norm": 2.2776873111724854,
+      "learning_rate": 4.9913472307114386e-05,
+      "loss": 1.9102,
+      "step": 2294
+    },
+    {
+      "epoch": 0.09908585309723199,
+      "grad_norm": 2.3031086921691895,
+      "learning_rate": 4.9906350471406446e-05,
+      "loss": 1.9193,
+      "step": 2325
+    },
+    {
+      "epoch": 0.10040699780519508,
+      "grad_norm": 2.2281057834625244,
+      "learning_rate": 4.989894757091861e-05,
+      "loss": 1.9143,
+      "step": 2356
+    },
+    {
+      "epoch": 0.10172814251315818,
+      "grad_norm": 2.39752459526062,
+      "learning_rate": 4.989126368919158e-05,
+      "loss": 1.905,
+      "step": 2387
+    },
+    {
+      "epoch": 0.10304928722112126,
+      "grad_norm": 2.3098957538604736,
+      "learning_rate": 4.988329891293693e-05,
+      "loss": 1.9204,
+      "step": 2418
+    },
+    {
+      "epoch": 0.10437043192908436,
+      "grad_norm": 2.244088649749756,
+      "learning_rate": 4.987505333203608e-05,
+      "loss": 1.8836,
+      "step": 2449
+    },
+    {
+      "epoch": 0.10569157663704745,
+      "grad_norm": 2.3035812377929688,
+      "learning_rate": 4.9866527039539276e-05,
+      "loss": 1.9097,
+      "step": 2480
+    },
+    {
+      "epoch": 0.10701272134501055,
+      "grad_norm": 2.2249033451080322,
+      "learning_rate": 4.9857720131664594e-05,
+      "loss": 1.8804,
+      "step": 2511
+    },
+    {
+      "epoch": 0.10833386605297364,
+      "grad_norm": 2.180612325668335,
+      "learning_rate": 4.9848632707796773e-05,
+      "loss": 1.8848,
+      "step": 2542
+    },
+    {
+      "epoch": 0.10965501076093673,
+      "grad_norm": 2.2042863368988037,
+      "learning_rate": 4.9839264870486155e-05,
+      "loss": 1.8879,
+      "step": 2573
+    },
+    {
+      "epoch": 0.11097615546889983,
+      "grad_norm": 2.278650999069214,
+      "learning_rate": 4.9829616725447526e-05,
+      "loss": 1.8728,
+      "step": 2604
+    },
+    {
+      "epoch": 0.11229730017686292,
+      "grad_norm": 2.2380714416503906,
+      "learning_rate": 4.981968838155888e-05,
+      "loss": 1.8927,
+      "step": 2635
+    },
+    {
+      "epoch": 0.11361844488482602,
+      "grad_norm": 2.062289237976074,
+      "learning_rate": 4.980947995086024e-05,
+      "loss": 1.8706,
+      "step": 2666
+    },
+    {
+      "epoch": 0.1149395895927891,
+      "grad_norm": 2.169900894165039,
+      "learning_rate": 4.979899154855234e-05,
+      "loss": 1.8753,
+      "step": 2697
+    },
+    {
+      "epoch": 0.1162607343007522,
+      "grad_norm": 2.1450119018554688,
+      "learning_rate": 4.9788223292995386e-05,
+      "loss": 1.8685,
+      "step": 2728
+    },
+    {
+      "epoch": 0.11758187900871529,
+      "grad_norm": 2.285205125808716,
+      "learning_rate": 4.977717530570768e-05,
+      "loss": 1.8601,
+      "step": 2759
+    },
+    {
+      "epoch": 0.11890302371667838,
+      "grad_norm": 2.1305880546569824,
+      "learning_rate": 4.976584771136425e-05,
+      "loss": 1.8557,
+      "step": 2790
+    },
+    {
+      "epoch": 0.12022416842464148,
+      "grad_norm": 2.0517284870147705,
+      "learning_rate": 4.975424063779547e-05,
+      "loss": 1.8587,
+      "step": 2821
+    },
+    {
+      "epoch": 0.12154531313260457,
+      "grad_norm": 2.190729856491089,
+      "learning_rate": 4.974235421598557e-05,
+      "loss": 1.8431,
+      "step": 2852
+    },
+    {
+      "epoch": 0.12286645784056767,
+      "grad_norm": 2.133803367614746,
+      "learning_rate": 4.973018858007122e-05,
+      "loss": 1.8492,
+      "step": 2883
+    },
+    {
+      "epoch": 0.12418760254853076,
+      "grad_norm": 2.3807687759399414,
+      "learning_rate": 4.9717743867339963e-05,
+      "loss": 1.8391,
+      "step": 2914
+    },
+    {
+      "epoch": 0.12550874725649386,
+      "grad_norm": 2.0250234603881836,
+      "learning_rate": 4.9705020218228695e-05,
+      "loss": 1.8465,
+      "step": 2945
+    },
+    {
+      "epoch": 0.12682989196445696,
+      "grad_norm": 2.1557273864746094,
+      "learning_rate": 4.969201777632205e-05,
+      "loss": 1.8346,
+      "step": 2976
+    },
+    {
+      "epoch": 0.12815103667242003,
+      "grad_norm": 1.9695699214935303,
+      "learning_rate": 4.9678736688350846e-05,
+      "loss": 1.8423,
+      "step": 3007
+    },
+    {
+      "epoch": 0.12947218138038313,
+      "grad_norm": 2.1525843143463135,
+      "learning_rate": 4.966517710419033e-05,
+      "loss": 1.851,
+      "step": 3038
+    },
+    {
+      "epoch": 0.13079332608834623,
+      "grad_norm": 1.9812391996383667,
+      "learning_rate": 4.965133917685858e-05,
+      "loss": 1.8338,
+      "step": 3069
+    },
+    {
+      "epoch": 0.1321144707963093,
+      "grad_norm": 2.2273776531219482,
+      "learning_rate": 4.9637223062514714e-05,
+      "loss": 1.8329,
+      "step": 3100
+    },
+    {
+      "epoch": 0.1334356155042724,
+      "grad_norm": 2.1103451251983643,
+      "learning_rate": 4.962282892045718e-05,
+      "loss": 1.8243,
+      "step": 3131
+    },
+    {
+      "epoch": 0.1347567602122355,
+      "grad_norm": 2.083226203918457,
+      "learning_rate": 4.9608156913121904e-05,
+      "loss": 1.8254,
+      "step": 3162
+    },
+    {
+      "epoch": 0.1360779049201986,
+      "grad_norm": 1.878396987915039,
+      "learning_rate": 4.959320720608049e-05,
+      "loss": 1.819,
+      "step": 3193
+    },
+    {
+      "epoch": 0.13739904962816168,
+      "grad_norm": 1.9980756044387817,
+      "learning_rate": 4.9577979968038354e-05,
+      "loss": 1.8391,
+      "step": 3224
+    },
+    {
+      "epoch": 0.13872019433612479,
+      "grad_norm": 1.977419376373291,
+      "learning_rate": 4.956247537083282e-05,
+      "loss": 1.8233,
+      "step": 3255
+    },
+    {
+      "epoch": 0.1400413390440879,
+      "grad_norm": 1.9977376461029053,
+      "learning_rate": 4.9546693589431145e-05,
+      "loss": 1.8254,
+      "step": 3286
+    },
+    {
+      "epoch": 0.14136248375205096,
+      "grad_norm": 1.8682738542556763,
+      "learning_rate": 4.9530634801928595e-05,
+      "loss": 1.8154,
+      "step": 3317
+    },
+    {
+      "epoch": 0.14268362846001406,
+      "grad_norm": 1.9372395277023315,
+      "learning_rate": 4.9514299189546395e-05,
+      "loss": 1.8229,
+      "step": 3348
+    },
+    {
+      "epoch": 0.14400477316797716,
+      "grad_norm": 1.9475616216659546,
+      "learning_rate": 4.949768693662973e-05,
+      "loss": 1.8122,
+      "step": 3379
+    },
+    {
+      "epoch": 0.14532591787594026,
+      "grad_norm": 1.8302438259124756,
+      "learning_rate": 4.948079823064559e-05,
+      "loss": 1.8275,
+      "step": 3410
+    },
+    {
+      "epoch": 0.14664706258390334,
+      "grad_norm": 1.9195733070373535,
+      "learning_rate": 4.946363326218074e-05,
+      "loss": 1.8288,
+      "step": 3441
+    },
+    {
+      "epoch": 0.14796820729186644,
+      "grad_norm": 2.056960105895996,
+      "learning_rate": 4.9446192224939525e-05,
+      "loss": 1.8267,
+      "step": 3472
+    },
+    {
+      "epoch": 0.14928935199982954,
+      "grad_norm": 1.9093526601791382,
+      "learning_rate": 4.942847531574167e-05,
+      "loss": 1.8118,
+      "step": 3503
+    },
+    {
+      "epoch": 0.1506104967077926,
+      "grad_norm": 1.800930380821228,
+      "learning_rate": 4.941048273452008e-05,
+      "loss": 1.7869,
+      "step": 3534
+    },
+    {
+      "epoch": 0.1519316414157557,
+      "grad_norm": 2.0274600982666016,
+      "learning_rate": 4.9392214684318605e-05,
+      "loss": 1.7937,
+      "step": 3565
+    },
+    {
+      "epoch": 0.15325278612371881,
+      "grad_norm": 1.880557656288147,
+      "learning_rate": 4.93736713712897e-05,
+      "loss": 1.793,
+      "step": 3596
+    },
+    {
+      "epoch": 0.15457393083168192,
+      "grad_norm": 1.9970703125,
+      "learning_rate": 4.9354853004692124e-05,
+      "loss": 1.7922,
+      "step": 3627
+    },
+    {
+      "epoch": 0.155895075539645,
+      "grad_norm": 2.0239968299865723,
+      "learning_rate": 4.93357597968886e-05,
+      "loss": 1.8026,
+      "step": 3658
+    },
+    {
+      "epoch": 0.1572162202476081,
+      "grad_norm": 1.8837428092956543,
+      "learning_rate": 4.931639196334338e-05,
+      "loss": 1.7898,
+      "step": 3689
+    },
+    {
+      "epoch": 0.1585373649555712,
+      "grad_norm": 1.8433760404586792,
+      "learning_rate": 4.9296749722619826e-05,
+      "loss": 1.7854,
+      "step": 3720
+    },
+    {
+      "epoch": 0.15985850966353427,
+      "grad_norm": 1.898555874824524,
+      "learning_rate": 4.9276833296377966e-05,
+      "loss": 1.8011,
+      "step": 3751
+    },
+    {
+      "epoch": 0.16117965437149737,
+      "grad_norm": 1.9116350412368774,
+      "learning_rate": 4.925664290937196e-05,
+      "loss": 1.7759,
+      "step": 3782
+    },
+    {
+      "epoch": 0.16250079907946047,
+      "grad_norm": 1.8142945766448975,
+      "learning_rate": 4.9236178789447576e-05,
+      "loss": 1.7712,
+      "step": 3813
+    },
+    {
+      "epoch": 0.16382194378742357,
+      "grad_norm": 1.8500661849975586,
+      "learning_rate": 4.921544116753962e-05,
+      "loss": 1.7885,
+      "step": 3844
+    },
+    {
+      "epoch": 0.16514308849538664,
+      "grad_norm": 1.9167587757110596,
+      "learning_rate": 4.919443027766935e-05,
+      "loss": 1.792,
+      "step": 3875
+    },
+    {
+      "epoch": 0.16646423320334974,
+      "grad_norm": 1.891123652458191,
+      "learning_rate": 4.91731463569418e-05,
+      "loss": 1.7963,
+      "step": 3906
+    },
+    {
+      "epoch": 0.16778537791131284,
+      "grad_norm": 1.7814100980758667,
+      "learning_rate": 4.915158964554312e-05,
+      "loss": 1.7836,
+      "step": 3937
+    },
+    {
+      "epoch": 0.16910652261927592,
+      "grad_norm": 2.0991482734680176,
+      "learning_rate": 4.912976038673786e-05,
+      "loss": 1.7788,
+      "step": 3968
+    },
+    {
+      "epoch": 0.17042766732723902,
+      "grad_norm": 1.8836603164672852,
+      "learning_rate": 4.9107658826866254e-05,
+      "loss": 1.7885,
+      "step": 3999
+    },
+    {
+      "epoch": 0.17174881203520212,
+      "grad_norm": 1.8759151697158813,
+      "learning_rate": 4.908528521534139e-05,
+      "loss": 1.7622,
+      "step": 4030
+    },
+    {
+      "epoch": 0.17306995674316522,
+      "grad_norm": 1.9080370664596558,
+      "learning_rate": 4.906263980464644e-05,
+      "loss": 1.7778,
+      "step": 4061
+    },
+    {
+      "epoch": 0.1743911014511283,
+      "grad_norm": 1.776382327079773,
+      "learning_rate": 4.903972285033178e-05,
+      "loss": 1.7415,
+      "step": 4092
+    },
+    {
+      "epoch": 0.1757122461590914,
+      "grad_norm": 1.8282791376113892,
+      "learning_rate": 4.901653461101213e-05,
+      "loss": 1.776,
+      "step": 4123
+    },
+    {
+      "epoch": 0.1770333908670545,
+      "grad_norm": 1.7473266124725342,
+      "learning_rate": 4.8993075348363626e-05,
+      "loss": 1.7625,
+      "step": 4154
+    },
+    {
+      "epoch": 0.17835453557501757,
+      "grad_norm": 1.849486231803894,
+      "learning_rate": 4.896934532712084e-05,
+      "loss": 1.7645,
+      "step": 4185
+    },
+    {
+      "epoch": 0.17967568028298067,
+      "grad_norm": 1.772829294204712,
+      "learning_rate": 4.8945344815073846e-05,
+      "loss": 1.7563,
+      "step": 4216
+    },
+    {
+      "epoch": 0.18099682499094377,
+      "grad_norm": 1.8278394937515259,
+      "learning_rate": 4.892107408306516e-05,
+      "loss": 1.7591,
+      "step": 4247
+    },
+    {
+      "epoch": 0.18231796969890685,
+      "grad_norm": 1.7458796501159668,
+      "learning_rate": 4.889653340498669e-05,
+      "loss": 1.7444,
+      "step": 4278
+    },
+    {
+      "epoch": 0.18363911440686995,
+      "grad_norm": 1.8150420188903809,
+      "learning_rate": 4.8871723057776664e-05,
+      "loss": 1.7486,
+      "step": 4309
+    },
+    {
+      "epoch": 0.18496025911483305,
+      "grad_norm": 1.914010763168335,
+      "learning_rate": 4.8846643321416476e-05,
+      "loss": 1.7456,
+      "step": 4340
+    },
+    {
+      "epoch": 0.18628140382279615,
+      "grad_norm": 1.7134442329406738,
+      "learning_rate": 4.882129447892753e-05,
+      "loss": 1.7521,
+      "step": 4371
+    },
+    {
+      "epoch": 0.18760254853075922,
+      "grad_norm": 1.6600676774978638,
+      "learning_rate": 4.8795676816368076e-05,
+      "loss": 1.7416,
+      "step": 4402
+    },
+    {
+      "epoch": 0.18892369323872232,
+      "grad_norm": 1.8451883792877197,
+      "learning_rate": 4.876979062282995e-05,
+      "loss": 1.7595,
+      "step": 4433
+    },
+    {
+      "epoch": 0.19024483794668542,
+      "grad_norm": 1.8544697761535645,
+      "learning_rate": 4.8743636190435325e-05,
+      "loss": 1.7669,
+      "step": 4464
+    },
+    {
+      "epoch": 0.1915659826546485,
+      "grad_norm": 1.704041600227356,
+      "learning_rate": 4.871721381433344e-05,
+      "loss": 1.7403,
+      "step": 4495
+    },
+    {
+      "epoch": 0.1928871273626116,
+      "grad_norm": 1.6593477725982666,
+      "learning_rate": 4.869052379269719e-05,
+      "loss": 1.7359,
+      "step": 4526
+    },
+    {
+      "epoch": 0.1942082720705747,
+      "grad_norm": 1.7819074392318726,
+      "learning_rate": 4.866356642671985e-05,
+      "loss": 1.7485,
+      "step": 4557
+    },
+    {
+      "epoch": 0.1955294167785378,
+      "grad_norm": 1.840776801109314,
+      "learning_rate": 4.8636342020611634e-05,
+      "loss": 1.7566,
+      "step": 4588
+    },
+    {
+      "epoch": 0.19685056148650087,
+      "grad_norm": 1.9981670379638672,
+      "learning_rate": 4.860885088159626e-05,
+      "loss": 1.747,
+      "step": 4619
+    },
+    {
+      "epoch": 0.19817170619446398,
+      "grad_norm": 1.8164016008377075,
+      "learning_rate": 4.858109331990751e-05,
+      "loss": 1.7575,
+      "step": 4650
+    },
+    {
+      "epoch": 0.19949285090242708,
+      "grad_norm": 1.7589538097381592,
+      "learning_rate": 4.855306964878567e-05,
+      "loss": 1.7387,
+      "step": 4681
+    },
+    {
+      "epoch": 0.20081399561039015,
+      "grad_norm": 1.8388264179229736,
+      "learning_rate": 4.8524780184474084e-05,
+      "loss": 1.7273,
+      "step": 4712
+    },
+    {
+      "epoch": 0.20213514031835325,
+      "grad_norm": 1.7422165870666504,
+      "learning_rate": 4.8496225246215496e-05,
+      "loss": 1.7301,
+      "step": 4743
+    },
+    {
+      "epoch": 0.20345628502631635,
+      "grad_norm": 1.6752463579177856,
+      "learning_rate": 4.8467405156248505e-05,
+      "loss": 1.7422,
+      "step": 4774
+    },
+    {
+      "epoch": 0.20477742973427945,
+      "grad_norm": 6.59528923034668,
+      "learning_rate": 4.843832023980392e-05,
+      "loss": 1.748,
+      "step": 4805
+    },
+    {
+      "epoch": 0.20609857444224253,
+      "grad_norm": 1.716232419013977,
+      "learning_rate": 4.840897082510106e-05,
+      "loss": 1.745,
+      "step": 4836
+    },
+    {
+      "epoch": 0.20741971915020563,
+      "grad_norm": 1.7691062688827515,
+      "learning_rate": 4.8379357243344084e-05,
+      "loss": 1.7208,
+      "step": 4867
+    },
+    {
+      "epoch": 0.20874086385816873,
+      "grad_norm": 1.6334593296051025,
+      "learning_rate": 4.8349479828718236e-05,
+      "loss": 1.7288,
+      "step": 4898
+    },
+    {
+      "epoch": 0.2100620085661318,
+      "grad_norm": 1.6801882982254028,
+      "learning_rate": 4.8319338918386075e-05,
+      "loss": 1.7428,
+      "step": 4929
+    },
+    {
+      "epoch": 0.2113831532740949,
+      "grad_norm": 1.6692436933517456,
+      "learning_rate": 4.828893485248369e-05,
+      "loss": 1.7286,
+      "step": 4960
+    },
+    {
+      "epoch": 0.212704297982058,
+      "grad_norm": 1.731826663017273,
+      "learning_rate": 4.825826797411682e-05,
+      "loss": 1.7455,
+      "step": 4991
+    },
+    {
+      "epoch": 0.2140254426900211,
+      "grad_norm": 1.689012050628662,
+      "learning_rate": 4.822733862935702e-05,
+      "loss": 1.7398,
+      "step": 5022
+    },
+    {
+      "epoch": 0.21534658739798418,
+      "grad_norm": 1.6423673629760742,
+      "learning_rate": 4.819614716723775e-05,
+      "loss": 1.7305,
+      "step": 5053
+    },
+    {
+      "epoch": 0.21666773210594728,
+      "grad_norm": 1.7029454708099365,
+      "learning_rate": 4.8164693939750425e-05,
+      "loss": 1.7355,
+      "step": 5084
+    },
+    {
+      "epoch": 0.21798887681391038,
+      "grad_norm": 1.7331738471984863,
+      "learning_rate": 4.813297930184042e-05,
+      "loss": 1.7167,
+      "step": 5115
+    },
+    {
+      "epoch": 0.21931002152187345,
+      "grad_norm": 1.729872226715088,
+      "learning_rate": 4.810100361140314e-05,
+      "loss": 1.7402,
+      "step": 5146
+    },
+    {
+      "epoch": 0.22063116622983656,
+      "grad_norm": 1.675826907157898,
+      "learning_rate": 4.8068767229279885e-05,
+      "loss": 1.7129,
+      "step": 5177
+    },
+    {
+      "epoch": 0.22195231093779966,
+      "grad_norm": 1.7073479890823364,
+      "learning_rate": 4.8036270519253854e-05,
+      "loss": 1.7292,
+      "step": 5208
+    },
+    {
+      "epoch": 0.22327345564576276,
+      "grad_norm": 1.67916738986969,
+      "learning_rate": 4.8003513848046e-05,
+      "loss": 1.7417,
+      "step": 5239
+    },
+    {
+      "epoch": 0.22459460035372583,
+      "grad_norm": 1.751838207244873,
+      "learning_rate": 4.79704975853109e-05,
+      "loss": 1.7167,
+      "step": 5270
+    },
+    {
+      "epoch": 0.22591574506168893,
+      "grad_norm": 1.7748228311538696,
+      "learning_rate": 4.793722210363262e-05,
+      "loss": 1.7175,
+      "step": 5301
+    },
+    {
+      "epoch": 0.22723688976965203,
+      "grad_norm": 1.6792418956756592,
+      "learning_rate": 4.7903687778520414e-05,
+      "loss": 1.7537,
+      "step": 5332
+    },
+    {
+      "epoch": 0.2285580344776151,
+      "grad_norm": 1.77089524269104,
+      "learning_rate": 4.7869894988404593e-05,
+      "loss": 1.7181,
+      "step": 5363
+    },
+    {
+      "epoch": 0.2298791791855782,
+      "grad_norm": 1.795487642288208,
+      "learning_rate": 4.783584411463221e-05,
+      "loss": 1.7216,
+      "step": 5394
+    },
+    {
+      "epoch": 0.2312003238935413,
+      "grad_norm": 1.8036439418792725,
+      "learning_rate": 4.780153554146274e-05,
+      "loss": 1.7172,
+      "step": 5425
+    },
+    {
+      "epoch": 0.2325214686015044,
+      "grad_norm": 1.683451771736145,
+      "learning_rate": 4.7766969656063766e-05,
+      "loss": 1.7371,
+      "step": 5456
+    },
+    {
+      "epoch": 0.23384261330946748,
+      "grad_norm": 1.7509490251541138,
+      "learning_rate": 4.773214684850662e-05,
+      "loss": 1.7245,
+      "step": 5487
+    },
+    {
+      "epoch": 0.23516375801743059,
+      "grad_norm": 1.6572686433792114,
+      "learning_rate": 4.769706751176193e-05,
+      "loss": 1.7108,
+      "step": 5518
+    },
+    {
+      "epoch": 0.2364849027253937,
+      "grad_norm": 1.7440228462219238,
+      "learning_rate": 4.7661732041695264e-05,
+      "loss": 1.7115,
+      "step": 5549
+    },
+    {
+      "epoch": 0.23780604743335676,
+      "grad_norm": 1.613687515258789,
+      "learning_rate": 4.762614083706258e-05,
+      "loss": 1.7079,
+      "step": 5580
+    },
+    {
+      "epoch": 0.23912719214131986,
+      "grad_norm": 1.4738078117370605,
+      "learning_rate": 4.759029429950581e-05,
+      "loss": 1.7097,
+      "step": 5611
+    },
+    {
+      "epoch": 0.24044833684928296,
+      "grad_norm": 1.6921734809875488,
+      "learning_rate": 4.7554192833548235e-05,
+      "loss": 1.7187,
+      "step": 5642
+    },
+    {
+      "epoch": 0.24176948155724606,
+      "grad_norm": 1.665814757347107,
+      "learning_rate": 4.751783684659e-05,
+      "loss": 1.7302,
+      "step": 5673
+    },
+    {
+      "epoch": 0.24309062626520914,
+      "grad_norm": 1.5973016023635864,
+      "learning_rate": 4.748122674890348e-05,
+      "loss": 1.718,
+      "step": 5704
+    },
+    {
+      "epoch": 0.24441177097317224,
+      "grad_norm": 1.6037402153015137,
+      "learning_rate": 4.7444362953628654e-05,
+      "loss": 1.7213,
+      "step": 5735
+    },
+    {
+      "epoch": 0.24573291568113534,
+      "grad_norm": 1.6214978694915771,
+      "learning_rate": 4.7407245876768424e-05,
+      "loss": 1.7343,
+      "step": 5766
+    },
+    {
+      "epoch": 0.2470540603890984,
+      "grad_norm": 1.758327603340149,
+      "learning_rate": 4.736987593718397e-05,
+      "loss": 1.6993,
+      "step": 5797
+    },
+    {
+      "epoch": 0.2483752050970615,
+      "grad_norm": 1.7455657720565796,
+      "learning_rate": 4.733225355658999e-05,
+      "loss": 1.7147,
+      "step": 5828
+    },
+    {
+      "epoch": 0.24969634980502461,
+      "grad_norm": 1.6355786323547363,
+      "learning_rate": 4.7294379159549926e-05,
+      "loss": 1.7284,
+      "step": 5859
+    },
+    {
+      "epoch": 0.2510174945129877,
+      "grad_norm": 1.6421799659729004,
+      "learning_rate": 4.725625317347119e-05,
+      "loss": 1.6986,
+      "step": 5890
+    },
+    {
+      "epoch": 0.2523386392209508,
+      "grad_norm": 1.6713591814041138,
+      "learning_rate": 4.7217876028600374e-05,
+      "loss": 1.7083,
+      "step": 5921
+    },
+    {
+      "epoch": 0.2536597839289139,
+      "grad_norm": 1.7056536674499512,
+      "learning_rate": 4.717924815801832e-05,
+      "loss": 1.7339,
+      "step": 5952
+    },
+    {
+      "epoch": 0.254980928636877,
+      "grad_norm": 1.6911596059799194,
+      "learning_rate": 4.714036999763532e-05,
+      "loss": 1.7068,
+      "step": 5983
+    },
+    {
+      "epoch": 0.25630207334484006,
+      "grad_norm": 1.68338143825531,
+      "learning_rate": 4.7101241986186116e-05,
+      "loss": 1.7153,
+      "step": 6014
+    },
+    {
+      "epoch": 0.2576232180528032,
+      "grad_norm": 1.6629693508148193,
+      "learning_rate": 4.7061864565225e-05,
+      "loss": 1.7178,
+      "step": 6045
+    },
+    {
+      "epoch": 0.25894436276076627,
+      "grad_norm": 1.8833751678466797,
+      "learning_rate": 4.702223817912081e-05,
+      "loss": 1.6759,
+      "step": 6076
+    },
+    {
+      "epoch": 0.26026550746872934,
+      "grad_norm": 1.8015220165252686,
+      "learning_rate": 4.698236327505195e-05,
+      "loss": 1.6893,
+      "step": 6107
+    },
+    {
+      "epoch": 0.26158665217669247,
+      "grad_norm": 1.761045217514038,
+      "learning_rate": 4.694224030300127e-05,
+      "loss": 1.7015,
+      "step": 6138
+    },
+    {
+      "epoch": 0.26290779688465554,
+      "grad_norm": 1.6620142459869385,
+      "learning_rate": 4.690186971575107e-05,
+      "loss": 1.6897,
+      "step": 6169
+    },
+    {
+      "epoch": 0.2642289415926186,
+      "grad_norm": 1.6513328552246094,
+      "learning_rate": 4.6861251968877916e-05,
+      "loss": 1.6926,
+      "step": 6200
+    },
+    {
+      "epoch": 0.26555008630058174,
+      "grad_norm": 1.6536865234375,
+      "learning_rate": 4.68203875207476e-05,
+      "loss": 1.7193,
+      "step": 6231
+    },
+    {
+      "epoch": 0.2668712310085448,
+      "grad_norm": 1.5718939304351807,
+      "learning_rate": 4.677927683250983e-05,
+      "loss": 1.6821,
+      "step": 6262
+    },
+    {
+      "epoch": 0.2681923757165079,
+      "grad_norm": 1.6237717866897583,
+      "learning_rate": 4.6737920368093156e-05,
+      "loss": 1.7116,
+      "step": 6293
+    },
+    {
+      "epoch": 0.269513520424471,
+      "grad_norm": 1.6111472845077515,
+      "learning_rate": 4.669631859419965e-05,
+      "loss": 1.7101,
+      "step": 6324
+    },
+    {
+      "epoch": 0.2708346651324341,
+      "grad_norm": 1.711414098739624,
+      "learning_rate": 4.6654471980299676e-05,
+      "loss": 1.7053,
+      "step": 6355
+    },
+    {
+      "epoch": 0.2721558098403972,
+      "grad_norm": 1.5804500579833984,
+      "learning_rate": 4.661238099862658e-05,
+      "loss": 1.6867,
+      "step": 6386
+    },
+    {
+      "epoch": 0.2734769545483603,
+      "grad_norm": 1.5932157039642334,
+      "learning_rate": 4.657004612417138e-05,
+      "loss": 1.7181,
+      "step": 6417
+    },
+    {
+      "epoch": 0.27479809925632337,
+      "grad_norm": 1.5526095628738403,
+      "learning_rate": 4.6527467834677374e-05,
+      "loss": 1.7178,
+      "step": 6448
+    },
+    {
+      "epoch": 0.2761192439642865,
+      "grad_norm": 1.5979021787643433,
+      "learning_rate": 4.648464661063478e-05,
+      "loss": 1.6885,
+      "step": 6479
+    },
+    {
+      "epoch": 0.27744038867224957,
+      "grad_norm": 1.5154697895050049,
+      "learning_rate": 4.6441582935275264e-05,
+      "loss": 1.6798,
+      "step": 6510
+    },
+    {
+      "epoch": 0.27876153338021264,
+      "grad_norm": 1.6004109382629395,
+      "learning_rate": 4.6398277294566586e-05,
+      "loss": 1.6877,
+      "step": 6541
+    },
+    {
+      "epoch": 0.2800826780881758,
+      "grad_norm": 1.7409225702285767,
+      "learning_rate": 4.6354730177207e-05,
+      "loss": 1.6877,
+      "step": 6572
+    },
+    {
+      "epoch": 0.28140382279613885,
+      "grad_norm": 1.5840095281600952,
+      "learning_rate": 4.6310942074619787e-05,
+      "loss": 1.6885,
+      "step": 6603
+    },
+    {
+      "epoch": 0.2827249675041019,
+      "grad_norm": 1.7218347787857056,
+      "learning_rate": 4.626691348094777e-05,
+      "loss": 1.6993,
+      "step": 6634
+    },
+    {
+      "epoch": 0.28404611221206505,
+      "grad_norm": 1.6517518758773804,
+      "learning_rate": 4.622264489304762e-05,
+      "loss": 1.7086,
+      "step": 6665
+    },
+    {
+      "epoch": 0.2853672569200281,
+      "grad_norm": 1.6828736066818237,
+      "learning_rate": 4.617813681048434e-05,
+      "loss": 1.6895,
+      "step": 6696
+    },
+    {
+      "epoch": 0.2866884016279912,
+      "grad_norm": 1.6049277782440186,
+      "learning_rate": 4.61333897355256e-05,
+      "loss": 1.6929,
+      "step": 6727
+    },
+    {
+      "epoch": 0.2880095463359543,
+      "grad_norm": 1.6272588968276978,
+      "learning_rate": 4.608840417313604e-05,
+      "loss": 1.693,
+      "step": 6758
+    },
+    {
+      "epoch": 0.2893306910439174,
+      "grad_norm": 1.6424005031585693,
+      "learning_rate": 4.6043180630971646e-05,
+      "loss": 1.6882,
+      "step": 6789
+    },
+    {
+      "epoch": 0.2906518357518805,
+      "grad_norm": 1.5532363653182983,
+      "learning_rate": 4.599771961937391e-05,
+      "loss": 1.6846,
+      "step": 6820
+    },
+    {
+      "epoch": 0.2919729804598436,
+      "grad_norm": 1.566545009613037,
+      "learning_rate": 4.5952021651364204e-05,
+      "loss": 1.6694,
+      "step": 6851
+    },
+    {
+      "epoch": 0.2932941251678067,
+      "grad_norm": 1.592480182647705,
+      "learning_rate": 4.590608724263786e-05,
+      "loss": 1.6689,
+      "step": 6882
+    },
+    {
+      "epoch": 0.2946152698757698,
+      "grad_norm": 1.6120296716690063,
+      "learning_rate": 4.585991691155845e-05,
+      "loss": 1.6654,
+      "step": 6913
+    },
+    {
+      "epoch": 0.2959364145837329,
+      "grad_norm": 1.5690969228744507,
+      "learning_rate": 4.581351117915188e-05,
+      "loss": 1.6586,
+      "step": 6944
+    },
+    {
+      "epoch": 0.29725755929169595,
+      "grad_norm": 1.5756118297576904,
+      "learning_rate": 4.5766870569100534e-05,
+      "loss": 1.6859,
+      "step": 6975
+    },
+    {
+      "epoch": 0.2985787039996591,
+      "grad_norm": 1.5164366960525513,
+      "learning_rate": 4.571999560773736e-05,
+      "loss": 1.6844,
+      "step": 7006
+    },
+    {
+      "epoch": 0.29989984870762215,
+      "grad_norm": 1.488936424255371,
+      "learning_rate": 4.5672886824039915e-05,
+      "loss": 1.6897,
+      "step": 7037
+    },
+    {
+      "epoch": 0.3012209934155852,
+      "grad_norm": 1.6014569997787476,
+      "learning_rate": 4.5625544749624435e-05,
+      "loss": 1.6821,
+      "step": 7068
+    },
+    {
+      "epoch": 0.30254213812354835,
+      "grad_norm": 1.5514070987701416,
+      "learning_rate": 4.5577969918739794e-05,
+      "loss": 1.7017,
+      "step": 7099
+    },
+    {
+      "epoch": 0.3038632828315114,
+      "grad_norm": 1.5641967058181763,
+      "learning_rate": 4.5530162868261486e-05,
+      "loss": 1.6809,
+      "step": 7130
+    },
+    {
+      "epoch": 0.3051844275394745,
+      "grad_norm": 1.5185332298278809,
+      "learning_rate": 4.548212413768558e-05,
+      "loss": 1.6719,
+      "step": 7161
+    },
+    {
+      "epoch": 0.30650557224743763,
+      "grad_norm": 1.497670292854309,
+      "learning_rate": 4.543385426912261e-05,
+      "loss": 1.6873,
+      "step": 7192
+    },
+    {
+      "epoch": 0.3078267169554007,
+      "grad_norm": 1.6817424297332764,
+      "learning_rate": 4.53853538072915e-05,
+      "loss": 1.668,
+      "step": 7223
+    },
+    {
+      "epoch": 0.30914786166336383,
+      "grad_norm": 1.561675786972046,
+      "learning_rate": 4.533662329951336e-05,
+      "loss": 1.678,
+      "step": 7254
+    },
+    {
+      "epoch": 0.3104690063713269,
+      "grad_norm": 1.7712374925613403,
+      "learning_rate": 4.528766329570536e-05,
+      "loss": 1.6817,
+      "step": 7285
+    },
+    {
+      "epoch": 0.31179015107929,
+      "grad_norm": 1.530474066734314,
+      "learning_rate": 4.523847434837447e-05,
+      "loss": 1.671,
+      "step": 7316
+    },
+    {
+      "epoch": 0.3131112957872531,
+      "grad_norm": 1.8106886148452759,
+      "learning_rate": 4.518905701261128e-05,
+      "loss": 1.6532,
+      "step": 7347
+    },
+    {
+      "epoch": 0.3144324404952162,
+      "grad_norm": 1.5035251379013062,
+      "learning_rate": 4.5139411846083715e-05,
+      "loss": 1.6651,
+      "step": 7378
+    },
+    {
+      "epoch": 0.31575358520317925,
+      "grad_norm": 1.6194860935211182,
+      "learning_rate": 4.508953940903073e-05,
+      "loss": 1.6664,
+      "step": 7409
+    },
+    {
+      "epoch": 0.3170747299111424,
+      "grad_norm": 1.5217773914337158,
+      "learning_rate": 4.5039440264255994e-05,
+      "loss": 1.6697,
+      "step": 7440
+    },
+    {
+      "epoch": 0.31839587461910546,
+      "grad_norm": 1.670386552810669,
+      "learning_rate": 4.498911497712155e-05,
+      "loss": 1.6637,
+      "step": 7471
+    },
+    {
+      "epoch": 0.31971701932706853,
+      "grad_norm": 1.680743932723999,
+      "learning_rate": 4.493856411554142e-05,
+      "loss": 1.6822,
+      "step": 7502
+    },
+    {
+      "epoch": 0.32103816403503166,
+      "grad_norm": 1.628322958946228,
+      "learning_rate": 4.4887788249975206e-05,
+      "loss": 1.6615,
+      "step": 7533
+    },
+    {
+      "epoch": 0.32235930874299473,
+      "grad_norm": 1.5849953889846802,
+      "learning_rate": 4.4836787953421656e-05,
+      "loss": 1.6637,
+      "step": 7564
+    },
+    {
+      "epoch": 0.3236804534509578,
+      "grad_norm": 1.5066392421722412,
+      "learning_rate": 4.478556380141218e-05,
+      "loss": 1.6839,
+      "step": 7595
+    },
+    {
+      "epoch": 0.32500159815892093,
+      "grad_norm": 1.6168526411056519,
+      "learning_rate": 4.4734116372004375e-05,
+      "loss": 1.6543,
+      "step": 7626
+    },
+    {
+      "epoch": 0.326322742866884,
+      "grad_norm": 1.5073281526565552,
+      "learning_rate": 4.4682446245775477e-05,
+      "loss": 1.6752,
+      "step": 7657
+    },
+    {
+      "epoch": 0.32764388757484714,
+      "grad_norm": 1.5694515705108643,
+      "learning_rate": 4.463055400581586e-05,
+      "loss": 1.6646,
+      "step": 7688
+    },
+    {
+      "epoch": 0.3289650322828102,
+      "grad_norm": 1.5031605958938599,
+      "learning_rate": 4.4578440237722374e-05,
+      "loss": 1.6807,
+      "step": 7719
+    },
+    {
+      "epoch": 0.3302861769907733,
+      "grad_norm": 1.4886877536773682,
+      "learning_rate": 4.452610552959183e-05,
+      "loss": 1.6644,
+      "step": 7750
+    },
+    {
+      "epoch": 0.3316073216987364,
+      "grad_norm": 1.5978937149047852,
+      "learning_rate": 4.447355047201428e-05,
+      "loss": 1.6573,
+      "step": 7781
+    },
+    {
+      "epoch": 0.3329284664066995,
+      "grad_norm": 1.4938329458236694,
+      "learning_rate": 4.4420775658066414e-05,
+      "loss": 1.6489,
+      "step": 7812
+    },
+    {
+      "epoch": 0.33424961111466256,
+      "grad_norm": 1.541128396987915,
+      "learning_rate": 4.436778168330484e-05,
+      "loss": 1.6673,
+      "step": 7843
+    },
+    {
+      "epoch": 0.3355707558226257,
+      "grad_norm": 1.5185847282409668,
+      "learning_rate": 4.4314569145759353e-05,
+      "loss": 1.6631,
+      "step": 7874
+    },
+    {
+      "epoch": 0.33689190053058876,
+      "grad_norm": 1.5545519590377808,
+      "learning_rate": 4.42611386459262e-05,
+      "loss": 1.6535,
+      "step": 7905
+    },
+    {
+      "epoch": 0.33821304523855183,
+      "grad_norm": 1.5892200469970703,
+      "learning_rate": 4.420749078676133e-05,
+      "loss": 1.6574,
+      "step": 7936
+    },
+    {
+      "epoch": 0.33953418994651496,
+      "grad_norm": 1.5779627561569214,
+      "learning_rate": 4.4153626173673516e-05,
+      "loss": 1.6683,
+      "step": 7967
+    },
+    {
+      "epoch": 0.34085533465447804,
+      "grad_norm": 1.5521479845046997,
+      "learning_rate": 4.409954541451762e-05,
+      "loss": 1.6599,
+      "step": 7998
+    },
+    {
+      "epoch": 0.3421764793624411,
+      "grad_norm": 1.4845163822174072,
+      "learning_rate": 4.404524911958764e-05,
+      "loss": 1.6357,
+      "step": 8029
+    },
+    {
+      "epoch": 0.34349762407040424,
+      "grad_norm": 1.4665591716766357,
+      "learning_rate": 4.399073790160989e-05,
+      "loss": 1.6577,
+      "step": 8060
+    },
+    {
+      "epoch": 0.3448187687783673,
+      "grad_norm": 1.4990118741989136,
+      "learning_rate": 4.393601237573607e-05,
+      "loss": 1.6589,
+      "step": 8091
+    },
+    {
+      "epoch": 0.34613991348633044,
+      "grad_norm": 1.6129504442214966,
+      "learning_rate": 4.388107315953628e-05,
+      "loss": 1.6719,
+      "step": 8122
+    },
+    {
+      "epoch": 0.3474610581942935,
+      "grad_norm": 1.5837408304214478,
+      "learning_rate": 4.382592087299212e-05,
+      "loss": 1.6389,
+      "step": 8153
+    },
+    {
+      "epoch": 0.3487822029022566,
+      "grad_norm": 1.4611998796463013,
+      "learning_rate": 4.377055613848964e-05,
+      "loss": 1.6568,
+      "step": 8184
+    },
+    {
+      "epoch": 0.3501033476102197,
+      "grad_norm": 1.5456607341766357,
+      "learning_rate": 4.3714979580812355e-05,
+      "loss": 1.6442,
+      "step": 8215
+    },
+    {
+      "epoch": 0.3514244923181828,
+      "grad_norm": 1.5393562316894531,
+      "learning_rate": 4.365919182713416e-05,
+      "loss": 1.6505,
+      "step": 8246
+    },
+    {
+      "epoch": 0.35274563702614586,
+      "grad_norm": 1.5040944814682007,
+      "learning_rate": 4.360319350701226e-05,
+      "loss": 1.6701,
+      "step": 8277
+    },
+    {
+      "epoch": 0.354066781734109,
+      "grad_norm": 1.5542012453079224,
+      "learning_rate": 4.3546985252380115e-05,
+      "loss": 1.6431,
+      "step": 8308
+    },
+    {
+      "epoch": 0.35538792644207207,
+      "grad_norm": 1.4878984689712524,
+      "learning_rate": 4.349056769754021e-05,
+      "loss": 1.6543,
+      "step": 8339
+    },
+    {
+      "epoch": 0.35670907115003514,
+      "grad_norm": 1.6049541234970093,
+      "learning_rate": 4.3433941479156994e-05,
+      "loss": 1.6415,
+      "step": 8370
+    },
+    {
+      "epoch": 0.35803021585799827,
+      "grad_norm": 1.6094954013824463,
+      "learning_rate": 4.3377107236249647e-05,
+      "loss": 1.646,
+      "step": 8401
+    },
+    {
+      "epoch": 0.35935136056596134,
+      "grad_norm": 1.5165326595306396,
+      "learning_rate": 4.332006561018488e-05,
+      "loss": 1.6575,
+      "step": 8432
+    },
+    {
+      "epoch": 0.3606725052739244,
+      "grad_norm": 1.5776197910308838,
+      "learning_rate": 4.3262817244669683e-05,
+      "loss": 1.6556,
+      "step": 8463
+    },
+    {
+      "epoch": 0.36199364998188754,
+      "grad_norm": 1.4522629976272583,
+      "learning_rate": 4.3205362785744083e-05,
+      "loss": 1.6583,
+      "step": 8494
+    },
+    {
+      "epoch": 0.3633147946898506,
+      "grad_norm": 1.4910801649093628,
+      "learning_rate": 4.314770288177384e-05,
+      "loss": 1.638,
+      "step": 8525
+    },
+    {
+      "epoch": 0.3646359393978137,
+      "grad_norm": 1.6230814456939697,
+      "learning_rate": 4.308983818344313e-05,
+      "loss": 1.6644,
+      "step": 8556
+    },
+    {
+      "epoch": 0.3659570841057768,
+      "grad_norm": 1.6548675298690796,
+      "learning_rate": 4.3031769343747206e-05,
+      "loss": 1.639,
+      "step": 8587
+    },
+    {
+      "epoch": 0.3672782288137399,
+      "grad_norm": 1.5451959371566772,
+      "learning_rate": 4.297349701798505e-05,
+      "loss": 1.6452,
+      "step": 8618
+    },
+    {
+      "epoch": 0.368599373521703,
+      "grad_norm": 1.46851646900177,
+      "learning_rate": 4.2915021863751916e-05,
+      "loss": 1.6645,
+      "step": 8649
+    },
+    {
+      "epoch": 0.3699205182296661,
+      "grad_norm": 1.4022047519683838,
+      "learning_rate": 4.285634454093198e-05,
+      "loss": 1.6556,
+      "step": 8680
+    },
+    {
+      "epoch": 0.37124166293762917,
+      "grad_norm": 1.4209098815917969,
+      "learning_rate": 4.279746571169086e-05,
+      "loss": 1.6699,
+      "step": 8711
+    },
+    {
+      "epoch": 0.3725628076455923,
+      "grad_norm": 1.4034055471420288,
+      "learning_rate": 4.2738386040468136e-05,
+      "loss": 1.6491,
+      "step": 8742
+    },
+    {
+      "epoch": 0.37388395235355537,
+      "grad_norm": 1.4482676982879639,
+      "learning_rate": 4.2679106193969866e-05,
+      "loss": 1.6299,
+      "step": 8773
+    },
+    {
+      "epoch": 0.37520509706151844,
+      "grad_norm": 1.542325496673584,
+      "learning_rate": 4.261962684116106e-05,
+      "loss": 1.6666,
+      "step": 8804
+    },
+    {
+      "epoch": 0.3765262417694816,
+      "grad_norm": 1.4821662902832031,
+      "learning_rate": 4.2559948653258145e-05,
+      "loss": 1.6383,
+      "step": 8835
+    },
+    {
+      "epoch": 0.37784738647744465,
+      "grad_norm": 1.445188283920288,
+      "learning_rate": 4.250007230372134e-05,
+      "loss": 1.6287,
+      "step": 8866
+    },
+    {
+      "epoch": 0.3791685311854077,
+      "grad_norm": 1.5624685287475586,
+      "learning_rate": 4.2439998468247126e-05,
+      "loss": 1.6337,
+      "step": 8897
+    },
+    {
+      "epoch": 0.38048967589337085,
+      "grad_norm": 1.7001335620880127,
+      "learning_rate": 4.2379727824760566e-05,
+      "loss": 1.6352,
+      "step": 8928
+    },
+    {
+      "epoch": 0.3818108206013339,
+      "grad_norm": 1.5825940370559692,
+      "learning_rate": 4.231926105340768e-05,
+      "loss": 1.6336,
+      "step": 8959
+    },
+    {
+      "epoch": 0.383131965309297,
+      "grad_norm": 1.4825541973114014,
+      "learning_rate": 4.225859883654776e-05,
+      "loss": 1.6359,
+      "step": 8990
+    },
+    {
+      "epoch": 0.3844531100172601,
+      "grad_norm": 1.4914085865020752,
+      "learning_rate": 4.219774185874569e-05,
+      "loss": 1.6623,
+      "step": 9021
+    },
+    {
+      "epoch": 0.3857742547252232,
+      "grad_norm": 1.491519570350647,
+      "learning_rate": 4.213669080676418e-05,
+      "loss": 1.6543,
+      "step": 9052
+    },
+    {
+      "epoch": 0.3870953994331863,
+      "grad_norm": 1.5370805263519287,
+      "learning_rate": 4.2075446369556056e-05,
+      "loss": 1.6325,
+      "step": 9083
+    },
+    {
+      "epoch": 0.3884165441411494,
+      "grad_norm": 1.423508644104004,
+      "learning_rate": 4.201400923825648e-05,
+      "loss": 1.6547,
+      "step": 9114
+    },
+    {
+      "epoch": 0.3897376888491125,
+      "grad_norm": 1.4396551847457886,
+      "learning_rate": 4.195238010617511e-05,
+      "loss": 1.6308,
+      "step": 9145
+    },
+    {
+      "epoch": 0.3910588335570756,
+      "grad_norm": 1.466518521308899,
+      "learning_rate": 4.1890559668788344e-05,
+      "loss": 1.6321,
+      "step": 9176
+    },
+    {
+      "epoch": 0.3923799782650387,
+      "grad_norm": 1.5302077531814575,
+      "learning_rate": 4.1828548623731405e-05,
+      "loss": 1.6455,
+      "step": 9207
+    },
+    {
+      "epoch": 0.39370112297300175,
+      "grad_norm": 1.5934138298034668,
+      "learning_rate": 4.1766347670790506e-05,
+      "loss": 1.6526,
+      "step": 9238
+    },
+    {
+      "epoch": 0.3950222676809649,
+      "grad_norm": 1.5880807638168335,
+      "learning_rate": 4.170395751189495e-05,
+      "loss": 1.6476,
+      "step": 9269
+    },
+    {
+      "epoch": 0.39634341238892795,
+      "grad_norm": 1.4420969486236572,
+      "learning_rate": 4.164137885110921e-05,
+      "loss": 1.6265,
+      "step": 9300
+    },
+    {
+      "epoch": 0.397664557096891,
+      "grad_norm": 1.489991307258606,
+      "learning_rate": 4.157861239462495e-05,
+      "loss": 1.647,
+      "step": 9331
+    },
+    {
+      "epoch": 0.39898570180485415,
+      "grad_norm": 1.5055150985717773,
+      "learning_rate": 4.1515658850753114e-05,
+      "loss": 1.6323,
+      "step": 9362
+    },
+    {
+      "epoch": 0.4003068465128172,
+      "grad_norm": 1.4316349029541016,
+      "learning_rate": 4.145251892991588e-05,
+      "loss": 1.6329,
+      "step": 9393
+    },
+    {
+      "epoch": 0.4016279912207803,
+      "grad_norm": 1.4182207584381104,
+      "learning_rate": 4.138919334463868e-05,
+      "loss": 1.6479,
+      "step": 9424
+    },
+    {
+      "epoch": 0.40294913592874343,
+      "grad_norm": 1.5128748416900635,
+      "learning_rate": 4.1325682809542124e-05,
+      "loss": 1.6419,
+      "step": 9455
+    },
+    {
+      "epoch": 0.4042702806367065,
+      "grad_norm": 1.4093265533447266,
+      "learning_rate": 4.126198804133398e-05,
+      "loss": 1.6128,
+      "step": 9486
+    },
+    {
+      "epoch": 0.40559142534466963,
+      "grad_norm": 1.4173692464828491,
+      "learning_rate": 4.1198109758801055e-05,
+      "loss": 1.6437,
+      "step": 9517
+    },
+    {
+      "epoch": 0.4069125700526327,
+      "grad_norm": 1.6048704385757446,
+      "learning_rate": 4.113404868280107e-05,
+      "loss": 1.6203,
+      "step": 9548
+    },
+    {
+      "epoch": 0.4082337147605958,
+      "grad_norm": 1.495764970779419,
+      "learning_rate": 4.106980553625457e-05,
+      "loss": 1.6294,
+      "step": 9579
+    },
+    {
+      "epoch": 0.4095548594685589,
+      "grad_norm": 1.530218482017517,
+      "learning_rate": 4.100538104413674e-05,
+      "loss": 1.6377,
+      "step": 9610
+    },
+    {
+      "epoch": 0.410876004176522,
+      "grad_norm": 1.5626767873764038,
+      "learning_rate": 4.09407759334692e-05,
+      "loss": 1.6293,
+      "step": 9641
+    },
+    {
+      "epoch": 0.41219714888448505,
+      "grad_norm": 1.3259074687957764,
+      "learning_rate": 4.087599093331186e-05,
+      "loss": 1.6209,
+      "step": 9672
+    },
+    {
+      "epoch": 0.4135182935924482,
+      "grad_norm": 1.4668447971343994,
+      "learning_rate": 4.081102677475462e-05,
+      "loss": 1.607,
+      "step": 9703
+    },
+    {
+      "epoch": 0.41483943830041126,
+      "grad_norm": 1.6264866590499878,
+      "learning_rate": 4.0745884190909194e-05,
+      "loss": 1.647,
+      "step": 9734
+    },
+    {
+      "epoch": 0.41616058300837433,
+      "grad_norm": 1.4699751138687134,
+      "learning_rate": 4.0680563916900796e-05,
+      "loss": 1.6433,
+      "step": 9765
+    },
+    {
+      "epoch": 0.41748172771633746,
+      "grad_norm": 1.5570849180221558,
+      "learning_rate": 4.0615066689859815e-05,
+      "loss": 1.6278,
+      "step": 9796
+    },
+    {
+      "epoch": 0.41880287242430053,
+      "grad_norm": 1.5093053579330444,
+      "learning_rate": 4.0549393248913584e-05,
+      "loss": 1.625,
+      "step": 9827
+    },
+    {
+      "epoch": 0.4201240171322636,
+      "grad_norm": 1.4233616590499878,
+      "learning_rate": 4.048354433517794e-05,
+      "loss": 1.6254,
+      "step": 9858
+    },
+    {
+      "epoch": 0.42144516184022673,
+      "grad_norm": 1.41616952419281,
+      "learning_rate": 4.0417520691748916e-05,
+      "loss": 1.6339,
+      "step": 9889
+    },
+    {
+      "epoch": 0.4227663065481898,
+      "grad_norm": 1.5434125661849976,
+      "learning_rate": 4.035132306369438e-05,
+      "loss": 1.6379,
+      "step": 9920
+    },
+    {
+      "epoch": 0.42408745125615294,
+      "grad_norm": 1.4518119096755981,
+      "learning_rate": 4.028495219804555e-05,
+      "loss": 1.6384,
+      "step": 9951
+    },
+    {
+      "epoch": 0.425408595964116,
+      "grad_norm": 1.4989590644836426,
+      "learning_rate": 4.021840884378864e-05,
+      "loss": 1.615,
+      "step": 9982
+    },
+    {
+      "epoch": 0.4267297406720791,
+      "grad_norm": 1.4292545318603516,
+      "learning_rate": 4.015169375185633e-05,
+      "loss": 1.6381,
+      "step": 10013
+    },
+    {
+      "epoch": 0.4280508853800422,
+      "grad_norm": 1.4783132076263428,
+      "learning_rate": 4.0084807675119396e-05,
+      "loss": 1.634,
+      "step": 10044
+    },
+    {
+      "epoch": 0.4293720300880053,
+      "grad_norm": 1.4413528442382812,
+      "learning_rate": 4.0017751368378106e-05,
+      "loss": 1.6356,
+      "step": 10075
+    },
+    {
+      "epoch": 0.43069317479596836,
+      "grad_norm": 1.4603348970413208,
+      "learning_rate": 3.995052558835377e-05,
+      "loss": 1.6248,
+      "step": 10106
+    },
+    {
+      "epoch": 0.4320143195039315,
+      "grad_norm": 1.5041816234588623,
+      "learning_rate": 3.988313109368017e-05,
+      "loss": 1.6352,
+      "step": 10137
+    },
+    {
+      "epoch": 0.43333546421189456,
+      "grad_norm": 1.483124852180481,
+      "learning_rate": 3.981556864489504e-05,
+      "loss": 1.6129,
+      "step": 10168
+    },
+    {
+      "epoch": 0.43465660891985763,
+      "grad_norm": 1.4493173360824585,
+      "learning_rate": 3.974783900443142e-05,
+      "loss": 1.624,
+      "step": 10199
+    },
+    {
+      "epoch": 0.43597775362782076,
+      "grad_norm": 1.553686499595642,
+      "learning_rate": 3.9679942936609095e-05,
+      "loss": 1.6099,
+      "step": 10230
+    },
+    {
+      "epoch": 0.43729889833578384,
+      "grad_norm": 1.4004040956497192,
+      "learning_rate": 3.961188120762596e-05,
+      "loss": 1.6152,
+      "step": 10261
+    },
+    {
+      "epoch": 0.4386200430437469,
+      "grad_norm": 1.4119765758514404,
+      "learning_rate": 3.954365458554938e-05,
+      "loss": 1.6161,
+      "step": 10292
+    },
+    {
+      "epoch": 0.43994118775171004,
+      "grad_norm": 1.4605529308319092,
+      "learning_rate": 3.947526384030751e-05,
+      "loss": 1.6234,
+      "step": 10323
+    },
+    {
+      "epoch": 0.4412623324596731,
+      "grad_norm": 1.5051487684249878,
+      "learning_rate": 3.9406709743680624e-05,
+      "loss": 1.6126,
+      "step": 10354
+    },
+    {
+      "epoch": 0.44258347716763624,
+      "grad_norm": 1.431213617324829,
+      "learning_rate": 3.9337993069292366e-05,
+      "loss": 1.6304,
+      "step": 10385
+    },
+    {
+      "epoch": 0.4439046218755993,
+      "grad_norm": 1.3373912572860718,
+      "learning_rate": 3.926911459260109e-05,
+      "loss": 1.6243,
+      "step": 10416
+    },
+    {
+      "epoch": 0.4452257665835624,
+      "grad_norm": 1.4510406255722046,
+      "learning_rate": 3.920007509089102e-05,
+      "loss": 1.6163,
+      "step": 10447
+    },
+    {
+      "epoch": 0.4465469112915255,
+      "grad_norm": 1.465527057647705,
+      "learning_rate": 3.913087534326357e-05,
+      "loss": 1.6092,
+      "step": 10478
+    },
+    {
+      "epoch": 0.4478680559994886,
+      "grad_norm": 1.4679442644119263,
+      "learning_rate": 3.9061516130628475e-05,
+      "loss": 1.617,
+      "step": 10509
+    },
+    {
+      "epoch": 0.44918920070745166,
+      "grad_norm": 1.5881776809692383,
+      "learning_rate": 3.8991998235695025e-05,
+      "loss": 1.615,
+      "step": 10540
+    },
+    {
+      "epoch": 0.4505103454154148,
+      "grad_norm": 1.4944725036621094,
+      "learning_rate": 3.8922322442963224e-05,
+      "loss": 1.6299,
+      "step": 10571
+    },
+    {
+      "epoch": 0.45183149012337787,
+      "grad_norm": 1.499192714691162,
+      "learning_rate": 3.885248953871491e-05,
+      "loss": 1.6173,
+      "step": 10602
+    },
+    {
+      "epoch": 0.45315263483134094,
+      "grad_norm": 1.409706473350525,
+      "learning_rate": 3.8782500311004915e-05,
+      "loss": 1.6083,
+      "step": 10633
+    },
+    {
+      "epoch": 0.45447377953930407,
+      "grad_norm": 1.3845860958099365,
+      "learning_rate": 3.871235554965218e-05,
+      "loss": 1.6104,
+      "step": 10664
+    },
+    {
+      "epoch": 0.45579492424726714,
+      "grad_norm": 1.4976452589035034,
+      "learning_rate": 3.864205604623078e-05,
+      "loss": 1.6123,
+      "step": 10695
+    },
+    {
+      "epoch": 0.4571160689552302,
+      "grad_norm": 1.496371865272522,
+      "learning_rate": 3.857160259406107e-05,
+      "loss": 1.594,
+      "step": 10726
+    },
+    {
+      "epoch": 0.45843721366319334,
+      "grad_norm": 1.364917516708374,
+      "learning_rate": 3.8500995988200674e-05,
+      "loss": 1.602,
+      "step": 10757
+    },
+    {
+      "epoch": 0.4597583583711564,
+      "grad_norm": 1.461065411567688,
+      "learning_rate": 3.843023702543556e-05,
+      "loss": 1.5936,
+      "step": 10788
+    },
+    {
+      "epoch": 0.46107950307911955,
+      "grad_norm": 1.422227382659912,
+      "learning_rate": 3.8359326504270984e-05,
+      "loss": 1.6275,
+      "step": 10819
+    },
+    {
+      "epoch": 0.4624006477870826,
+      "grad_norm": 1.3532990217208862,
+      "learning_rate": 3.828826522492255e-05,
+      "loss": 1.5982,
+      "step": 10850
+    },
+    {
+      "epoch": 0.4637217924950457,
+      "grad_norm": 1.6219929456710815,
+      "learning_rate": 3.821705398930713e-05,
+      "loss": 1.6225,
+      "step": 10881
+    },
+    {
+      "epoch": 0.4650429372030088,
+      "grad_norm": 1.4138596057891846,
+      "learning_rate": 3.814569360103385e-05,
+      "loss": 1.6216,
+      "step": 10912
+    },
+    {
+      "epoch": 0.4663640819109719,
+      "grad_norm": 1.492357611656189,
+      "learning_rate": 3.807418486539499e-05,
+      "loss": 1.6049,
+      "step": 10943
+    },
+    {
+      "epoch": 0.46768522661893497,
+      "grad_norm": 1.4756381511688232,
+      "learning_rate": 3.80025285893569e-05,
+      "loss": 1.6077,
+      "step": 10974
+    },
+    {
+      "epoch": 0.4690063713268981,
+      "grad_norm": 1.4480361938476562,
+      "learning_rate": 3.793072558155093e-05,
+      "loss": 1.6131,
+      "step": 11005
+    },
+    {
+      "epoch": 0.47032751603486117,
+      "grad_norm": 1.5386732816696167,
+      "learning_rate": 3.785877665226426e-05,
+      "loss": 1.6124,
+      "step": 11036
+    },
+    {
+      "epoch": 0.47164866074282424,
+      "grad_norm": 1.4616602659225464,
+      "learning_rate": 3.778668261343079e-05,
+      "loss": 1.6177,
+      "step": 11067
+    },
+    {
+      "epoch": 0.4729698054507874,
+      "grad_norm": 1.4740066528320312,
+      "learning_rate": 3.771444427862192e-05,
+      "loss": 1.6194,
+      "step": 11098
+    },
+    {
+      "epoch": 0.47429095015875045,
+      "grad_norm": 1.3635661602020264,
+      "learning_rate": 3.7642062463037465e-05,
+      "loss": 1.6085,
+      "step": 11129
+    },
+    {
+      "epoch": 0.4756120948667135,
+      "grad_norm": 1.4568270444869995,
+      "learning_rate": 3.7569537983496373e-05,
+      "loss": 1.6171,
+      "step": 11160
+    },
+    {
+      "epoch": 0.47693323957467665,
+      "grad_norm": 1.4151735305786133,
+      "learning_rate": 3.749687165842753e-05,
+      "loss": 1.6176,
+      "step": 11191
+    },
+    {
+      "epoch": 0.4782543842826397,
+      "grad_norm": 1.3976393938064575,
+      "learning_rate": 3.7424064307860536e-05,
+      "loss": 1.6326,
+      "step": 11222
+    },
+    {
+      "epoch": 0.47957552899060285,
+      "grad_norm": 1.4214377403259277,
+      "learning_rate": 3.735111675341645e-05,
+      "loss": 1.6353,
+      "step": 11253
+    },
+    {
+      "epoch": 0.4808966736985659,
+      "grad_norm": 1.453192949295044,
+      "learning_rate": 3.7278029818298524e-05,
+      "loss": 1.6225,
+      "step": 11284
+    },
+    {
+      "epoch": 0.482217818406529,
+      "grad_norm": 1.3557348251342773,
+      "learning_rate": 3.720480432728287e-05,
+      "loss": 1.6195,
+      "step": 11315
+    },
+    {
+      "epoch": 0.4835389631144921,
+      "grad_norm": 1.411064863204956,
+      "learning_rate": 3.71314411067092e-05,
+      "loss": 1.6026,
+      "step": 11346
+    },
+    {
+      "epoch": 0.4848601078224552,
+      "grad_norm": 1.4603424072265625,
+      "learning_rate": 3.70579409844715e-05,
+      "loss": 1.6133,
+      "step": 11377
+    },
+    {
+      "epoch": 0.4861812525304183,
+      "grad_norm": 1.484947681427002,
+      "learning_rate": 3.698430479000865e-05,
+      "loss": 1.5991,
+      "step": 11408
+    },
+    {
+      "epoch": 0.4875023972383814,
+      "grad_norm": 1.4569475650787354,
+      "learning_rate": 3.691053335429509e-05,
+      "loss": 1.6016,
+      "step": 11439
+    },
+    {
+      "epoch": 0.4888235419463445,
+      "grad_norm": 1.3495190143585205,
+      "learning_rate": 3.683662750983147e-05,
+      "loss": 1.6119,
+      "step": 11470
+    },
+    {
+      "epoch": 0.49014468665430755,
+      "grad_norm": 1.481056809425354,
+      "learning_rate": 3.676258809063518e-05,
+      "loss": 1.5983,
+      "step": 11501
+    },
+    {
+      "epoch": 0.4914658313622707,
+      "grad_norm": 1.4746501445770264,
+      "learning_rate": 3.6688415932231004e-05,
+      "loss": 1.624,
+      "step": 11532
+    },
+    {
+      "epoch": 0.49278697607023375,
+      "grad_norm": 1.340066909790039,
+      "learning_rate": 3.661411187164166e-05,
+      "loss": 1.5697,
+      "step": 11563
+    },
+    {
+      "epoch": 0.4941081207781968,
+      "grad_norm": 1.3957117795944214,
+      "learning_rate": 3.65396767473784e-05,
+      "loss": 1.608,
+      "step": 11594
+    },
+    {
+      "epoch": 0.49542926548615995,
+      "grad_norm": 1.448289155960083,
+      "learning_rate": 3.6465111399431465e-05,
+      "loss": 1.6024,
+      "step": 11625
+    },
+    {
+      "epoch": 0.496750410194123,
+      "grad_norm": 1.3630342483520508,
+      "learning_rate": 3.6390416669260674e-05,
+      "loss": 1.5887,
+      "step": 11656
+    },
+    {
+      "epoch": 0.49807155490208616,
+      "grad_norm": 1.4242172241210938,
+      "learning_rate": 3.63155933997859e-05,
+      "loss": 1.5989,
+      "step": 11687
+    },
+    {
+      "epoch": 0.49939269961004923,
+      "grad_norm": 1.4106690883636475,
+      "learning_rate": 3.624064243537758e-05,
+      "loss": 1.6133,
+      "step": 11718
+    },
+    {
+      "epoch": 0.5007138443180124,
+      "grad_norm": 1.3867900371551514,
+      "learning_rate": 3.616556462184716e-05,
+      "loss": 1.5766,
+      "step": 11749
+    },
+    {
+      "epoch": 0.5020349890259754,
+      "grad_norm": 1.4181773662567139,
+      "learning_rate": 3.609036080643755e-05,
+      "loss": 1.6259,
+      "step": 11780
+    },
+    {
+      "epoch": 0.5033561337339385,
+      "grad_norm": 1.434336543083191,
+      "learning_rate": 3.60150318378136e-05,
+      "loss": 1.6021,
+      "step": 11811
+    },
+    {
+      "epoch": 0.5046772784419016,
+      "grad_norm": 1.370049238204956,
+      "learning_rate": 3.5939578566052465e-05,
+      "loss": 1.5802,
+      "step": 11842
+    },
+    {
+      "epoch": 0.5059984231498647,
+      "grad_norm": 1.3584986925125122,
+      "learning_rate": 3.586400184263408e-05,
+      "loss": 1.5964,
+      "step": 11873
+    },
+    {
+      "epoch": 0.5073195678578278,
+      "grad_norm": 1.458137035369873,
+      "learning_rate": 3.578830252043148e-05,
+      "loss": 1.5962,
+      "step": 11904
+    },
+    {
+      "epoch": 0.5086407125657909,
+      "grad_norm": 1.4329200983047485,
+      "learning_rate": 3.571248145370125e-05,
+      "loss": 1.6046,
+      "step": 11935
+    },
+    {
+      "epoch": 0.509961857273754,
+      "grad_norm": 1.3814787864685059,
+      "learning_rate": 3.5636539498073794e-05,
+      "loss": 1.5901,
+      "step": 11966
+    },
+    {
+      "epoch": 0.5112830019817171,
+      "grad_norm": 1.3795075416564941,
+      "learning_rate": 3.556047751054378e-05,
+      "loss": 1.621,
+      "step": 11997
+    },
+    {
+      "epoch": 0.5126041466896801,
+      "grad_norm": 1.4559143781661987,
+      "learning_rate": 3.548429634946039e-05,
+      "loss": 1.5845,
+      "step": 12028
+    },
+    {
+      "epoch": 0.5139252913976432,
+      "grad_norm": 1.468773365020752,
+      "learning_rate": 3.540799687451768e-05,
+      "loss": 1.5956,
+      "step": 12059
+    },
+    {
+      "epoch": 0.5152464361056064,
+      "grad_norm": 1.5287885665893555,
+      "learning_rate": 3.533157994674485e-05,
+      "loss": 1.5801,
+      "step": 12090
+    },
+    {
+      "epoch": 0.5165675808135695,
+      "grad_norm": 1.4891117811203003,
+      "learning_rate": 3.5255046428496546e-05,
+      "loss": 1.6034,
+      "step": 12121
+    },
+    {
+      "epoch": 0.5178887255215325,
+      "grad_norm": 1.541502594947815,
+      "learning_rate": 3.517839718344311e-05,
+      "loss": 1.6076,
+      "step": 12152
+    },
+    {
+      "epoch": 0.5192098702294956,
+      "grad_norm": 1.3209787607192993,
+      "learning_rate": 3.510163307656086e-05,
+      "loss": 1.5967,
+      "step": 12183
+    }
+  ],
+  "logging_steps": 31,
+  "max_steps": 30517,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 2,
+  "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": 9.055780216089084e+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 sha256:302cac6bc00d87f6c817113f3074af210729435b710bdeed411138aae3385920
+size 5304

checkpoint-15260/config.json

@@ -0,0 +1,39 @@
+{
+  "_name_or_path": "/mnt/parscratch/users/acp23ay/private/models/Llama-3.1-8B-Instruct-gu-madlad-mean/",
+  "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": 138256
+}

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 sha256:3d7a2c17c2a4ad937e36df2c523fa0a84d6f5d0c768be98a61c3cfd14dc4e110
+size 4983197184

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

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:168629c67732309d49f60a5ec48a6d160212bc987365e82e183dfbf74ba0c1f3
+size 4899116432

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 sha256:6c10f85e5ba648187f4442f89b72f1cd21f148521978f49b0f0210c05d15eb77
+size 4999813120

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

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:16253ab34e94a3612effa3b8b8fd2246aead2b043b4c5d3d4189ef3e28f354ef
+size 2734998184

checkpoint-15260/model.safetensors.index.json

@@ -0,0 +1,298 @@
+{
+  "metadata": {
+    "total_size": 32448724992
+  },
+  "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-00002-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:aa6b1d44aa1a007a4cb207ccf2db06799fc58bb1876cf62286a78f309a2bcac7
+size 16040396334

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": 0.6503441368876388,
+  "eval_steps": 500,
+  "global_step": 15260,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.001321144707963093,
+      "grad_norm": 29.45369529724121,
+      "learning_rate": 1.0157273918741808e-06,
+      "loss": 9.356,
+      "step": 31
+    },
+    {
+      "epoch": 0.002642289415926186,
+      "grad_norm": 17.081205368041992,
+      "learning_rate": 2.0314547837483616e-06,
+      "loss": 7.7447,
+      "step": 62
+    },
+    {
+      "epoch": 0.00396343412388928,
+      "grad_norm": 17.875776290893555,
+      "learning_rate": 3.0471821756225426e-06,
+      "loss": 6.3787,
+      "step": 93
+    },
+    {
+      "epoch": 0.005284578831852372,
+      "grad_norm": 38.05577087402344,
+      "learning_rate": 4.062909567496723e-06,
+      "loss": 5.1973,
+      "step": 124
+    },
+    {
+      "epoch": 0.006605723539815466,
+      "grad_norm": 20.137739181518555,
+      "learning_rate": 5.078636959370905e-06,
+      "loss": 4.6404,
+      "step": 155
+    },
+    {
+      "epoch": 0.00792686824777856,
+      "grad_norm": 15.796778678894043,
+      "learning_rate": 6.094364351245085e-06,
+      "loss": 4.3168,
+      "step": 186
+    },
+    {
+      "epoch": 0.009248012955741652,
+      "grad_norm": 18.784709930419922,
+      "learning_rate": 7.110091743119267e-06,
+      "loss": 4.0553,
+      "step": 217
+    },
+    {
+      "epoch": 0.010569157663704745,
+      "grad_norm": 23.772998809814453,
+      "learning_rate": 8.125819134993446e-06,
+      "loss": 3.84,
+      "step": 248
+    },
+    {
+      "epoch": 0.011890302371667839,
+      "grad_norm": 14.564051628112793,
+      "learning_rate": 9.141546526867629e-06,
+      "loss": 3.7012,
+      "step": 279
+    },
+    {
+      "epoch": 0.013211447079630931,
+      "grad_norm": 14.735116004943848,
+      "learning_rate": 1.015727391874181e-05,
+      "loss": 3.5595,
+      "step": 310
+    },
+    {
+      "epoch": 0.014532591787594026,
+      "grad_norm": 13.88759708404541,
+      "learning_rate": 1.117300131061599e-05,
+      "loss": 3.4515,
+      "step": 341
+    },
+    {
+      "epoch": 0.01585373649555712,
+      "grad_norm": 10.884685516357422,
+      "learning_rate": 1.218872870249017e-05,
+      "loss": 3.3299,
+      "step": 372
+    },
+    {
+      "epoch": 0.01717488120352021,
+      "grad_norm": 12.495129585266113,
+      "learning_rate": 1.3204456094364351e-05,
+      "loss": 3.2453,
+      "step": 403
+    },
+    {
+      "epoch": 0.018496025911483305,
+      "grad_norm": 11.930519104003906,
+      "learning_rate": 1.4220183486238533e-05,
+      "loss": 3.1447,
+      "step": 434
+    },
+    {
+      "epoch": 0.0198171706194464,
+      "grad_norm": 7.951854228973389,
+      "learning_rate": 1.5235910878112714e-05,
+      "loss": 3.0765,
+      "step": 465
+    },
+    {
+      "epoch": 0.02113831532740949,
+      "grad_norm": 8.87326717376709,
+      "learning_rate": 1.6251638269986893e-05,
+      "loss": 3.0296,
+      "step": 496
+    },
+    {
+      "epoch": 0.022459460035372584,
+      "grad_norm": 8.348654747009277,
+      "learning_rate": 1.7267365661861077e-05,
+      "loss": 2.9469,
+      "step": 527
+    },
+    {
+      "epoch": 0.023780604743335678,
+      "grad_norm": 7.966938495635986,
+      "learning_rate": 1.8283093053735257e-05,
+      "loss": 2.8941,
+      "step": 558
+    },
+    {
+      "epoch": 0.02510174945129877,
+      "grad_norm": 6.656867027282715,
+      "learning_rate": 1.9298820445609438e-05,
+      "loss": 2.8428,
+      "step": 589
+    },
+    {
+      "epoch": 0.026422894159261863,
+      "grad_norm": 6.973931312561035,
+      "learning_rate": 2.031454783748362e-05,
+      "loss": 2.7875,
+      "step": 620
+    },
+    {
+      "epoch": 0.027744038867224957,
+      "grad_norm": 6.903556823730469,
+      "learning_rate": 2.13302752293578e-05,
+      "loss": 2.7264,
+      "step": 651
+    },
+    {
+      "epoch": 0.02906518357518805,
+      "grad_norm": 7.422015190124512,
+      "learning_rate": 2.234600262123198e-05,
+      "loss": 2.6833,
+      "step": 682
+    },
+    {
+      "epoch": 0.030386328283151142,
+      "grad_norm": 6.663395404815674,
+      "learning_rate": 2.336173001310616e-05,
+      "loss": 2.6303,
+      "step": 713
+    },
+    {
+      "epoch": 0.03170747299111424,
+      "grad_norm": 6.743407249450684,
+      "learning_rate": 2.437745740498034e-05,
+      "loss": 2.578,
+      "step": 744
+    },
+    {
+      "epoch": 0.03302861769907733,
+      "grad_norm": 5.852759838104248,
+      "learning_rate": 2.5393184796854525e-05,
+      "loss": 2.5248,
+      "step": 775
+    },
+    {
+      "epoch": 0.03434976240704042,
+      "grad_norm": 5.186204433441162,
+      "learning_rate": 2.6408912188728702e-05,
+      "loss": 2.5051,
+      "step": 806
+    },
+    {
+      "epoch": 0.035670907115003515,
+      "grad_norm": 5.873666286468506,
+      "learning_rate": 2.7424639580602886e-05,
+      "loss": 2.4878,
+      "step": 837
+    },
+    {
+      "epoch": 0.03699205182296661,
+      "grad_norm": 6.269757270812988,
+      "learning_rate": 2.8440366972477066e-05,
+      "loss": 2.4272,
+      "step": 868
+    },
+    {
+      "epoch": 0.038313196530929704,
+      "grad_norm": 5.196881294250488,
+      "learning_rate": 2.9456094364351244e-05,
+      "loss": 2.3863,
+      "step": 899
+    },
+    {
+      "epoch": 0.0396343412388928,
+      "grad_norm": 4.899730682373047,
+      "learning_rate": 3.0471821756225428e-05,
+      "loss": 2.3714,
+      "step": 930
+    },
+    {
+      "epoch": 0.04095548594685589,
+      "grad_norm": 4.69920015335083,
+      "learning_rate": 3.148754914809961e-05,
+      "loss": 2.3456,
+      "step": 961
+    },
+    {
+      "epoch": 0.04227663065481898,
+      "grad_norm": 5.293018341064453,
+      "learning_rate": 3.2503276539973785e-05,
+      "loss": 2.3162,
+      "step": 992
+    },
+    {
+      "epoch": 0.043597775362782074,
+      "grad_norm": 4.992343902587891,
+      "learning_rate": 3.351900393184797e-05,
+      "loss": 2.2939,
+      "step": 1023
+    },
+    {
+      "epoch": 0.04491892007074517,
+      "grad_norm": 4.305331230163574,
+      "learning_rate": 3.453473132372215e-05,
+      "loss": 2.275,
+      "step": 1054
+    },
+    {
+      "epoch": 0.04624006477870826,
+      "grad_norm": 4.863211631774902,
+      "learning_rate": 3.555045871559633e-05,
+      "loss": 2.2618,
+      "step": 1085
+    },
+    {
+      "epoch": 0.047561209486671356,
+      "grad_norm": 4.6537251472473145,
+      "learning_rate": 3.6566186107470514e-05,
+      "loss": 2.2584,
+      "step": 1116
+    },
+    {
+      "epoch": 0.04888235419463445,
+      "grad_norm": 7.502182483673096,
+      "learning_rate": 3.7581913499344695e-05,
+      "loss": 2.244,
+      "step": 1147
+    },
+    {
+      "epoch": 0.05020349890259754,
+      "grad_norm": 4.560093879699707,
+      "learning_rate": 3.8597640891218876e-05,
+      "loss": 2.2054,
+      "step": 1178
+    },
+    {
+      "epoch": 0.05152464361056063,
+      "grad_norm": 4.21810245513916,
+      "learning_rate": 3.9613368283093056e-05,
+      "loss": 2.1923,
+      "step": 1209
+    },
+    {
+      "epoch": 0.052845788318523726,
+      "grad_norm": 3.815044403076172,
+      "learning_rate": 4.062909567496724e-05,
+      "loss": 2.1746,
+      "step": 1240
+    },
+    {
+      "epoch": 0.05416693302648682,
+      "grad_norm": 3.4870846271514893,
+      "learning_rate": 4.164482306684142e-05,
+      "loss": 2.1751,
+      "step": 1271
+    },
+    {
+      "epoch": 0.055488077734449914,
+      "grad_norm": 3.8318536281585693,
+      "learning_rate": 4.26605504587156e-05,
+      "loss": 2.1773,
+      "step": 1302
+    },
+    {
+      "epoch": 0.05680922244241301,
+      "grad_norm": 3.6395630836486816,
+      "learning_rate": 4.367627785058978e-05,
+      "loss": 2.1534,
+      "step": 1333
+    },
+    {
+      "epoch": 0.0581303671503761,
+      "grad_norm": 3.4870851039886475,
+      "learning_rate": 4.469200524246396e-05,
+      "loss": 2.1373,
+      "step": 1364
+    },
+    {
+      "epoch": 0.05945151185833919,
+      "grad_norm": 3.533464193344116,
+      "learning_rate": 4.570773263433814e-05,
+      "loss": 2.104,
+      "step": 1395
+    },
+    {
+      "epoch": 0.060772656566302284,
+      "grad_norm": 3.7605605125427246,
+      "learning_rate": 4.672346002621232e-05,
+      "loss": 2.1497,
+      "step": 1426
+    },
+    {
+      "epoch": 0.06209380127426538,
+      "grad_norm": 3.331261396408081,
+      "learning_rate": 4.77391874180865e-05,
+      "loss": 2.103,
+      "step": 1457
+    },
+    {
+      "epoch": 0.06341494598222848,
+      "grad_norm": 3.4388535022735596,
+      "learning_rate": 4.875491480996068e-05,
+      "loss": 2.1017,
+      "step": 1488
+    },
+    {
+      "epoch": 0.06473609069019157,
+      "grad_norm": 3.3586745262145996,
+      "learning_rate": 4.977064220183487e-05,
+      "loss": 2.0883,
+      "step": 1519
+    },
+    {
+      "epoch": 0.06605723539815465,
+      "grad_norm": 3.0805885791778564,
+      "learning_rate": 4.9999915451558777e-05,
+      "loss": 2.0741,
+      "step": 1550
+    },
+    {
+      "epoch": 0.06737838010611776,
+      "grad_norm": 3.1708438396453857,
+      "learning_rate": 4.999955597496219e-05,
+      "loss": 2.0718,
+      "step": 1581
+    },
+    {
+      "epoch": 0.06869952481408084,
+      "grad_norm": 3.03056001663208,
+      "learning_rate": 4.9998914381774255e-05,
+      "loss": 2.0673,
+      "step": 1612
+    },
+    {
+      "epoch": 0.07002066952204394,
+      "grad_norm": 2.7995362281799316,
+      "learning_rate": 4.999799067923527e-05,
+      "loss": 2.0683,
+      "step": 1643
+    },
+    {
+      "epoch": 0.07134181423000703,
+      "grad_norm": 3.116344451904297,
+      "learning_rate": 4.999678487776908e-05,
+      "loss": 2.0231,
+      "step": 1674
+    },
+    {
+      "epoch": 0.07266295893797013,
+      "grad_norm": 3.1621696949005127,
+      "learning_rate": 4.9995296990983006e-05,
+      "loss": 2.0391,
+      "step": 1705
+    },
+    {
+      "epoch": 0.07398410364593322,
+      "grad_norm": 2.8406624794006348,
+      "learning_rate": 4.999352703566763e-05,
+      "loss": 2.0132,
+      "step": 1736
+    },
+    {
+      "epoch": 0.0753052483538963,
+      "grad_norm": 2.570624351501465,
+      "learning_rate": 4.999147503179668e-05,
+      "loss": 2.0116,
+      "step": 1767
+    },
+    {
+      "epoch": 0.07662639306185941,
+      "grad_norm": 2.7663609981536865,
+      "learning_rate": 4.998914100252672e-05,
+      "loss": 2.0127,
+      "step": 1798
+    },
+    {
+      "epoch": 0.0779475377698225,
+      "grad_norm": 2.5649290084838867,
+      "learning_rate": 4.998652497419696e-05,
+      "loss": 2.0033,
+      "step": 1829
+    },
+    {
+      "epoch": 0.0792686824777856,
+      "grad_norm": 2.575486183166504,
+      "learning_rate": 4.9983626976328927e-05,
+      "loss": 2.005,
+      "step": 1860
+    },
+    {
+      "epoch": 0.08058982718574868,
+      "grad_norm": 2.509678840637207,
+      "learning_rate": 4.998044704162613e-05,
+      "loss": 1.9824,
+      "step": 1891
+    },
+    {
+      "epoch": 0.08191097189371178,
+      "grad_norm": 2.6347391605377197,
+      "learning_rate": 4.9976985205973705e-05,
+      "loss": 1.9571,
+      "step": 1922
+    },
+    {
+      "epoch": 0.08323211660167487,
+      "grad_norm": 2.5539751052856445,
+      "learning_rate": 4.997324150843799e-05,
+      "loss": 1.9782,
+      "step": 1953
+    },
+    {
+      "epoch": 0.08455326130963796,
+      "grad_norm": 10.463569641113281,
+      "learning_rate": 4.99692159912661e-05,
+      "loss": 1.9713,
+      "step": 1984
+    },
+    {
+      "epoch": 0.08587440601760106,
+      "grad_norm": 2.604071617126465,
+      "learning_rate": 4.996490869988546e-05,
+      "loss": 1.9769,
+      "step": 2015
+    },
+    {
+      "epoch": 0.08719555072556415,
+      "grad_norm": 2.4612362384796143,
+      "learning_rate": 4.996031968290326e-05,
+      "loss": 1.9463,
+      "step": 2046
+    },
+    {
+      "epoch": 0.08851669543352725,
+      "grad_norm": 2.539299488067627,
+      "learning_rate": 4.995544899210594e-05,
+      "loss": 1.9414,
+      "step": 2077
+    },
+    {
+      "epoch": 0.08983784014149034,
+      "grad_norm": 2.6650774478912354,
+      "learning_rate": 4.9950296682458583e-05,
+      "loss": 1.9371,
+      "step": 2108
+    },
+    {
+      "epoch": 0.09115898484945342,
+      "grad_norm": 2.5480291843414307,
+      "learning_rate": 4.994486281210429e-05,
+      "loss": 1.9581,
+      "step": 2139
+    },
+    {
+      "epoch": 0.09248012955741652,
+      "grad_norm": 2.4799551963806152,
+      "learning_rate": 4.9939147442363566e-05,
+      "loss": 1.9524,
+      "step": 2170
+    },
+    {
+      "epoch": 0.09380127426537961,
+      "grad_norm": 2.3326504230499268,
+      "learning_rate": 4.9933150637733574e-05,
+      "loss": 1.9319,
+      "step": 2201
+    },
+    {
+      "epoch": 0.09512241897334271,
+      "grad_norm": 2.3550055027008057,
+      "learning_rate": 4.992687246588743e-05,
+      "loss": 1.9015,
+      "step": 2232
+    },
+    {
+      "epoch": 0.0964435636813058,
+      "grad_norm": 2.422797918319702,
+      "learning_rate": 4.992031299767347e-05,
+      "loss": 1.9364,
+      "step": 2263
+    },
+    {
+      "epoch": 0.0977647083892689,
+      "grad_norm": 2.2776873111724854,
+      "learning_rate": 4.9913472307114386e-05,
+      "loss": 1.9102,
+      "step": 2294
+    },
+    {
+      "epoch": 0.09908585309723199,
+      "grad_norm": 2.3031086921691895,
+      "learning_rate": 4.9906350471406446e-05,
+      "loss": 1.9193,
+      "step": 2325
+    },
+    {
+      "epoch": 0.10040699780519508,
+      "grad_norm": 2.2281057834625244,
+      "learning_rate": 4.989894757091861e-05,
+      "loss": 1.9143,
+      "step": 2356
+    },
+    {
+      "epoch": 0.10172814251315818,
+      "grad_norm": 2.39752459526062,
+      "learning_rate": 4.989126368919158e-05,
+      "loss": 1.905,
+      "step": 2387
+    },
+    {
+      "epoch": 0.10304928722112126,
+      "grad_norm": 2.3098957538604736,
+      "learning_rate": 4.988329891293693e-05,
+      "loss": 1.9204,
+      "step": 2418
+    },
+    {
+      "epoch": 0.10437043192908436,
+      "grad_norm": 2.244088649749756,
+      "learning_rate": 4.987505333203608e-05,
+      "loss": 1.8836,
+      "step": 2449
+    },
+    {
+      "epoch": 0.10569157663704745,
+      "grad_norm": 2.3035812377929688,
+      "learning_rate": 4.9866527039539276e-05,
+      "loss": 1.9097,
+      "step": 2480
+    },
+    {
+      "epoch": 0.10701272134501055,
+      "grad_norm": 2.2249033451080322,
+      "learning_rate": 4.9857720131664594e-05,
+      "loss": 1.8804,
+      "step": 2511
+    },
+    {
+      "epoch": 0.10833386605297364,
+      "grad_norm": 2.180612325668335,
+      "learning_rate": 4.9848632707796773e-05,
+      "loss": 1.8848,
+      "step": 2542
+    },
+    {
+      "epoch": 0.10965501076093673,
+      "grad_norm": 2.2042863368988037,
+      "learning_rate": 4.9839264870486155e-05,
+      "loss": 1.8879,
+      "step": 2573
+    },
+    {
+      "epoch": 0.11097615546889983,
+      "grad_norm": 2.278650999069214,
+      "learning_rate": 4.9829616725447526e-05,
+      "loss": 1.8728,
+      "step": 2604
+    },
+    {
+      "epoch": 0.11229730017686292,
+      "grad_norm": 2.2380714416503906,
+      "learning_rate": 4.981968838155888e-05,
+      "loss": 1.8927,
+      "step": 2635
+    },
+    {
+      "epoch": 0.11361844488482602,
+      "grad_norm": 2.062289237976074,
+      "learning_rate": 4.980947995086024e-05,
+      "loss": 1.8706,
+      "step": 2666
+    },
+    {
+      "epoch": 0.1149395895927891,
+      "grad_norm": 2.169900894165039,
+      "learning_rate": 4.979899154855234e-05,
+      "loss": 1.8753,
+      "step": 2697
+    },
+    {
+      "epoch": 0.1162607343007522,
+      "grad_norm": 2.1450119018554688,
+      "learning_rate": 4.9788223292995386e-05,
+      "loss": 1.8685,
+      "step": 2728
+    },
+    {
+      "epoch": 0.11758187900871529,
+      "grad_norm": 2.285205125808716,
+      "learning_rate": 4.977717530570768e-05,
+      "loss": 1.8601,
+      "step": 2759
+    },
+    {
+      "epoch": 0.11890302371667838,
+      "grad_norm": 2.1305880546569824,
+      "learning_rate": 4.976584771136425e-05,
+      "loss": 1.8557,
+      "step": 2790
+    },
+    {
+      "epoch": 0.12022416842464148,
+      "grad_norm": 2.0517284870147705,
+      "learning_rate": 4.975424063779547e-05,
+      "loss": 1.8587,
+      "step": 2821
+    },
+    {
+      "epoch": 0.12154531313260457,
+      "grad_norm": 2.190729856491089,
+      "learning_rate": 4.974235421598557e-05,
+      "loss": 1.8431,
+      "step": 2852
+    },
+    {
+      "epoch": 0.12286645784056767,
+      "grad_norm": 2.133803367614746,
+      "learning_rate": 4.973018858007122e-05,
+      "loss": 1.8492,
+      "step": 2883
+    },
+    {
+      "epoch": 0.12418760254853076,
+      "grad_norm": 2.3807687759399414,
+      "learning_rate": 4.9717743867339963e-05,
+      "loss": 1.8391,
+      "step": 2914
+    },
+    {
+      "epoch": 0.12550874725649386,
+      "grad_norm": 2.0250234603881836,
+      "learning_rate": 4.9705020218228695e-05,
+      "loss": 1.8465,
+      "step": 2945
+    },
+    {
+      "epoch": 0.12682989196445696,
+      "grad_norm": 2.1557273864746094,
+      "learning_rate": 4.969201777632205e-05,
+      "loss": 1.8346,
+      "step": 2976
+    },
+    {
+      "epoch": 0.12815103667242003,
+      "grad_norm": 1.9695699214935303,
+      "learning_rate": 4.9678736688350846e-05,
+      "loss": 1.8423,
+      "step": 3007
+    },
+    {
+      "epoch": 0.12947218138038313,
+      "grad_norm": 2.1525843143463135,
+      "learning_rate": 4.966517710419033e-05,
+      "loss": 1.851,
+      "step": 3038
+    },
+    {
+      "epoch": 0.13079332608834623,
+      "grad_norm": 1.9812391996383667,
+      "learning_rate": 4.965133917685858e-05,
+      "loss": 1.8338,
+      "step": 3069
+    },
+    {
+      "epoch": 0.1321144707963093,
+      "grad_norm": 2.2273776531219482,
+      "learning_rate": 4.9637223062514714e-05,
+      "loss": 1.8329,
+      "step": 3100
+    },
+    {
+      "epoch": 0.1334356155042724,
+      "grad_norm": 2.1103451251983643,
+      "learning_rate": 4.962282892045718e-05,
+      "loss": 1.8243,
+      "step": 3131
+    },
+    {
+      "epoch": 0.1347567602122355,
+      "grad_norm": 2.083226203918457,
+      "learning_rate": 4.9608156913121904e-05,
+      "loss": 1.8254,
+      "step": 3162
+    },
+    {
+      "epoch": 0.1360779049201986,
+      "grad_norm": 1.878396987915039,
+      "learning_rate": 4.959320720608049e-05,
+      "loss": 1.819,
+      "step": 3193
+    },
+    {
+      "epoch": 0.13739904962816168,
+      "grad_norm": 1.9980756044387817,
+      "learning_rate": 4.9577979968038354e-05,
+      "loss": 1.8391,
+      "step": 3224
+    },
+    {
+      "epoch": 0.13872019433612479,
+      "grad_norm": 1.977419376373291,
+      "learning_rate": 4.956247537083282e-05,
+      "loss": 1.8233,
+      "step": 3255
+    },
+    {
+      "epoch": 0.1400413390440879,
+      "grad_norm": 1.9977376461029053,
+      "learning_rate": 4.9546693589431145e-05,
+      "loss": 1.8254,
+      "step": 3286
+    },
+    {
+      "epoch": 0.14136248375205096,
+      "grad_norm": 1.8682738542556763,
+      "learning_rate": 4.9530634801928595e-05,
+      "loss": 1.8154,
+      "step": 3317
+    },
+    {
+      "epoch": 0.14268362846001406,
+      "grad_norm": 1.9372395277023315,
+      "learning_rate": 4.9514299189546395e-05,
+      "loss": 1.8229,
+      "step": 3348
+    },
+    {
+      "epoch": 0.14400477316797716,
+      "grad_norm": 1.9475616216659546,
+      "learning_rate": 4.949768693662973e-05,
+      "loss": 1.8122,
+      "step": 3379
+    },
+    {
+      "epoch": 0.14532591787594026,
+      "grad_norm": 1.8302438259124756,
+      "learning_rate": 4.948079823064559e-05,
+      "loss": 1.8275,
+      "step": 3410
+    },
+    {
+      "epoch": 0.14664706258390334,
+      "grad_norm": 1.9195733070373535,
+      "learning_rate": 4.946363326218074e-05,
+      "loss": 1.8288,
+      "step": 3441
+    },
+    {
+      "epoch": 0.14796820729186644,
+      "grad_norm": 2.056960105895996,
+      "learning_rate": 4.9446192224939525e-05,
+      "loss": 1.8267,
+      "step": 3472
+    },
+    {
+      "epoch": 0.14928935199982954,
+      "grad_norm": 1.9093526601791382,
+      "learning_rate": 4.942847531574167e-05,
+      "loss": 1.8118,
+      "step": 3503
+    },
+    {
+      "epoch": 0.1506104967077926,
+      "grad_norm": 1.800930380821228,
+      "learning_rate": 4.941048273452008e-05,
+      "loss": 1.7869,
+      "step": 3534
+    },
+    {
+      "epoch": 0.1519316414157557,
+      "grad_norm": 2.0274600982666016,
+      "learning_rate": 4.9392214684318605e-05,
+      "loss": 1.7937,
+      "step": 3565
+    },
+    {
+      "epoch": 0.15325278612371881,
+      "grad_norm": 1.880557656288147,
+      "learning_rate": 4.93736713712897e-05,
+      "loss": 1.793,
+      "step": 3596
+    },
+    {
+      "epoch": 0.15457393083168192,
+      "grad_norm": 1.9970703125,
+      "learning_rate": 4.9354853004692124e-05,
+      "loss": 1.7922,
+      "step": 3627
+    },
+    {
+      "epoch": 0.155895075539645,
+      "grad_norm": 2.0239968299865723,
+      "learning_rate": 4.93357597968886e-05,
+      "loss": 1.8026,
+      "step": 3658
+    },
+    {
+      "epoch": 0.1572162202476081,
+      "grad_norm": 1.8837428092956543,
+      "learning_rate": 4.931639196334338e-05,
+      "loss": 1.7898,
+      "step": 3689
+    },
+    {
+      "epoch": 0.1585373649555712,
+      "grad_norm": 1.8433760404586792,
+      "learning_rate": 4.9296749722619826e-05,
+      "loss": 1.7854,
+      "step": 3720
+    },
+    {
+      "epoch": 0.15985850966353427,
+      "grad_norm": 1.898555874824524,
+      "learning_rate": 4.9276833296377966e-05,
+      "loss": 1.8011,
+      "step": 3751
+    },
+    {
+      "epoch": 0.16117965437149737,
+      "grad_norm": 1.9116350412368774,
+      "learning_rate": 4.925664290937196e-05,
+      "loss": 1.7759,
+      "step": 3782
+    },
+    {
+      "epoch": 0.16250079907946047,
+      "grad_norm": 1.8142945766448975,
+      "learning_rate": 4.9236178789447576e-05,
+      "loss": 1.7712,
+      "step": 3813
+    },
+    {
+      "epoch": 0.16382194378742357,
+      "grad_norm": 1.8500661849975586,
+      "learning_rate": 4.921544116753962e-05,
+      "loss": 1.7885,
+      "step": 3844
+    },
+    {
+      "epoch": 0.16514308849538664,
+      "grad_norm": 1.9167587757110596,
+      "learning_rate": 4.919443027766935e-05,
+      "loss": 1.792,
+      "step": 3875
+    },
+    {
+      "epoch": 0.16646423320334974,
+      "grad_norm": 1.891123652458191,
+      "learning_rate": 4.91731463569418e-05,
+      "loss": 1.7963,
+      "step": 3906
+    },
+    {
+      "epoch": 0.16778537791131284,
+      "grad_norm": 1.7814100980758667,
+      "learning_rate": 4.915158964554312e-05,
+      "loss": 1.7836,
+      "step": 3937
+    },
+    {
+      "epoch": 0.16910652261927592,
+      "grad_norm": 2.0991482734680176,
+      "learning_rate": 4.912976038673786e-05,
+      "loss": 1.7788,
+      "step": 3968
+    },
+    {
+      "epoch": 0.17042766732723902,
+      "grad_norm": 1.8836603164672852,
+      "learning_rate": 4.9107658826866254e-05,
+      "loss": 1.7885,
+      "step": 3999
+    },
+    {
+      "epoch": 0.17174881203520212,
+      "grad_norm": 1.8759151697158813,
+      "learning_rate": 4.908528521534139e-05,
+      "loss": 1.7622,
+      "step": 4030
+    },
+    {
+      "epoch": 0.17306995674316522,
+      "grad_norm": 1.9080370664596558,
+      "learning_rate": 4.906263980464644e-05,
+      "loss": 1.7778,
+      "step": 4061
+    },
+    {
+      "epoch": 0.1743911014511283,
+      "grad_norm": 1.776382327079773,
+      "learning_rate": 4.903972285033178e-05,
+      "loss": 1.7415,
+      "step": 4092
+    },
+    {
+      "epoch": 0.1757122461590914,
+      "grad_norm": 1.8282791376113892,
+      "learning_rate": 4.901653461101213e-05,
+      "loss": 1.776,
+      "step": 4123
+    },
+    {
+      "epoch": 0.1770333908670545,
+      "grad_norm": 1.7473266124725342,
+      "learning_rate": 4.8993075348363626e-05,
+      "loss": 1.7625,
+      "step": 4154
+    },
+    {
+      "epoch": 0.17835453557501757,
+      "grad_norm": 1.849486231803894,
+      "learning_rate": 4.896934532712084e-05,
+      "loss": 1.7645,
+      "step": 4185
+    },
+    {
+      "epoch": 0.17967568028298067,
+      "grad_norm": 1.772829294204712,
+      "learning_rate": 4.8945344815073846e-05,
+      "loss": 1.7563,
+      "step": 4216
+    },
+    {
+      "epoch": 0.18099682499094377,
+      "grad_norm": 1.8278394937515259,
+      "learning_rate": 4.892107408306516e-05,
+      "loss": 1.7591,
+      "step": 4247
+    },
+    {
+      "epoch": 0.18231796969890685,
+      "grad_norm": 1.7458796501159668,
+      "learning_rate": 4.889653340498669e-05,
+      "loss": 1.7444,
+      "step": 4278
+    },
+    {
+      "epoch": 0.18363911440686995,
+      "grad_norm": 1.8150420188903809,
+      "learning_rate": 4.8871723057776664e-05,
+      "loss": 1.7486,
+      "step": 4309
+    },
+    {
+      "epoch": 0.18496025911483305,
+      "grad_norm": 1.914010763168335,
+      "learning_rate": 4.8846643321416476e-05,
+      "loss": 1.7456,
+      "step": 4340
+    },
+    {
+      "epoch": 0.18628140382279615,
+      "grad_norm": 1.7134442329406738,
+      "learning_rate": 4.882129447892753e-05,
+      "loss": 1.7521,
+      "step": 4371
+    },
+    {
+      "epoch": 0.18760254853075922,
+      "grad_norm": 1.6600676774978638,
+      "learning_rate": 4.8795676816368076e-05,
+      "loss": 1.7416,
+      "step": 4402
+    },
+    {
+      "epoch": 0.18892369323872232,
+      "grad_norm": 1.8451883792877197,
+      "learning_rate": 4.876979062282995e-05,
+      "loss": 1.7595,
+      "step": 4433
+    },
+    {
+      "epoch": 0.19024483794668542,
+      "grad_norm": 1.8544697761535645,
+      "learning_rate": 4.8743636190435325e-05,
+      "loss": 1.7669,
+      "step": 4464
+    },
+    {
+      "epoch": 0.1915659826546485,
+      "grad_norm": 1.704041600227356,
+      "learning_rate": 4.871721381433344e-05,
+      "loss": 1.7403,
+      "step": 4495
+    },
+    {
+      "epoch": 0.1928871273626116,
+      "grad_norm": 1.6593477725982666,
+      "learning_rate": 4.869052379269719e-05,
+      "loss": 1.7359,
+      "step": 4526
+    },
+    {
+      "epoch": 0.1942082720705747,
+      "grad_norm": 1.7819074392318726,
+      "learning_rate": 4.866356642671985e-05,
+      "loss": 1.7485,
+      "step": 4557
+    },
+    {
+      "epoch": 0.1955294167785378,
+      "grad_norm": 1.840776801109314,
+      "learning_rate": 4.8636342020611634e-05,
+      "loss": 1.7566,
+      "step": 4588
+    },
+    {
+      "epoch": 0.19685056148650087,
+      "grad_norm": 1.9981670379638672,
+      "learning_rate": 4.860885088159626e-05,
+      "loss": 1.747,
+      "step": 4619
+    },
+    {
+      "epoch": 0.19817170619446398,
+      "grad_norm": 1.8164016008377075,
+      "learning_rate": 4.858109331990751e-05,
+      "loss": 1.7575,
+      "step": 4650
+    },
+    {
+      "epoch": 0.19949285090242708,
+      "grad_norm": 1.7589538097381592,
+      "learning_rate": 4.855306964878567e-05,
+      "loss": 1.7387,
+      "step": 4681
+    },
+    {
+      "epoch": 0.20081399561039015,
+      "grad_norm": 1.8388264179229736,
+      "learning_rate": 4.8524780184474084e-05,
+      "loss": 1.7273,
+      "step": 4712
+    },
+    {
+      "epoch": 0.20213514031835325,
+      "grad_norm": 1.7422165870666504,
+      "learning_rate": 4.8496225246215496e-05,
+      "loss": 1.7301,
+      "step": 4743
+    },
+    {
+      "epoch": 0.20345628502631635,
+      "grad_norm": 1.6752463579177856,
+      "learning_rate": 4.8467405156248505e-05,
+      "loss": 1.7422,
+      "step": 4774
+    },
+    {
+      "epoch": 0.20477742973427945,
+      "grad_norm": 6.59528923034668,
+      "learning_rate": 4.843832023980392e-05,
+      "loss": 1.748,
+      "step": 4805
+    },
+    {
+      "epoch": 0.20609857444224253,
+      "grad_norm": 1.716232419013977,
+      "learning_rate": 4.840897082510106e-05,
+      "loss": 1.745,
+      "step": 4836
+    },
+    {
+      "epoch": 0.20741971915020563,
+      "grad_norm": 1.7691062688827515,
+      "learning_rate": 4.8379357243344084e-05,
+      "loss": 1.7208,
+      "step": 4867
+    },
+    {
+      "epoch": 0.20874086385816873,
+      "grad_norm": 1.6334593296051025,
+      "learning_rate": 4.8349479828718236e-05,
+      "loss": 1.7288,
+      "step": 4898
+    },
+    {
+      "epoch": 0.2100620085661318,
+      "grad_norm": 1.6801882982254028,
+      "learning_rate": 4.8319338918386075e-05,
+      "loss": 1.7428,
+      "step": 4929
+    },
+    {
+      "epoch": 0.2113831532740949,
+      "grad_norm": 1.6692436933517456,
+      "learning_rate": 4.828893485248369e-05,
+      "loss": 1.7286,
+      "step": 4960
+    },
+    {
+      "epoch": 0.212704297982058,
+      "grad_norm": 1.731826663017273,
+      "learning_rate": 4.825826797411682e-05,
+      "loss": 1.7455,
+      "step": 4991
+    },
+    {
+      "epoch": 0.2140254426900211,
+      "grad_norm": 1.689012050628662,
+      "learning_rate": 4.822733862935702e-05,
+      "loss": 1.7398,
+      "step": 5022
+    },
+    {
+      "epoch": 0.21534658739798418,
+      "grad_norm": 1.6423673629760742,
+      "learning_rate": 4.819614716723775e-05,
+      "loss": 1.7305,
+      "step": 5053
+    },
+    {
+      "epoch": 0.21666773210594728,
+      "grad_norm": 1.7029454708099365,
+      "learning_rate": 4.8164693939750425e-05,
+      "loss": 1.7355,
+      "step": 5084
+    },
+    {
+      "epoch": 0.21798887681391038,
+      "grad_norm": 1.7331738471984863,
+      "learning_rate": 4.813297930184042e-05,
+      "loss": 1.7167,
+      "step": 5115
+    },
+    {
+      "epoch": 0.21931002152187345,
+      "grad_norm": 1.729872226715088,
+      "learning_rate": 4.810100361140314e-05,
+      "loss": 1.7402,
+      "step": 5146
+    },
+    {
+      "epoch": 0.22063116622983656,
+      "grad_norm": 1.675826907157898,
+      "learning_rate": 4.8068767229279885e-05,
+      "loss": 1.7129,
+      "step": 5177
+    },
+    {
+      "epoch": 0.22195231093779966,
+      "grad_norm": 1.7073479890823364,
+      "learning_rate": 4.8036270519253854e-05,
+      "loss": 1.7292,
+      "step": 5208
+    },
+    {
+      "epoch": 0.22327345564576276,
+      "grad_norm": 1.67916738986969,
+      "learning_rate": 4.8003513848046e-05,
+      "loss": 1.7417,
+      "step": 5239
+    },
+    {
+      "epoch": 0.22459460035372583,
+      "grad_norm": 1.751838207244873,
+      "learning_rate": 4.79704975853109e-05,
+      "loss": 1.7167,
+      "step": 5270
+    },
+    {
+      "epoch": 0.22591574506168893,
+      "grad_norm": 1.7748228311538696,
+      "learning_rate": 4.793722210363262e-05,
+      "loss": 1.7175,
+      "step": 5301
+    },
+    {
+      "epoch": 0.22723688976965203,
+      "grad_norm": 1.6792418956756592,
+      "learning_rate": 4.7903687778520414e-05,
+      "loss": 1.7537,
+      "step": 5332
+    },
+    {
+      "epoch": 0.2285580344776151,
+      "grad_norm": 1.77089524269104,
+      "learning_rate": 4.7869894988404593e-05,
+      "loss": 1.7181,
+      "step": 5363
+    },
+    {
+      "epoch": 0.2298791791855782,
+      "grad_norm": 1.795487642288208,
+      "learning_rate": 4.783584411463221e-05,
+      "loss": 1.7216,
+      "step": 5394
+    },
+    {
+      "epoch": 0.2312003238935413,
+      "grad_norm": 1.8036439418792725,
+      "learning_rate": 4.780153554146274e-05,
+      "loss": 1.7172,
+      "step": 5425
+    },
+    {
+      "epoch": 0.2325214686015044,
+      "grad_norm": 1.683451771736145,
+      "learning_rate": 4.7766969656063766e-05,
+      "loss": 1.7371,
+      "step": 5456
+    },
+    {
+      "epoch": 0.23384261330946748,
+      "grad_norm": 1.7509490251541138,
+      "learning_rate": 4.773214684850662e-05,
+      "loss": 1.7245,
+      "step": 5487
+    },
+    {
+      "epoch": 0.23516375801743059,
+      "grad_norm": 1.6572686433792114,
+      "learning_rate": 4.769706751176193e-05,
+      "loss": 1.7108,
+      "step": 5518
+    },
+    {
+      "epoch": 0.2364849027253937,
+      "grad_norm": 1.7440228462219238,
+      "learning_rate": 4.7661732041695264e-05,
+      "loss": 1.7115,
+      "step": 5549
+    },
+    {
+      "epoch": 0.23780604743335676,
+      "grad_norm": 1.613687515258789,
+      "learning_rate": 4.762614083706258e-05,
+      "loss": 1.7079,
+      "step": 5580
+    },
+    {
+      "epoch": 0.23912719214131986,
+      "grad_norm": 1.4738078117370605,
+      "learning_rate": 4.759029429950581e-05,
+      "loss": 1.7097,
+      "step": 5611
+    },
+    {
+      "epoch": 0.24044833684928296,
+      "grad_norm": 1.6921734809875488,
+      "learning_rate": 4.7554192833548235e-05,
+      "loss": 1.7187,
+      "step": 5642
+    },
+    {
+      "epoch": 0.24176948155724606,
+      "grad_norm": 1.665814757347107,
+      "learning_rate": 4.751783684659e-05,
+      "loss": 1.7302,
+      "step": 5673
+    },
+    {
+      "epoch": 0.24309062626520914,
+      "grad_norm": 1.5973016023635864,
+      "learning_rate": 4.748122674890348e-05,
+      "loss": 1.718,
+      "step": 5704
+    },
+    {
+      "epoch": 0.24441177097317224,
+      "grad_norm": 1.6037402153015137,
+      "learning_rate": 4.7444362953628654e-05,
+      "loss": 1.7213,
+      "step": 5735
+    },
+    {
+      "epoch": 0.24573291568113534,
+      "grad_norm": 1.6214978694915771,
+      "learning_rate": 4.7407245876768424e-05,
+      "loss": 1.7343,
+      "step": 5766
+    },
+    {
+      "epoch": 0.2470540603890984,
+      "grad_norm": 1.758327603340149,
+      "learning_rate": 4.736987593718397e-05,
+      "loss": 1.6993,
+      "step": 5797
+    },
+    {
+      "epoch": 0.2483752050970615,
+      "grad_norm": 1.7455657720565796,
+      "learning_rate": 4.733225355658999e-05,
+      "loss": 1.7147,
+      "step": 5828
+    },
+    {
+      "epoch": 0.24969634980502461,
+      "grad_norm": 1.6355786323547363,
+      "learning_rate": 4.7294379159549926e-05,
+      "loss": 1.7284,
+      "step": 5859
+    },
+    {
+      "epoch": 0.2510174945129877,
+      "grad_norm": 1.6421799659729004,
+      "learning_rate": 4.725625317347119e-05,
+      "loss": 1.6986,
+      "step": 5890
+    },
+    {
+      "epoch": 0.2523386392209508,
+      "grad_norm": 1.6713591814041138,
+      "learning_rate": 4.7217876028600374e-05,
+      "loss": 1.7083,
+      "step": 5921
+    },
+    {
+      "epoch": 0.2536597839289139,
+      "grad_norm": 1.7056536674499512,
+      "learning_rate": 4.717924815801832e-05,
+      "loss": 1.7339,
+      "step": 5952
+    },
+    {
+      "epoch": 0.254980928636877,
+      "grad_norm": 1.6911596059799194,
+      "learning_rate": 4.714036999763532e-05,
+      "loss": 1.7068,
+      "step": 5983
+    },
+    {
+      "epoch": 0.25630207334484006,
+      "grad_norm": 1.68338143825531,
+      "learning_rate": 4.7101241986186116e-05,
+      "loss": 1.7153,
+      "step": 6014
+    },
+    {
+      "epoch": 0.2576232180528032,
+      "grad_norm": 1.6629693508148193,
+      "learning_rate": 4.7061864565225e-05,
+      "loss": 1.7178,
+      "step": 6045
+    },
+    {
+      "epoch": 0.25894436276076627,
+      "grad_norm": 1.8833751678466797,
+      "learning_rate": 4.702223817912081e-05,
+      "loss": 1.6759,
+      "step": 6076
+    },
+    {
+      "epoch": 0.26026550746872934,
+      "grad_norm": 1.8015220165252686,
+      "learning_rate": 4.698236327505195e-05,
+      "loss": 1.6893,
+      "step": 6107
+    },
+    {
+      "epoch": 0.26158665217669247,
+      "grad_norm": 1.761045217514038,
+      "learning_rate": 4.694224030300127e-05,
+      "loss": 1.7015,
+      "step": 6138
+    },
+    {
+      "epoch": 0.26290779688465554,
+      "grad_norm": 1.6620142459869385,
+      "learning_rate": 4.690186971575107e-05,
+      "loss": 1.6897,
+      "step": 6169
+    },
+    {
+      "epoch": 0.2642289415926186,
+      "grad_norm": 1.6513328552246094,
+      "learning_rate": 4.6861251968877916e-05,
+      "loss": 1.6926,
+      "step": 6200
+    },
+    {
+      "epoch": 0.26555008630058174,
+      "grad_norm": 1.6536865234375,
+      "learning_rate": 4.68203875207476e-05,
+      "loss": 1.7193,
+      "step": 6231
+    },
+    {
+      "epoch": 0.2668712310085448,
+      "grad_norm": 1.5718939304351807,
+      "learning_rate": 4.677927683250983e-05,
+      "loss": 1.6821,
+      "step": 6262
+    },
+    {
+      "epoch": 0.2681923757165079,
+      "grad_norm": 1.6237717866897583,
+      "learning_rate": 4.6737920368093156e-05,
+      "loss": 1.7116,
+      "step": 6293
+    },
+    {
+      "epoch": 0.269513520424471,
+      "grad_norm": 1.6111472845077515,
+      "learning_rate": 4.669631859419965e-05,
+      "loss": 1.7101,
+      "step": 6324
+    },
+    {
+      "epoch": 0.2708346651324341,
+      "grad_norm": 1.711414098739624,
+      "learning_rate": 4.6654471980299676e-05,
+      "loss": 1.7053,
+      "step": 6355
+    },
+    {
+      "epoch": 0.2721558098403972,
+      "grad_norm": 1.5804500579833984,
+      "learning_rate": 4.661238099862658e-05,
+      "loss": 1.6867,
+      "step": 6386
+    },
+    {
+      "epoch": 0.2734769545483603,
+      "grad_norm": 1.5932157039642334,
+      "learning_rate": 4.657004612417138e-05,
+      "loss": 1.7181,
+      "step": 6417
+    },
+    {
+      "epoch": 0.27479809925632337,
+      "grad_norm": 1.5526095628738403,
+      "learning_rate": 4.6527467834677374e-05,
+      "loss": 1.7178,
+      "step": 6448
+    },
+    {
+      "epoch": 0.2761192439642865,
+      "grad_norm": 1.5979021787643433,
+      "learning_rate": 4.648464661063478e-05,
+      "loss": 1.6885,
+      "step": 6479
+    },
+    {
+      "epoch": 0.27744038867224957,
+      "grad_norm": 1.5154697895050049,
+      "learning_rate": 4.6441582935275264e-05,
+      "loss": 1.6798,
+      "step": 6510
+    },
+    {
+      "epoch": 0.27876153338021264,
+      "grad_norm": 1.6004109382629395,
+      "learning_rate": 4.6398277294566586e-05,
+      "loss": 1.6877,
+      "step": 6541
+    },
+    {
+      "epoch": 0.2800826780881758,
+      "grad_norm": 1.7409225702285767,
+      "learning_rate": 4.6354730177207e-05,
+      "loss": 1.6877,
+      "step": 6572
+    },
+    {
+      "epoch": 0.28140382279613885,
+      "grad_norm": 1.5840095281600952,
+      "learning_rate": 4.6310942074619787e-05,
+      "loss": 1.6885,
+      "step": 6603
+    },
+    {
+      "epoch": 0.2827249675041019,
+      "grad_norm": 1.7218347787857056,
+      "learning_rate": 4.626691348094777e-05,
+      "loss": 1.6993,
+      "step": 6634
+    },
+    {
+      "epoch": 0.28404611221206505,
+      "grad_norm": 1.6517518758773804,
+      "learning_rate": 4.622264489304762e-05,
+      "loss": 1.7086,
+      "step": 6665
+    },
+    {
+      "epoch": 0.2853672569200281,
+      "grad_norm": 1.6828736066818237,
+      "learning_rate": 4.617813681048434e-05,
+      "loss": 1.6895,
+      "step": 6696
+    },
+    {
+      "epoch": 0.2866884016279912,
+      "grad_norm": 1.6049277782440186,
+      "learning_rate": 4.61333897355256e-05,
+      "loss": 1.6929,
+      "step": 6727
+    },
+    {
+      "epoch": 0.2880095463359543,
+      "grad_norm": 1.6272588968276978,
+      "learning_rate": 4.608840417313604e-05,
+      "loss": 1.693,
+      "step": 6758
+    },
+    {
+      "epoch": 0.2893306910439174,
+      "grad_norm": 1.6424005031585693,
+      "learning_rate": 4.6043180630971646e-05,
+      "loss": 1.6882,
+      "step": 6789
+    },
+    {
+      "epoch": 0.2906518357518805,
+      "grad_norm": 1.5532363653182983,
+      "learning_rate": 4.599771961937391e-05,
+      "loss": 1.6846,
+      "step": 6820
+    },
+    {
+      "epoch": 0.2919729804598436,
+      "grad_norm": 1.566545009613037,
+      "learning_rate": 4.5952021651364204e-05,
+      "loss": 1.6694,
+      "step": 6851
+    },
+    {
+      "epoch": 0.2932941251678067,
+      "grad_norm": 1.592480182647705,
+      "learning_rate": 4.590608724263786e-05,
+      "loss": 1.6689,
+      "step": 6882
+    },
+    {
+      "epoch": 0.2946152698757698,
+      "grad_norm": 1.6120296716690063,
+      "learning_rate": 4.585991691155845e-05,
+      "loss": 1.6654,
+      "step": 6913
+    },
+    {
+      "epoch": 0.2959364145837329,
+      "grad_norm": 1.5690969228744507,
+      "learning_rate": 4.581351117915188e-05,
+      "loss": 1.6586,
+      "step": 6944
+    },
+    {
+      "epoch": 0.29725755929169595,
+      "grad_norm": 1.5756118297576904,
+      "learning_rate": 4.5766870569100534e-05,
+      "loss": 1.6859,
+      "step": 6975
+    },
+    {
+      "epoch": 0.2985787039996591,
+      "grad_norm": 1.5164366960525513,
+      "learning_rate": 4.571999560773736e-05,
+      "loss": 1.6844,
+      "step": 7006
+    },
+    {
+      "epoch": 0.29989984870762215,
+      "grad_norm": 1.488936424255371,
+      "learning_rate": 4.5672886824039915e-05,
+      "loss": 1.6897,
+      "step": 7037
+    },
+    {
+      "epoch": 0.3012209934155852,
+      "grad_norm": 1.6014569997787476,
+      "learning_rate": 4.5625544749624435e-05,
+      "loss": 1.6821,
+      "step": 7068
+    },
+    {
+      "epoch": 0.30254213812354835,
+      "grad_norm": 1.5514070987701416,
+      "learning_rate": 4.5577969918739794e-05,
+      "loss": 1.7017,
+      "step": 7099
+    },
+    {
+      "epoch": 0.3038632828315114,
+      "grad_norm": 1.5641967058181763,
+      "learning_rate": 4.5530162868261486e-05,
+      "loss": 1.6809,
+      "step": 7130
+    },
+    {
+      "epoch": 0.3051844275394745,
+      "grad_norm": 1.5185332298278809,
+      "learning_rate": 4.548212413768558e-05,
+      "loss": 1.6719,
+      "step": 7161
+    },
+    {
+      "epoch": 0.30650557224743763,
+      "grad_norm": 1.497670292854309,
+      "learning_rate": 4.543385426912261e-05,
+      "loss": 1.6873,
+      "step": 7192
+    },
+    {
+      "epoch": 0.3078267169554007,
+      "grad_norm": 1.6817424297332764,
+      "learning_rate": 4.53853538072915e-05,
+      "loss": 1.668,
+      "step": 7223
+    },
+    {
+      "epoch": 0.30914786166336383,
+      "grad_norm": 1.561675786972046,
+      "learning_rate": 4.533662329951336e-05,
+      "loss": 1.678,
+      "step": 7254
+    },
+    {
+      "epoch": 0.3104690063713269,
+      "grad_norm": 1.7712374925613403,
+      "learning_rate": 4.528766329570536e-05,
+      "loss": 1.6817,
+      "step": 7285
+    },
+    {
+      "epoch": 0.31179015107929,
+      "grad_norm": 1.530474066734314,
+      "learning_rate": 4.523847434837447e-05,
+      "loss": 1.671,
+      "step": 7316
+    },
+    {
+      "epoch": 0.3131112957872531,
+      "grad_norm": 1.8106886148452759,
+      "learning_rate": 4.518905701261128e-05,
+      "loss": 1.6532,
+      "step": 7347
+    },
+    {
+      "epoch": 0.3144324404952162,
+      "grad_norm": 1.5035251379013062,
+      "learning_rate": 4.5139411846083715e-05,
+      "loss": 1.6651,
+      "step": 7378
+    },
+    {
+      "epoch": 0.31575358520317925,
+      "grad_norm": 1.6194860935211182,
+      "learning_rate": 4.508953940903073e-05,
+      "loss": 1.6664,
+      "step": 7409
+    },
+    {
+      "epoch": 0.3170747299111424,
+      "grad_norm": 1.5217773914337158,
+      "learning_rate": 4.5039440264255994e-05,
+      "loss": 1.6697,
+      "step": 7440
+    },
+    {
+      "epoch": 0.31839587461910546,
+      "grad_norm": 1.670386552810669,
+      "learning_rate": 4.498911497712155e-05,
+      "loss": 1.6637,
+      "step": 7471
+    },
+    {
+      "epoch": 0.31971701932706853,
+      "grad_norm": 1.680743932723999,
+      "learning_rate": 4.493856411554142e-05,
+      "loss": 1.6822,
+      "step": 7502
+    },
+    {
+      "epoch": 0.32103816403503166,
+      "grad_norm": 1.628322958946228,
+      "learning_rate": 4.4887788249975206e-05,
+      "loss": 1.6615,
+      "step": 7533
+    },
+    {
+      "epoch": 0.32235930874299473,
+      "grad_norm": 1.5849953889846802,
+      "learning_rate": 4.4836787953421656e-05,
+      "loss": 1.6637,
+      "step": 7564
+    },
+    {
+      "epoch": 0.3236804534509578,
+      "grad_norm": 1.5066392421722412,
+      "learning_rate": 4.478556380141218e-05,
+      "loss": 1.6839,
+      "step": 7595
+    },
+    {
+      "epoch": 0.32500159815892093,
+      "grad_norm": 1.6168526411056519,
+      "learning_rate": 4.4734116372004375e-05,
+      "loss": 1.6543,
+      "step": 7626
+    },
+    {
+      "epoch": 0.326322742866884,
+      "grad_norm": 1.5073281526565552,
+      "learning_rate": 4.4682446245775477e-05,
+      "loss": 1.6752,
+      "step": 7657
+    },
+    {
+      "epoch": 0.32764388757484714,
+      "grad_norm": 1.5694515705108643,
+      "learning_rate": 4.463055400581586e-05,
+      "loss": 1.6646,
+      "step": 7688
+    },
+    {
+      "epoch": 0.3289650322828102,
+      "grad_norm": 1.5031605958938599,
+      "learning_rate": 4.4578440237722374e-05,
+      "loss": 1.6807,
+      "step": 7719
+    },
+    {
+      "epoch": 0.3302861769907733,
+      "grad_norm": 1.4886877536773682,
+      "learning_rate": 4.452610552959183e-05,
+      "loss": 1.6644,
+      "step": 7750
+    },
+    {
+      "epoch": 0.3316073216987364,
+      "grad_norm": 1.5978937149047852,
+      "learning_rate": 4.447355047201428e-05,
+      "loss": 1.6573,
+      "step": 7781
+    },
+    {
+      "epoch": 0.3329284664066995,
+      "grad_norm": 1.4938329458236694,
+      "learning_rate": 4.4420775658066414e-05,
+      "loss": 1.6489,
+      "step": 7812
+    },
+    {
+      "epoch": 0.33424961111466256,
+      "grad_norm": 1.541128396987915,
+      "learning_rate": 4.436778168330484e-05,
+      "loss": 1.6673,
+      "step": 7843
+    },
+    {
+      "epoch": 0.3355707558226257,
+      "grad_norm": 1.5185847282409668,
+      "learning_rate": 4.4314569145759353e-05,
+      "loss": 1.6631,
+      "step": 7874
+    },
+    {
+      "epoch": 0.33689190053058876,
+      "grad_norm": 1.5545519590377808,
+      "learning_rate": 4.42611386459262e-05,
+      "loss": 1.6535,
+      "step": 7905
+    },
+    {
+      "epoch": 0.33821304523855183,
+      "grad_norm": 1.5892200469970703,
+      "learning_rate": 4.420749078676133e-05,
+      "loss": 1.6574,
+      "step": 7936
+    },
+    {
+      "epoch": 0.33953418994651496,
+      "grad_norm": 1.5779627561569214,
+      "learning_rate": 4.4153626173673516e-05,
+      "loss": 1.6683,
+      "step": 7967
+    },
+    {
+      "epoch": 0.34085533465447804,
+      "grad_norm": 1.5521479845046997,
+      "learning_rate": 4.409954541451762e-05,
+      "loss": 1.6599,
+      "step": 7998
+    },
+    {
+      "epoch": 0.3421764793624411,
+      "grad_norm": 1.4845163822174072,
+      "learning_rate": 4.404524911958764e-05,
+      "loss": 1.6357,
+      "step": 8029
+    },
+    {
+      "epoch": 0.34349762407040424,
+      "grad_norm": 1.4665591716766357,
+      "learning_rate": 4.399073790160989e-05,
+      "loss": 1.6577,
+      "step": 8060
+    },
+    {
+      "epoch": 0.3448187687783673,
+      "grad_norm": 1.4990118741989136,
+      "learning_rate": 4.393601237573607e-05,
+      "loss": 1.6589,
+      "step": 8091
+    },
+    {
+      "epoch": 0.34613991348633044,
+      "grad_norm": 1.6129504442214966,
+      "learning_rate": 4.388107315953628e-05,
+      "loss": 1.6719,
+      "step": 8122
+    },
+    {
+      "epoch": 0.3474610581942935,
+      "grad_norm": 1.5837408304214478,
+      "learning_rate": 4.382592087299212e-05,
+      "loss": 1.6389,
+      "step": 8153
+    },
+    {
+      "epoch": 0.3487822029022566,
+      "grad_norm": 1.4611998796463013,
+      "learning_rate": 4.377055613848964e-05,
+      "loss": 1.6568,
+      "step": 8184
+    },
+    {
+      "epoch": 0.3501033476102197,
+      "grad_norm": 1.5456607341766357,
+      "learning_rate": 4.3714979580812355e-05,
+      "loss": 1.6442,
+      "step": 8215
+    },
+    {
+      "epoch": 0.3514244923181828,
+      "grad_norm": 1.5393562316894531,
+      "learning_rate": 4.365919182713416e-05,
+      "loss": 1.6505,
+      "step": 8246
+    },
+    {
+      "epoch": 0.35274563702614586,
+      "grad_norm": 1.5040944814682007,
+      "learning_rate": 4.360319350701226e-05,
+      "loss": 1.6701,
+      "step": 8277
+    },
+    {
+      "epoch": 0.354066781734109,
+      "grad_norm": 1.5542012453079224,
+      "learning_rate": 4.3546985252380115e-05,
+      "loss": 1.6431,
+      "step": 8308
+    },
+    {
+      "epoch": 0.35538792644207207,
+      "grad_norm": 1.4878984689712524,
+      "learning_rate": 4.349056769754021e-05,
+      "loss": 1.6543,
+      "step": 8339
+    },
+    {
+      "epoch": 0.35670907115003514,
+      "grad_norm": 1.6049541234970093,
+      "learning_rate": 4.3433941479156994e-05,
+      "loss": 1.6415,
+      "step": 8370
+    },
+    {
+      "epoch": 0.35803021585799827,
+      "grad_norm": 1.6094954013824463,
+      "learning_rate": 4.3377107236249647e-05,
+      "loss": 1.646,
+      "step": 8401
+    },
+    {
+      "epoch": 0.35935136056596134,
+      "grad_norm": 1.5165326595306396,
+      "learning_rate": 4.332006561018488e-05,
+      "loss": 1.6575,
+      "step": 8432
+    },
+    {
+      "epoch": 0.3606725052739244,
+      "grad_norm": 1.5776197910308838,
+      "learning_rate": 4.3262817244669683e-05,
+      "loss": 1.6556,
+      "step": 8463
+    },
+    {
+      "epoch": 0.36199364998188754,
+      "grad_norm": 1.4522629976272583,
+      "learning_rate": 4.3205362785744083e-05,
+      "loss": 1.6583,
+      "step": 8494
+    },
+    {
+      "epoch": 0.3633147946898506,
+      "grad_norm": 1.4910801649093628,
+      "learning_rate": 4.314770288177384e-05,
+      "loss": 1.638,
+      "step": 8525
+    },
+    {
+      "epoch": 0.3646359393978137,
+      "grad_norm": 1.6230814456939697,
+      "learning_rate": 4.308983818344313e-05,
+      "loss": 1.6644,
+      "step": 8556
+    },
+    {
+      "epoch": 0.3659570841057768,
+      "grad_norm": 1.6548675298690796,
+      "learning_rate": 4.3031769343747206e-05,
+      "loss": 1.639,
+      "step": 8587
+    },
+    {
+      "epoch": 0.3672782288137399,
+      "grad_norm": 1.5451959371566772,
+      "learning_rate": 4.297349701798505e-05,
+      "loss": 1.6452,
+      "step": 8618
+    },
+    {
+      "epoch": 0.368599373521703,
+      "grad_norm": 1.46851646900177,
+      "learning_rate": 4.2915021863751916e-05,
+      "loss": 1.6645,
+      "step": 8649
+    },
+    {
+      "epoch": 0.3699205182296661,
+      "grad_norm": 1.4022047519683838,
+      "learning_rate": 4.285634454093198e-05,
+      "loss": 1.6556,
+      "step": 8680
+    },
+    {
+      "epoch": 0.37124166293762917,
+      "grad_norm": 1.4209098815917969,
+      "learning_rate": 4.279746571169086e-05,
+      "loss": 1.6699,
+      "step": 8711
+    },
+    {
+      "epoch": 0.3725628076455923,
+      "grad_norm": 1.4034055471420288,
+      "learning_rate": 4.2738386040468136e-05,
+      "loss": 1.6491,
+      "step": 8742
+    },
+    {
+      "epoch": 0.37388395235355537,
+      "grad_norm": 1.4482676982879639,
+      "learning_rate": 4.2679106193969866e-05,
+      "loss": 1.6299,
+      "step": 8773
+    },
+    {
+      "epoch": 0.37520509706151844,
+      "grad_norm": 1.542325496673584,
+      "learning_rate": 4.261962684116106e-05,
+      "loss": 1.6666,
+      "step": 8804
+    },
+    {
+      "epoch": 0.3765262417694816,
+      "grad_norm": 1.4821662902832031,
+      "learning_rate": 4.2559948653258145e-05,
+      "loss": 1.6383,
+      "step": 8835
+    },
+    {
+      "epoch": 0.37784738647744465,
+      "grad_norm": 1.445188283920288,
+      "learning_rate": 4.250007230372134e-05,
+      "loss": 1.6287,
+      "step": 8866
+    },
+    {
+      "epoch": 0.3791685311854077,
+      "grad_norm": 1.5624685287475586,
+      "learning_rate": 4.2439998468247126e-05,
+      "loss": 1.6337,
+      "step": 8897
+    },
+    {
+      "epoch": 0.38048967589337085,
+      "grad_norm": 1.7001335620880127,
+      "learning_rate": 4.2379727824760566e-05,
+      "loss": 1.6352,
+      "step": 8928
+    },
+    {
+      "epoch": 0.3818108206013339,
+      "grad_norm": 1.5825940370559692,
+      "learning_rate": 4.231926105340768e-05,
+      "loss": 1.6336,
+      "step": 8959
+    },
+    {
+      "epoch": 0.383131965309297,
+      "grad_norm": 1.4825541973114014,
+      "learning_rate": 4.225859883654776e-05,
+      "loss": 1.6359,
+      "step": 8990
+    },
+    {
+      "epoch": 0.3844531100172601,
+      "grad_norm": 1.4914085865020752,
+      "learning_rate": 4.219774185874569e-05,
+      "loss": 1.6623,
+      "step": 9021
+    },
+    {
+      "epoch": 0.3857742547252232,
+      "grad_norm": 1.491519570350647,
+      "learning_rate": 4.213669080676418e-05,
+      "loss": 1.6543,
+      "step": 9052
+    },
+    {
+      "epoch": 0.3870953994331863,
+      "grad_norm": 1.5370805263519287,
+      "learning_rate": 4.2075446369556056e-05,
+      "loss": 1.6325,
+      "step": 9083
+    },
+    {
+      "epoch": 0.3884165441411494,
+      "grad_norm": 1.423508644104004,
+      "learning_rate": 4.201400923825648e-05,
+      "loss": 1.6547,
+      "step": 9114
+    },
+    {
+      "epoch": 0.3897376888491125,
+      "grad_norm": 1.4396551847457886,
+      "learning_rate": 4.195238010617511e-05,
+      "loss": 1.6308,
+      "step": 9145
+    },
+    {
+      "epoch": 0.3910588335570756,
+      "grad_norm": 1.466518521308899,
+      "learning_rate": 4.1890559668788344e-05,
+      "loss": 1.6321,
+      "step": 9176
+    },
+    {
+      "epoch": 0.3923799782650387,
+      "grad_norm": 1.5302077531814575,
+      "learning_rate": 4.1828548623731405e-05,
+      "loss": 1.6455,
+      "step": 9207
+    },
+    {
+      "epoch": 0.39370112297300175,
+      "grad_norm": 1.5934138298034668,
+      "learning_rate": 4.1766347670790506e-05,
+      "loss": 1.6526,
+      "step": 9238
+    },
+    {
+      "epoch": 0.3950222676809649,
+      "grad_norm": 1.5880807638168335,
+      "learning_rate": 4.170395751189495e-05,
+      "loss": 1.6476,
+      "step": 9269
+    },
+    {
+      "epoch": 0.39634341238892795,
+      "grad_norm": 1.4420969486236572,
+      "learning_rate": 4.164137885110921e-05,
+      "loss": 1.6265,
+      "step": 9300
+    },
+    {
+      "epoch": 0.397664557096891,
+      "grad_norm": 1.489991307258606,
+      "learning_rate": 4.157861239462495e-05,
+      "loss": 1.647,
+      "step": 9331
+    },
+    {
+      "epoch": 0.39898570180485415,
+      "grad_norm": 1.5055150985717773,
+      "learning_rate": 4.1515658850753114e-05,
+      "loss": 1.6323,
+      "step": 9362
+    },
+    {
+      "epoch": 0.4003068465128172,
+      "grad_norm": 1.4316349029541016,
+      "learning_rate": 4.145251892991588e-05,
+      "loss": 1.6329,
+      "step": 9393
+    },
+    {
+      "epoch": 0.4016279912207803,
+      "grad_norm": 1.4182207584381104,
+      "learning_rate": 4.138919334463868e-05,
+      "loss": 1.6479,
+      "step": 9424
+    },
+    {
+      "epoch": 0.40294913592874343,
+      "grad_norm": 1.5128748416900635,
+      "learning_rate": 4.1325682809542124e-05,
+      "loss": 1.6419,
+      "step": 9455
+    },
+    {
+      "epoch": 0.4042702806367065,
+      "grad_norm": 1.4093265533447266,
+      "learning_rate": 4.126198804133398e-05,
+      "loss": 1.6128,
+      "step": 9486
+    },
+    {
+      "epoch": 0.40559142534466963,
+      "grad_norm": 1.4173692464828491,
+      "learning_rate": 4.1198109758801055e-05,
+      "loss": 1.6437,
+      "step": 9517
+    },
+    {
+      "epoch": 0.4069125700526327,
+      "grad_norm": 1.6048704385757446,
+      "learning_rate": 4.113404868280107e-05,
+      "loss": 1.6203,
+      "step": 9548
+    },
+    {
+      "epoch": 0.4082337147605958,
+      "grad_norm": 1.495764970779419,
+      "learning_rate": 4.106980553625457e-05,
+      "loss": 1.6294,
+      "step": 9579
+    },
+    {
+      "epoch": 0.4095548594685589,
+      "grad_norm": 1.530218482017517,
+      "learning_rate": 4.100538104413674e-05,
+      "loss": 1.6377,
+      "step": 9610
+    },
+    {
+      "epoch": 0.410876004176522,
+      "grad_norm": 1.5626767873764038,
+      "learning_rate": 4.09407759334692e-05,
+      "loss": 1.6293,
+      "step": 9641
+    },
+    {
+      "epoch": 0.41219714888448505,
+      "grad_norm": 1.3259074687957764,
+      "learning_rate": 4.087599093331186e-05,
+      "loss": 1.6209,
+      "step": 9672
+    },
+    {
+      "epoch": 0.4135182935924482,
+      "grad_norm": 1.4668447971343994,
+      "learning_rate": 4.081102677475462e-05,
+      "loss": 1.607,
+      "step": 9703
+    },
+    {
+      "epoch": 0.41483943830041126,
+      "grad_norm": 1.6264866590499878,
+      "learning_rate": 4.0745884190909194e-05,
+      "loss": 1.647,
+      "step": 9734
+    },
+    {
+      "epoch": 0.41616058300837433,
+      "grad_norm": 1.4699751138687134,
+      "learning_rate": 4.0680563916900796e-05,
+      "loss": 1.6433,
+      "step": 9765
+    },
+    {
+      "epoch": 0.41748172771633746,
+      "grad_norm": 1.5570849180221558,
+      "learning_rate": 4.0615066689859815e-05,
+      "loss": 1.6278,
+      "step": 9796
+    },
+    {
+      "epoch": 0.41880287242430053,
+      "grad_norm": 1.5093053579330444,
+      "learning_rate": 4.0549393248913584e-05,
+      "loss": 1.625,
+      "step": 9827
+    },
+    {
+      "epoch": 0.4201240171322636,
+      "grad_norm": 1.4233616590499878,
+      "learning_rate": 4.048354433517794e-05,
+      "loss": 1.6254,
+      "step": 9858
+    },
+    {
+      "epoch": 0.42144516184022673,
+      "grad_norm": 1.41616952419281,
+      "learning_rate": 4.0417520691748916e-05,
+      "loss": 1.6339,
+      "step": 9889
+    },
+    {
+      "epoch": 0.4227663065481898,
+      "grad_norm": 1.5434125661849976,
+      "learning_rate": 4.035132306369438e-05,
+      "loss": 1.6379,
+      "step": 9920
+    },
+    {
+      "epoch": 0.42408745125615294,
+      "grad_norm": 1.4518119096755981,
+      "learning_rate": 4.028495219804555e-05,
+      "loss": 1.6384,
+      "step": 9951
+    },
+    {
+      "epoch": 0.425408595964116,
+      "grad_norm": 1.4989590644836426,
+      "learning_rate": 4.021840884378864e-05,
+      "loss": 1.615,
+      "step": 9982
+    },
+    {
+      "epoch": 0.4267297406720791,
+      "grad_norm": 1.4292545318603516,
+      "learning_rate": 4.015169375185633e-05,
+      "loss": 1.6381,
+      "step": 10013
+    },
+    {
+      "epoch": 0.4280508853800422,
+      "grad_norm": 1.4783132076263428,
+      "learning_rate": 4.0084807675119396e-05,
+      "loss": 1.634,
+      "step": 10044
+    },
+    {
+      "epoch": 0.4293720300880053,
+      "grad_norm": 1.4413528442382812,
+      "learning_rate": 4.0017751368378106e-05,
+      "loss": 1.6356,
+      "step": 10075
+    },
+    {
+      "epoch": 0.43069317479596836,
+      "grad_norm": 1.4603348970413208,
+      "learning_rate": 3.995052558835377e-05,
+      "loss": 1.6248,
+      "step": 10106
+    },
+    {
+      "epoch": 0.4320143195039315,
+      "grad_norm": 1.5041816234588623,
+      "learning_rate": 3.988313109368017e-05,
+      "loss": 1.6352,
+      "step": 10137
+    },
+    {
+      "epoch": 0.43333546421189456,
+      "grad_norm": 1.483124852180481,
+      "learning_rate": 3.981556864489504e-05,
+      "loss": 1.6129,
+      "step": 10168
+    },
+    {
+      "epoch": 0.43465660891985763,
+      "grad_norm": 1.4493173360824585,
+      "learning_rate": 3.974783900443142e-05,
+      "loss": 1.624,
+      "step": 10199
+    },
+    {
+      "epoch": 0.43597775362782076,
+      "grad_norm": 1.553686499595642,
+      "learning_rate": 3.9679942936609095e-05,
+      "loss": 1.6099,
+      "step": 10230
+    },
+    {
+      "epoch": 0.43729889833578384,
+      "grad_norm": 1.4004040956497192,
+      "learning_rate": 3.961188120762596e-05,
+      "loss": 1.6152,
+      "step": 10261
+    },
+    {
+      "epoch": 0.4386200430437469,
+      "grad_norm": 1.4119765758514404,
+      "learning_rate": 3.954365458554938e-05,
+      "loss": 1.6161,
+      "step": 10292
+    },
+    {
+      "epoch": 0.43994118775171004,
+      "grad_norm": 1.4605529308319092,
+      "learning_rate": 3.947526384030751e-05,
+      "loss": 1.6234,
+      "step": 10323
+    },
+    {
+      "epoch": 0.4412623324596731,
+      "grad_norm": 1.5051487684249878,
+      "learning_rate": 3.9406709743680624e-05,
+      "loss": 1.6126,
+      "step": 10354
+    },
+    {
+      "epoch": 0.44258347716763624,
+      "grad_norm": 1.431213617324829,
+      "learning_rate": 3.9337993069292366e-05,
+      "loss": 1.6304,
+      "step": 10385
+    },
+    {
+      "epoch": 0.4439046218755993,
+      "grad_norm": 1.3373912572860718,
+      "learning_rate": 3.926911459260109e-05,
+      "loss": 1.6243,
+      "step": 10416
+    },
+    {
+      "epoch": 0.4452257665835624,
+      "grad_norm": 1.4510406255722046,
+      "learning_rate": 3.920007509089102e-05,
+      "loss": 1.6163,
+      "step": 10447
+    },
+    {
+      "epoch": 0.4465469112915255,
+      "grad_norm": 1.465527057647705,
+      "learning_rate": 3.913087534326357e-05,
+      "loss": 1.6092,
+      "step": 10478
+    },
+    {
+      "epoch": 0.4478680559994886,
+      "grad_norm": 1.4679442644119263,
+      "learning_rate": 3.9061516130628475e-05,
+      "loss": 1.617,
+      "step": 10509
+    },
+    {
+      "epoch": 0.44918920070745166,
+      "grad_norm": 1.5881776809692383,
+      "learning_rate": 3.8991998235695025e-05,
+      "loss": 1.615,
+      "step": 10540
+    },
+    {
+      "epoch": 0.4505103454154148,
+      "grad_norm": 1.4944725036621094,
+      "learning_rate": 3.8922322442963224e-05,
+      "loss": 1.6299,
+      "step": 10571
+    },
+    {
+      "epoch": 0.45183149012337787,
+      "grad_norm": 1.499192714691162,
+      "learning_rate": 3.885248953871491e-05,
+      "loss": 1.6173,
+      "step": 10602
+    },
+    {
+      "epoch": 0.45315263483134094,
+      "grad_norm": 1.409706473350525,
+      "learning_rate": 3.8782500311004915e-05,
+      "loss": 1.6083,
+      "step": 10633
+    },
+    {
+      "epoch": 0.45447377953930407,
+      "grad_norm": 1.3845860958099365,
+      "learning_rate": 3.871235554965218e-05,
+      "loss": 1.6104,
+      "step": 10664
+    },
+    {
+      "epoch": 0.45579492424726714,
+      "grad_norm": 1.4976452589035034,
+      "learning_rate": 3.864205604623078e-05,
+      "loss": 1.6123,
+      "step": 10695
+    },
+    {
+      "epoch": 0.4571160689552302,
+      "grad_norm": 1.496371865272522,
+      "learning_rate": 3.857160259406107e-05,
+      "loss": 1.594,
+      "step": 10726
+    },
+    {
+      "epoch": 0.45843721366319334,
+      "grad_norm": 1.364917516708374,
+      "learning_rate": 3.8500995988200674e-05,
+      "loss": 1.602,
+      "step": 10757
+    },
+    {
+      "epoch": 0.4597583583711564,
+      "grad_norm": 1.461065411567688,
+      "learning_rate": 3.843023702543556e-05,
+      "loss": 1.5936,
+      "step": 10788
+    },
+    {
+      "epoch": 0.46107950307911955,
+      "grad_norm": 1.422227382659912,
+      "learning_rate": 3.8359326504270984e-05,
+      "loss": 1.6275,
+      "step": 10819
+    },
+    {
+      "epoch": 0.4624006477870826,
+      "grad_norm": 1.3532990217208862,
+      "learning_rate": 3.828826522492255e-05,
+      "loss": 1.5982,
+      "step": 10850
+    },
+    {
+      "epoch": 0.4637217924950457,
+      "grad_norm": 1.6219929456710815,
+      "learning_rate": 3.821705398930713e-05,
+      "loss": 1.6225,
+      "step": 10881
+    },
+    {
+      "epoch": 0.4650429372030088,
+      "grad_norm": 1.4138596057891846,
+      "learning_rate": 3.814569360103385e-05,
+      "loss": 1.6216,
+      "step": 10912
+    },
+    {
+      "epoch": 0.4663640819109719,
+      "grad_norm": 1.492357611656189,
+      "learning_rate": 3.807418486539499e-05,
+      "loss": 1.6049,
+      "step": 10943
+    },
+    {
+      "epoch": 0.46768522661893497,
+      "grad_norm": 1.4756381511688232,
+      "learning_rate": 3.80025285893569e-05,
+      "loss": 1.6077,
+      "step": 10974
+    },
+    {
+      "epoch": 0.4690063713268981,
+      "grad_norm": 1.4480361938476562,
+      "learning_rate": 3.793072558155093e-05,
+      "loss": 1.6131,
+      "step": 11005
+    },
+    {
+      "epoch": 0.47032751603486117,
+      "grad_norm": 1.5386732816696167,
+      "learning_rate": 3.785877665226426e-05,
+      "loss": 1.6124,
+      "step": 11036
+    },
+    {
+      "epoch": 0.47164866074282424,
+      "grad_norm": 1.4616602659225464,
+      "learning_rate": 3.778668261343079e-05,
+      "loss": 1.6177,
+      "step": 11067
+    },
+    {
+      "epoch": 0.4729698054507874,
+      "grad_norm": 1.4740066528320312,
+      "learning_rate": 3.771444427862192e-05,
+      "loss": 1.6194,
+      "step": 11098
+    },
+    {
+      "epoch": 0.47429095015875045,
+      "grad_norm": 1.3635661602020264,
+      "learning_rate": 3.7642062463037465e-05,
+      "loss": 1.6085,
+      "step": 11129
+    },
+    {
+      "epoch": 0.4756120948667135,
+      "grad_norm": 1.4568270444869995,
+      "learning_rate": 3.7569537983496373e-05,
+      "loss": 1.6171,
+      "step": 11160
+    },
+    {
+      "epoch": 0.47693323957467665,
+      "grad_norm": 1.4151735305786133,
+      "learning_rate": 3.749687165842753e-05,
+      "loss": 1.6176,
+      "step": 11191
+    },
+    {
+      "epoch": 0.4782543842826397,
+      "grad_norm": 1.3976393938064575,
+      "learning_rate": 3.7424064307860536e-05,
+      "loss": 1.6326,
+      "step": 11222
+    },
+    {
+      "epoch": 0.47957552899060285,
+      "grad_norm": 1.4214377403259277,
+      "learning_rate": 3.735111675341645e-05,
+      "loss": 1.6353,
+      "step": 11253
+    },
+    {
+      "epoch": 0.4808966736985659,
+      "grad_norm": 1.453192949295044,
+      "learning_rate": 3.7278029818298524e-05,
+      "loss": 1.6225,
+      "step": 11284
+    },
+    {
+      "epoch": 0.482217818406529,
+      "grad_norm": 1.3557348251342773,
+      "learning_rate": 3.720480432728287e-05,
+      "loss": 1.6195,
+      "step": 11315
+    },
+    {
+      "epoch": 0.4835389631144921,
+      "grad_norm": 1.411064863204956,
+      "learning_rate": 3.71314411067092e-05,
+      "loss": 1.6026,
+      "step": 11346
+    },
+    {
+      "epoch": 0.4848601078224552,
+      "grad_norm": 1.4603424072265625,
+      "learning_rate": 3.70579409844715e-05,
+      "loss": 1.6133,
+      "step": 11377
+    },
+    {
+      "epoch": 0.4861812525304183,
+      "grad_norm": 1.484947681427002,
+      "learning_rate": 3.698430479000865e-05,
+      "loss": 1.5991,
+      "step": 11408
+    },
+    {
+      "epoch": 0.4875023972383814,
+      "grad_norm": 1.4569475650787354,
+      "learning_rate": 3.691053335429509e-05,
+      "loss": 1.6016,
+      "step": 11439
+    },
+    {
+      "epoch": 0.4888235419463445,
+      "grad_norm": 1.3495190143585205,
+      "learning_rate": 3.683662750983147e-05,
+      "loss": 1.6119,
+      "step": 11470
+    },
+    {
+      "epoch": 0.49014468665430755,
+      "grad_norm": 1.481056809425354,
+      "learning_rate": 3.676258809063518e-05,
+      "loss": 1.5983,
+      "step": 11501
+    },
+    {
+      "epoch": 0.4914658313622707,
+      "grad_norm": 1.4746501445770264,
+      "learning_rate": 3.6688415932231004e-05,
+      "loss": 1.624,
+      "step": 11532
+    },
+    {
+      "epoch": 0.49278697607023375,
+      "grad_norm": 1.340066909790039,
+      "learning_rate": 3.661411187164166e-05,
+      "loss": 1.5697,
+      "step": 11563
+    },
+    {
+      "epoch": 0.4941081207781968,
+      "grad_norm": 1.3957117795944214,
+      "learning_rate": 3.65396767473784e-05,
+      "loss": 1.608,
+      "step": 11594
+    },
+    {
+      "epoch": 0.49542926548615995,
+      "grad_norm": 1.448289155960083,
+      "learning_rate": 3.6465111399431465e-05,
+      "loss": 1.6024,
+      "step": 11625
+    },
+    {
+      "epoch": 0.496750410194123,
+      "grad_norm": 1.3630342483520508,
+      "learning_rate": 3.6390416669260674e-05,
+      "loss": 1.5887,
+      "step": 11656
+    },
+    {
+      "epoch": 0.49807155490208616,
+      "grad_norm": 1.4242172241210938,
+      "learning_rate": 3.63155933997859e-05,
+      "loss": 1.5989,
+      "step": 11687
+    },
+    {
+      "epoch": 0.49939269961004923,
+      "grad_norm": 1.4106690883636475,
+      "learning_rate": 3.624064243537758e-05,
+      "loss": 1.6133,
+      "step": 11718
+    },
+    {
+      "epoch": 0.5007138443180124,
+      "grad_norm": 1.3867900371551514,
+      "learning_rate": 3.616556462184716e-05,
+      "loss": 1.5766,
+      "step": 11749
+    },
+    {
+      "epoch": 0.5020349890259754,
+      "grad_norm": 1.4181773662567139,
+      "learning_rate": 3.609036080643755e-05,
+      "loss": 1.6259,
+      "step": 11780
+    },
+    {
+      "epoch": 0.5033561337339385,
+      "grad_norm": 1.434336543083191,
+      "learning_rate": 3.60150318378136e-05,
+      "loss": 1.6021,
+      "step": 11811
+    },
+    {
+      "epoch": 0.5046772784419016,
+      "grad_norm": 1.370049238204956,
+      "learning_rate": 3.5939578566052465e-05,
+      "loss": 1.5802,
+      "step": 11842
+    },
+    {
+      "epoch": 0.5059984231498647,
+      "grad_norm": 1.3584986925125122,
+      "learning_rate": 3.586400184263408e-05,
+      "loss": 1.5964,
+      "step": 11873
+    },
+    {
+      "epoch": 0.5073195678578278,
+      "grad_norm": 1.458137035369873,
+      "learning_rate": 3.578830252043148e-05,
+      "loss": 1.5962,
+      "step": 11904
+    },
+    {
+      "epoch": 0.5086407125657909,
+      "grad_norm": 1.4329200983047485,
+      "learning_rate": 3.571248145370125e-05,
+      "loss": 1.6046,
+      "step": 11935
+    },
+    {
+      "epoch": 0.509961857273754,
+      "grad_norm": 1.3814787864685059,
+      "learning_rate": 3.5636539498073794e-05,
+      "loss": 1.5901,
+      "step": 11966
+    },
+    {
+      "epoch": 0.5112830019817171,
+      "grad_norm": 1.3795075416564941,
+      "learning_rate": 3.556047751054378e-05,
+      "loss": 1.621,
+      "step": 11997
+    },
+    {
+      "epoch": 0.5126041466896801,
+      "grad_norm": 1.4559143781661987,
+      "learning_rate": 3.548429634946039e-05,
+      "loss": 1.5845,
+      "step": 12028
+    },
+    {
+      "epoch": 0.5139252913976432,
+      "grad_norm": 1.468773365020752,
+      "learning_rate": 3.540799687451768e-05,
+      "loss": 1.5956,
+      "step": 12059
+    },
+    {
+      "epoch": 0.5152464361056064,
+      "grad_norm": 1.5287885665893555,
+      "learning_rate": 3.533157994674485e-05,
+      "loss": 1.5801,
+      "step": 12090
+    },
+    {
+      "epoch": 0.5165675808135695,
+      "grad_norm": 1.4891117811203003,
+      "learning_rate": 3.5255046428496546e-05,
+      "loss": 1.6034,
+      "step": 12121
+    },
+    {
+      "epoch": 0.5178887255215325,
+      "grad_norm": 1.541502594947815,
+      "learning_rate": 3.517839718344311e-05,
+      "loss": 1.6076,
+      "step": 12152
+    },
+    {
+      "epoch": 0.5192098702294956,
+      "grad_norm": 1.3209787607192993,
+      "learning_rate": 3.510163307656086e-05,
+      "loss": 1.5967,
+      "step": 12183
+    },
+    {
+      "epoch": 0.5205310149374587,
+      "grad_norm": 1.534022569656372,
+      "learning_rate": 3.5024754974122324e-05,
+      "loss": 1.5808,
+      "step": 12214
+    },
+    {
+      "epoch": 0.5218521596454218,
+      "grad_norm": 1.460730791091919,
+      "learning_rate": 3.494776374368643e-05,
+      "loss": 1.6052,
+      "step": 12245
+    },
+    {
+      "epoch": 0.5231733043533849,
+      "grad_norm": 1.4722791910171509,
+      "learning_rate": 3.4870660254088724e-05,
+      "loss": 1.5959,
+      "step": 12276
+    },
+    {
+      "epoch": 0.524494449061348,
+      "grad_norm": 1.345739722251892,
+      "learning_rate": 3.479344537543164e-05,
+      "loss": 1.5888,
+      "step": 12307
+    },
+    {
+      "epoch": 0.5258155937693111,
+      "grad_norm": 1.3360117673873901,
+      "learning_rate": 3.4716119979074565e-05,
+      "loss": 1.5933,
+      "step": 12338
+    },
+    {
+      "epoch": 0.5271367384772742,
+      "grad_norm": 1.4767347574234009,
+      "learning_rate": 3.463868493762412e-05,
+      "loss": 1.5741,
+      "step": 12369
+    },
+    {
+      "epoch": 0.5284578831852372,
+      "grad_norm": 1.364464521408081,
+      "learning_rate": 3.456114112492418e-05,
+      "loss": 1.6054,
+      "step": 12400
+    },
+    {
+      "epoch": 0.5297790278932004,
+      "grad_norm": 1.4622302055358887,
+      "learning_rate": 3.4483489416046164e-05,
+      "loss": 1.5907,
+      "step": 12431
+    },
+    {
+      "epoch": 0.5311001726011635,
+      "grad_norm": 1.4068593978881836,
+      "learning_rate": 3.440573068727905e-05,
+      "loss": 1.6045,
+      "step": 12462
+    },
+    {
+      "epoch": 0.5324213173091266,
+      "grad_norm": 1.449144959449768,
+      "learning_rate": 3.4327865816119495e-05,
+      "loss": 1.5987,
+      "step": 12493
+    },
+    {
+      "epoch": 0.5337424620170896,
+      "grad_norm": 1.4499784708023071,
+      "learning_rate": 3.4249895681262025e-05,
+      "loss": 1.585,
+      "step": 12524
+    },
+    {
+      "epoch": 0.5350636067250527,
+      "grad_norm": 1.4232062101364136,
+      "learning_rate": 3.417182116258899e-05,
+      "loss": 1.605,
+      "step": 12555
+    },
+    {
+      "epoch": 0.5363847514330158,
+      "grad_norm": 1.454372763633728,
+      "learning_rate": 3.409364314116074e-05,
+      "loss": 1.6045,
+      "step": 12586
+    },
+    {
+      "epoch": 0.537705896140979,
+      "grad_norm": 1.3590772151947021,
+      "learning_rate": 3.401536249920559e-05,
+      "loss": 1.5772,
+      "step": 12617
+    },
+    {
+      "epoch": 0.539027040848942,
+      "grad_norm": 1.3528358936309814,
+      "learning_rate": 3.393698012010998e-05,
+      "loss": 1.5792,
+      "step": 12648
+    },
+    {
+      "epoch": 0.5403481855569051,
+      "grad_norm": 1.3940211534500122,
+      "learning_rate": 3.385849688840839e-05,
+      "loss": 1.5981,
+      "step": 12679
+    },
+    {
+      "epoch": 0.5416693302648682,
+      "grad_norm": 1.5671747922897339,
+      "learning_rate": 3.3779913689773414e-05,
+      "loss": 1.596,
+      "step": 12710
+    },
+    {
+      "epoch": 0.5429904749728313,
+      "grad_norm": 1.5250004529953003,
+      "learning_rate": 3.370123141100578e-05,
+      "loss": 1.6173,
+      "step": 12741
+    },
+    {
+      "epoch": 0.5443116196807944,
+      "grad_norm": 1.4243561029434204,
+      "learning_rate": 3.3622450940024305e-05,
+      "loss": 1.5748,
+      "step": 12772
+    },
+    {
+      "epoch": 0.5456327643887575,
+      "grad_norm": 1.3759068250656128,
+      "learning_rate": 3.35435731658559e-05,
+      "loss": 1.5953,
+      "step": 12803
+    },
+    {
+      "epoch": 0.5469539090967206,
+      "grad_norm": 1.3891198635101318,
+      "learning_rate": 3.346459897862552e-05,
+      "loss": 1.5693,
+      "step": 12834
+    },
+    {
+      "epoch": 0.5482750538046837,
+      "grad_norm": 1.341163992881775,
+      "learning_rate": 3.338552926954613e-05,
+      "loss": 1.5937,
+      "step": 12865
+    },
+    {
+      "epoch": 0.5495961985126467,
+      "grad_norm": 1.4457162618637085,
+      "learning_rate": 3.330636493090868e-05,
+      "loss": 1.5821,
+      "step": 12896
+    },
+    {
+      "epoch": 0.5509173432206098,
+      "grad_norm": 1.3722436428070068,
+      "learning_rate": 3.322710685607193e-05,
+      "loss": 1.5847,
+      "step": 12927
+    },
+    {
+      "epoch": 0.552238487928573,
+      "grad_norm": 1.3744261264801025,
+      "learning_rate": 3.314775593945251e-05,
+      "loss": 1.5668,
+      "step": 12958
+    },
+    {
+      "epoch": 0.5535596326365361,
+      "grad_norm": 1.5389301776885986,
+      "learning_rate": 3.3068313076514714e-05,
+      "loss": 1.6037,
+      "step": 12989
+    },
+    {
+      "epoch": 0.5548807773444991,
+      "grad_norm": 1.3751068115234375,
+      "learning_rate": 3.298877916376047e-05,
+      "loss": 1.5938,
+      "step": 13020
+    },
+    {
+      "epoch": 0.5562019220524622,
+      "grad_norm": 1.442494511604309,
+      "learning_rate": 3.290915509871915e-05,
+      "loss": 1.5744,
+      "step": 13051
+    },
+    {
+      "epoch": 0.5575230667604253,
+      "grad_norm": 1.4287244081497192,
+      "learning_rate": 3.282944177993753e-05,
+      "loss": 1.5788,
+      "step": 13082
+    },
+    {
+      "epoch": 0.5588442114683884,
+      "grad_norm": 1.4278240203857422,
+      "learning_rate": 3.274964010696957e-05,
+      "loss": 1.569,
+      "step": 13113
+    },
+    {
+      "epoch": 0.5601653561763515,
+      "grad_norm": 1.3635144233703613,
+      "learning_rate": 3.266975098036629e-05,
+      "loss": 1.5824,
+      "step": 13144
+    },
+    {
+      "epoch": 0.5614865008843146,
+      "grad_norm": 1.4243277311325073,
+      "learning_rate": 3.258977530166562e-05,
+      "loss": 1.5832,
+      "step": 13175
+    },
+    {
+      "epoch": 0.5628076455922777,
+      "grad_norm": 1.4209073781967163,
+      "learning_rate": 3.250971397338227e-05,
+      "loss": 1.5813,
+      "step": 13206
+    },
+    {
+      "epoch": 0.5641287903002408,
+      "grad_norm": 1.4427789449691772,
+      "learning_rate": 3.2429567898997404e-05,
+      "loss": 1.5671,
+      "step": 13237
+    },
+    {
+      "epoch": 0.5654499350082038,
+      "grad_norm": 1.40023672580719,
+      "learning_rate": 3.234933798294859e-05,
+      "loss": 1.5906,
+      "step": 13268
+    },
+    {
+      "epoch": 0.566771079716167,
+      "grad_norm": 1.3432368040084839,
+      "learning_rate": 3.2269025130619535e-05,
+      "loss": 1.5754,
+      "step": 13299
+    },
+    {
+      "epoch": 0.5680922244241301,
+      "grad_norm": 1.3551653623580933,
+      "learning_rate": 3.218863024832985e-05,
+      "loss": 1.5729,
+      "step": 13330
+    },
+    {
+      "epoch": 0.5694133691320932,
+      "grad_norm": 1.4137156009674072,
+      "learning_rate": 3.2108154243324864e-05,
+      "loss": 1.5852,
+      "step": 13361
+    },
+    {
+      "epoch": 0.5707345138400562,
+      "grad_norm": 1.408336877822876,
+      "learning_rate": 3.2027598023765345e-05,
+      "loss": 1.5923,
+      "step": 13392
+    },
+    {
+      "epoch": 0.5720556585480193,
+      "grad_norm": 1.4031184911727905,
+      "learning_rate": 3.194696249871729e-05,
+      "loss": 1.5958,
+      "step": 13423
+    },
+    {
+      "epoch": 0.5733768032559824,
+      "grad_norm": 1.3647385835647583,
+      "learning_rate": 3.186624857814164e-05,
+      "loss": 1.6011,
+      "step": 13454
+    },
+    {
+      "epoch": 0.5746979479639456,
+      "grad_norm": 1.3675132989883423,
+      "learning_rate": 3.178545717288401e-05,
+      "loss": 1.5905,
+      "step": 13485
+    },
+    {
+      "epoch": 0.5760190926719087,
+      "grad_norm": 1.3849265575408936,
+      "learning_rate": 3.170458919466444e-05,
+      "loss": 1.5834,
+      "step": 13516
+    },
+    {
+      "epoch": 0.5773402373798717,
+      "grad_norm": 1.3320952653884888,
+      "learning_rate": 3.1623645556067063e-05,
+      "loss": 1.5897,
+      "step": 13547
+    },
+    {
+      "epoch": 0.5786613820878348,
+      "grad_norm": 1.3240253925323486,
+      "learning_rate": 3.154262717052985e-05,
+      "loss": 1.5587,
+      "step": 13578
+    },
+    {
+      "epoch": 0.5799825267957979,
+      "grad_norm": 1.4849603176116943,
+      "learning_rate": 3.146153495233426e-05,
+      "loss": 1.5845,
+      "step": 13609
+    },
+    {
+      "epoch": 0.581303671503761,
+      "grad_norm": 1.3303260803222656,
+      "learning_rate": 3.1380369816594944e-05,
+      "loss": 1.5895,
+      "step": 13640
+    },
+    {
+      "epoch": 0.5826248162117241,
+      "grad_norm": 1.4952738285064697,
+      "learning_rate": 3.129913267924946e-05,
+      "loss": 1.5909,
+      "step": 13671
+    },
+    {
+      "epoch": 0.5839459609196872,
+      "grad_norm": 1.3588736057281494,
+      "learning_rate": 3.121782445704782e-05,
+      "loss": 1.5923,
+      "step": 13702
+    },
+    {
+      "epoch": 0.5852671056276503,
+      "grad_norm": 1.2940070629119873,
+      "learning_rate": 3.11364460675423e-05,
+      "loss": 1.5703,
+      "step": 13733
+    },
+    {
+      "epoch": 0.5865882503356133,
+      "grad_norm": 1.4225587844848633,
+      "learning_rate": 3.1054998429076934e-05,
+      "loss": 1.6073,
+      "step": 13764
+    },
+    {
+      "epoch": 0.5879093950435764,
+      "grad_norm": 1.3482099771499634,
+      "learning_rate": 3.097348246077728e-05,
+      "loss": 1.5872,
+      "step": 13795
+    },
+    {
+      "epoch": 0.5892305397515396,
+      "grad_norm": 1.4100350141525269,
+      "learning_rate": 3.0891899082539924e-05,
+      "loss": 1.5852,
+      "step": 13826
+    },
+    {
+      "epoch": 0.5905516844595027,
+      "grad_norm": 1.4312938451766968,
+      "learning_rate": 3.0810249215022233e-05,
+      "loss": 1.5729,
+      "step": 13857
+    },
+    {
+      "epoch": 0.5918728291674658,
+      "grad_norm": 1.5027344226837158,
+      "learning_rate": 3.0728533779631865e-05,
+      "loss": 1.5712,
+      "step": 13888
+    },
+    {
+      "epoch": 0.5931939738754288,
+      "grad_norm": 1.3554552793502808,
+      "learning_rate": 3.064675369851637e-05,
+      "loss": 1.5733,
+      "step": 13919
+    },
+    {
+      "epoch": 0.5945151185833919,
+      "grad_norm": 1.356598973274231,
+      "learning_rate": 3.056490989455289e-05,
+      "loss": 1.5594,
+      "step": 13950
+    },
+    {
+      "epoch": 0.595836263291355,
+      "grad_norm": 1.406317114830017,
+      "learning_rate": 3.0483003291337596e-05,
+      "loss": 1.5817,
+      "step": 13981
+    },
+    {
+      "epoch": 0.5971574079993182,
+      "grad_norm": 1.3264002799987793,
+      "learning_rate": 3.040103481317539e-05,
+      "loss": 1.6009,
+      "step": 14012
+    },
+    {
+      "epoch": 0.5984785527072812,
+      "grad_norm": 1.4850720167160034,
+      "learning_rate": 3.03190053850694e-05,
+      "loss": 1.5703,
+      "step": 14043
+    },
+    {
+      "epoch": 0.5997996974152443,
+      "grad_norm": 1.45437753200531,
+      "learning_rate": 3.0236915932710573e-05,
+      "loss": 1.5872,
+      "step": 14074
+    },
+    {
+      "epoch": 0.6011208421232074,
+      "grad_norm": 1.4236788749694824,
+      "learning_rate": 3.0154767382467232e-05,
+      "loss": 1.5897,
+      "step": 14105
+    },
+    {
+      "epoch": 0.6024419868311705,
+      "grad_norm": 1.3803141117095947,
+      "learning_rate": 3.0072560661374582e-05,
+      "loss": 1.5826,
+      "step": 14136
+    },
+    {
+      "epoch": 0.6037631315391336,
+      "grad_norm": 1.3711870908737183,
+      "learning_rate": 2.999029669712431e-05,
+      "loss": 1.5959,
+      "step": 14167
+    },
+    {
+      "epoch": 0.6050842762470967,
+      "grad_norm": 1.3403728008270264,
+      "learning_rate": 2.990797641805408e-05,
+      "loss": 1.571,
+      "step": 14198
+    },
+    {
+      "epoch": 0.6064054209550598,
+      "grad_norm": 1.4036363363265991,
+      "learning_rate": 2.982560075313704e-05,
+      "loss": 1.5784,
+      "step": 14229
+    },
+    {
+      "epoch": 0.6077265656630229,
+      "grad_norm": 1.361581563949585,
+      "learning_rate": 2.9743170631971368e-05,
+      "loss": 1.5865,
+      "step": 14260
+    },
+    {
+      "epoch": 0.6090477103709859,
+      "grad_norm": 1.4343942403793335,
+      "learning_rate": 2.9660686984769792e-05,
+      "loss": 1.5636,
+      "step": 14291
+    },
+    {
+      "epoch": 0.610368855078949,
+      "grad_norm": 1.4140040874481201,
+      "learning_rate": 2.9578150742349047e-05,
+      "loss": 1.5643,
+      "step": 14322
+    },
+    {
+      "epoch": 0.6116899997869122,
+      "grad_norm": 1.4035499095916748,
+      "learning_rate": 2.949556283611942e-05,
+      "loss": 1.5924,
+      "step": 14353
+    },
+    {
+      "epoch": 0.6130111444948753,
+      "grad_norm": 1.4702353477478027,
+      "learning_rate": 2.9412924198074206e-05,
+      "loss": 1.5725,
+      "step": 14384
+    },
+    {
+      "epoch": 0.6143322892028383,
+      "grad_norm": 1.342893362045288,
+      "learning_rate": 2.9330235760779208e-05,
+      "loss": 1.5971,
+      "step": 14415
+    },
+    {
+      "epoch": 0.6156534339108014,
+      "grad_norm": 1.3674631118774414,
+      "learning_rate": 2.9247498457362188e-05,
+      "loss": 1.5733,
+      "step": 14446
+    },
+    {
+      "epoch": 0.6169745786187645,
+      "grad_norm": 1.3972622156143188,
+      "learning_rate": 2.9164713221502373e-05,
+      "loss": 1.562,
+      "step": 14477
+    },
+    {
+      "epoch": 0.6182957233267277,
+      "grad_norm": 1.3173855543136597,
+      "learning_rate": 2.9081880987419912e-05,
+      "loss": 1.5786,
+      "step": 14508
+    },
+    {
+      "epoch": 0.6196168680346907,
+      "grad_norm": 1.3843812942504883,
+      "learning_rate": 2.8999002689865296e-05,
+      "loss": 1.5532,
+      "step": 14539
+    },
+    {
+      "epoch": 0.6209380127426538,
+      "grad_norm": 1.350704550743103,
+      "learning_rate": 2.8916079264108852e-05,
+      "loss": 1.5588,
+      "step": 14570
+    },
+    {
+      "epoch": 0.6222591574506169,
+      "grad_norm": 1.460552453994751,
+      "learning_rate": 2.883311164593017e-05,
+      "loss": 1.5636,
+      "step": 14601
+    },
+    {
+      "epoch": 0.62358030215858,
+      "grad_norm": 1.288114070892334,
+      "learning_rate": 2.875010077160754e-05,
+      "loss": 1.5355,
+      "step": 14632
+    },
+    {
+      "epoch": 0.624901446866543,
+      "grad_norm": 1.3940645456314087,
+      "learning_rate": 2.866704757790741e-05,
+      "loss": 1.5562,
+      "step": 14663
+    },
+    {
+      "epoch": 0.6262225915745062,
+      "grad_norm": 1.4258652925491333,
+      "learning_rate": 2.858395300207376e-05,
+      "loss": 1.5741,
+      "step": 14694
+    },
+    {
+      "epoch": 0.6275437362824693,
+      "grad_norm": 1.2987011671066284,
+      "learning_rate": 2.8500817981817607e-05,
+      "loss": 1.5716,
+      "step": 14725
+    },
+    {
+      "epoch": 0.6288648809904324,
+      "grad_norm": 1.345184326171875,
+      "learning_rate": 2.8417643455306336e-05,
+      "loss": 1.5757,
+      "step": 14756
+    },
+    {
+      "epoch": 0.6301860256983954,
+      "grad_norm": 1.3593329191207886,
+      "learning_rate": 2.8334430361153185e-05,
+      "loss": 1.5769,
+      "step": 14787
+    },
+    {
+      "epoch": 0.6315071704063585,
+      "grad_norm": 1.369004726409912,
+      "learning_rate": 2.8251179638406612e-05,
+      "loss": 1.5681,
+      "step": 14818
+    },
+    {
+      "epoch": 0.6328283151143216,
+      "grad_norm": 1.3379971981048584,
+      "learning_rate": 2.8167892226539704e-05,
+      "loss": 1.561,
+      "step": 14849
+    },
+    {
+      "epoch": 0.6341494598222848,
+      "grad_norm": 1.3294727802276611,
+      "learning_rate": 2.8084569065439588e-05,
+      "loss": 1.583,
+      "step": 14880
+    },
+    {
+      "epoch": 0.6354706045302478,
+      "grad_norm": 1.3562073707580566,
+      "learning_rate": 2.8001211095396807e-05,
+      "loss": 1.5758,
+      "step": 14911
+    },
+    {
+      "epoch": 0.6367917492382109,
+      "grad_norm": 1.359453797340393,
+      "learning_rate": 2.791781925709473e-05,
+      "loss": 1.5689,
+      "step": 14942
+    },
+    {
+      "epoch": 0.638112893946174,
+      "grad_norm": 1.3837451934814453,
+      "learning_rate": 2.7834394491598908e-05,
+      "loss": 1.5644,
+      "step": 14973
+    },
+    {
+      "epoch": 0.6394340386541371,
+      "grad_norm": 1.37912118434906,
+      "learning_rate": 2.7750937740346485e-05,
+      "loss": 1.5692,
+      "step": 15004
+    },
+    {
+      "epoch": 0.6407551833621002,
+      "grad_norm": 1.3171929121017456,
+      "learning_rate": 2.7667449945135564e-05,
+      "loss": 1.5773,
+      "step": 15035
+    },
+    {
+      "epoch": 0.6420763280700633,
+      "grad_norm": 1.3926173448562622,
+      "learning_rate": 2.7583932048114557e-05,
+      "loss": 1.5637,
+      "step": 15066
+    },
+    {
+      "epoch": 0.6433974727780264,
+      "grad_norm": 1.3726190328598022,
+      "learning_rate": 2.7500384991771587e-05,
+      "loss": 1.5847,
+      "step": 15097
+    },
+    {
+      "epoch": 0.6447186174859895,
+      "grad_norm": 1.4371352195739746,
+      "learning_rate": 2.7416809718923825e-05,
+      "loss": 1.5591,
+      "step": 15128
+    },
+    {
+      "epoch": 0.6460397621939525,
+      "grad_norm": 1.2980587482452393,
+      "learning_rate": 2.7333207172706864e-05,
+      "loss": 1.564,
+      "step": 15159
+    },
+    {
+      "epoch": 0.6473609069019156,
+      "grad_norm": 1.3802196979522705,
+      "learning_rate": 2.7249578296564088e-05,
+      "loss": 1.5816,
+      "step": 15190
+    },
+    {
+      "epoch": 0.6486820516098788,
+      "grad_norm": 1.349716067314148,
+      "learning_rate": 2.7165924034235973e-05,
+      "loss": 1.5671,
+      "step": 15221
+    },
+    {
+      "epoch": 0.6500031963178419,
+      "grad_norm": 1.3532193899154663,
+      "learning_rate": 2.708224532974953e-05,
+      "loss": 1.5547,
+      "step": 15252
+    }
+  ],
+  "logging_steps": 31,
+  "max_steps": 30517,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 2,
+  "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": 1.1319725270111355e+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 sha256:302cac6bc00d87f6c817113f3074af210729435b710bdeed411138aae3385920
+size 5304

checkpoint-18312/config.json

@@ -0,0 +1,39 @@
+{
+  "_name_or_path": "/mnt/parscratch/users/acp23ay/private/models/Llama-3.1-8B-Instruct-gu-madlad-mean/",
+  "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": 138256
+}

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 sha256:820370f31c37e7b2060189d26f94b9f6165c96207ca32094f388708752278940
+size 4983197184

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

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:168629c67732309d49f60a5ec48a6d160212bc987365e82e183dfbf74ba0c1f3
+size 4899116432

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 sha256:7d4765816ecf8913b7a7566f63b472d3214ad2258990eacac3cfe3479de8b1a9
+size 4999813120

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

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1e2e6ca923e9e3c8735f2dbe22eaa0cd0a32cac4d2d7317f0024d5728b6479cb
+size 2734998184

checkpoint-18312/model.safetensors.index.json

@@ -0,0 +1,298 @@
+{
+  "metadata": {
+    "total_size": 32448724992
+  },
+  "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-00002-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 sha256:2c916de849de54289b84553212967eb2009a89339ad571ce1f71f9688bb0e099
+size 16040396334

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 sha256:302cac6bc00d87f6c817113f3074af210729435b710bdeed411138aae3385920
+size 5304

checkpoint-21364/config.json

@@ -0,0 +1,39 @@
+{
+  "_name_or_path": "/mnt/parscratch/users/acp23ay/private/models/Llama-3.1-8B-Instruct-gu-madlad-mean/",
+  "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": 138256
+}

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:c2a421aa1d64b3d5a7553eeeb1f06f18bf3882cec251bd92b0058372780febaa
+size 4983197184

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

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:168629c67732309d49f60a5ec48a6d160212bc987365e82e183dfbf74ba0c1f3
+size 4899116432

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

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