Upload folder using huggingface_hub

config.json

@@ -0,0 +1,38 @@
+{
+  "_name_or_path": "/mnt/stardoc/escher/checkpoints/Emu3-Base-SFT-without_cot-Mar04_256_lr1e-4/checkpoint-8400",
+  "architectures": [
+    "Emu3ForCausalLM"
+  ],
+  "attention_dropout": 0.1,
+  "auto_map": {
+    "AutoConfig": "BAAI/Emu3-Stage1--configuration_emu3.Emu3Config",
+    "AutoModelForCausalLM": "BAAI/Emu3-Stage1--modeling_emu3.Emu3ForCausalLM"
+  },
+  "boi_token_id": 151852,
+  "bos_token_id": 151849,
+  "eof_token_id": 151847,
+  "eoi_token_id": 151853,
+  "eol_token_id": 151846,
+  "eos_token_id": 151850,
+  "hidden_act": "silu",
+  "hidden_size": 4096,
+  "image_area": 65536,
+  "img_token_id": 151851,
+  "initializer_range": 0.02,
+  "intermediate_size": 14336,
+  "max_position_embeddings": 4300,
+  "model_type": "Emu3",
+  "num_attention_heads": 32,
+  "num_hidden_layers": 32,
+  "num_key_value_heads": 8,
+  "pad_token_id": 151643,
+  "pretraining_tp": 1,
+  "rms_norm_eps": 1e-05,
+  "rope_scaling": null,
+  "rope_theta": 1000000.0,
+  "tie_word_embeddings": false,
+  "torch_dtype": "bfloat16",
+  "transformers_version": "4.44.0",
+  "use_cache": false,
+  "vocab_size": 184622
+}

generation_config.json

@@ -0,0 +1,7 @@
+{
+  "_from_model_config": true,
+  "bos_token_id": 151849,
+  "eos_token_id": 151850,
+  "pad_token_id": 151643,
+  "transformers_version": "4.44.0"
+}

latest

@@ -0,0 +1 @@
+global_step10800

model-00001-of-00004.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c950d53ff85a875821ef8d5cb755585ca5113fea940843a458f280a6625c2d2d
+size 4884766656

model-00002-of-00004.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6015ee78d8bbce3a82ecdfb889a81f0ca7f163a441ff312d06c578402528f9b5
+size 4999819320

model-00003-of-00004.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2ec098dadee1d0ec80425cd842767ff3f9bdb624db80d414167515503cf1c1b5
+size 4915916184

model-00004-of-00004.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3d59e70779b24544e822436ab27875b960c8d9ae8e2725aa0dfb535e64119bae
+size 2183554760

model.safetensors.index.json

@@ -0,0 +1,298 @@
+{
+  "metadata": {
+    "total_size": 16984023040
+  },
+  "weight_map": {
+    "lm_head.weight": "model-00004-of-00004.safetensors",
+    "model.embed_tokens.weight": "model-00001-of-00004.safetensors",
+    "model.layers.0.input_layernorm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.0.mlp.down_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.0.mlp.gate_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.0.mlp.up_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.0.post_attention_layernorm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.0.self_attn.k_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.0.self_attn.o_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.0.self_attn.q_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.0.self_attn.v_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.1.input_layernorm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.1.mlp.down_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.1.mlp.gate_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.1.mlp.up_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.1.post_attention_layernorm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.1.self_attn.k_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.1.self_attn.o_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.1.self_attn.q_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.1.self_attn.v_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.10.input_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.10.mlp.down_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.10.mlp.gate_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.10.mlp.up_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.10.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.10.self_attn.k_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.10.self_attn.o_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.10.self_attn.q_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.10.self_attn.v_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.11.input_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.11.mlp.down_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.11.mlp.gate_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.11.mlp.up_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.11.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.11.self_attn.k_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.11.self_attn.o_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.11.self_attn.q_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.11.self_attn.v_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.12.input_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.12.mlp.down_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.12.mlp.gate_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.12.mlp.up_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.12.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.12.self_attn.k_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.12.self_attn.o_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.12.self_attn.q_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.12.self_attn.v_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.13.input_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.13.mlp.down_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.13.mlp.gate_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.13.mlp.up_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.13.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.13.self_attn.k_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.13.self_attn.o_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.13.self_attn.q_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.13.self_attn.v_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.14.input_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.14.mlp.down_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.14.mlp.gate_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.14.mlp.up_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.14.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.14.self_attn.k_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.14.self_attn.o_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.14.self_attn.q_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.14.self_attn.v_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.15.input_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.15.mlp.down_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.15.mlp.gate_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.15.mlp.up_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.15.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.15.self_attn.k_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.15.self_attn.o_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.15.self_attn.q_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.15.self_attn.v_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.16.input_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.16.mlp.down_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.16.mlp.gate_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.16.mlp.up_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.16.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.16.self_attn.k_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.16.self_attn.o_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.16.self_attn.q_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.16.self_attn.v_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.17.input_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.17.mlp.down_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.17.mlp.gate_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.17.mlp.up_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.17.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.17.self_attn.k_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.17.self_attn.o_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.17.self_attn.q_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.17.self_attn.v_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.18.input_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.18.mlp.down_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.18.mlp.gate_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.18.mlp.up_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.18.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.18.self_attn.k_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.18.self_attn.o_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.18.self_attn.q_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.18.self_attn.v_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.19.input_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.19.mlp.down_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.19.mlp.gate_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.19.mlp.up_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.19.post_attention_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.19.self_attn.k_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.19.self_attn.o_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.19.self_attn.q_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.19.self_attn.v_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.2.input_layernorm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.2.mlp.down_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.2.mlp.gate_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.2.mlp.up_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.2.post_attention_layernorm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.2.self_attn.k_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.2.self_attn.o_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.2.self_attn.q_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.2.self_attn.v_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.20.input_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.20.mlp.down_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.20.mlp.gate_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.20.mlp.up_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.20.post_attention_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.20.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.20.self_attn.o_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.20.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.20.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.21.input_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.21.mlp.down_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.21.mlp.gate_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.21.mlp.up_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.21.post_attention_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.21.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.21.self_attn.o_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.21.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.21.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.22.input_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.22.mlp.down_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.22.mlp.gate_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.22.mlp.up_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.22.post_attention_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.22.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.22.self_attn.o_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.22.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.22.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.23.input_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.23.mlp.down_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.23.mlp.gate_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.23.mlp.up_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.23.post_attention_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.23.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.23.self_attn.o_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.23.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.23.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.24.input_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.24.mlp.down_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.24.mlp.gate_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.24.mlp.up_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.24.post_attention_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.24.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.24.self_attn.o_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.24.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.24.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.25.input_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.25.mlp.down_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.25.mlp.gate_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.25.mlp.up_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.25.post_attention_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.25.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.25.self_attn.o_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.25.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.25.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.26.input_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.26.mlp.down_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.26.mlp.gate_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.26.mlp.up_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.26.post_attention_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.26.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.26.self_attn.o_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.26.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.26.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.27.input_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.27.mlp.down_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.27.mlp.gate_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.27.mlp.up_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.27.post_attention_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.27.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.27.self_attn.o_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.27.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.27.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.28.input_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.28.mlp.down_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.28.mlp.gate_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.28.mlp.up_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.28.post_attention_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.28.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.28.self_attn.o_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.28.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.28.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.29.input_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.29.mlp.down_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.29.mlp.gate_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.29.mlp.up_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.29.post_attention_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.29.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.29.self_attn.o_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.29.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.29.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.3.input_layernorm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.3.mlp.down_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.3.mlp.gate_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.3.mlp.up_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.3.post_attention_layernorm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.3.self_attn.k_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.3.self_attn.o_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.3.self_attn.q_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.3.self_attn.v_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.30.input_layernorm.weight": "model-00004-of-00004.safetensors",
+    "model.layers.30.mlp.down_proj.weight": "model-00004-of-00004.safetensors",
+    "model.layers.30.mlp.gate_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.30.mlp.up_proj.weight": "model-00004-of-00004.safetensors",
+    "model.layers.30.post_attention_layernorm.weight": "model-00004-of-00004.safetensors",
+    "model.layers.30.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.30.self_attn.o_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.30.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.30.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.31.input_layernorm.weight": "model-00004-of-00004.safetensors",
+    "model.layers.31.mlp.down_proj.weight": "model-00004-of-00004.safetensors",
+    "model.layers.31.mlp.gate_proj.weight": "model-00004-of-00004.safetensors",
+    "model.layers.31.mlp.up_proj.weight": "model-00004-of-00004.safetensors",
+    "model.layers.31.post_attention_layernorm.weight": "model-00004-of-00004.safetensors",
+    "model.layers.31.self_attn.k_proj.weight": "model-00004-of-00004.safetensors",
+    "model.layers.31.self_attn.o_proj.weight": "model-00004-of-00004.safetensors",
+    "model.layers.31.self_attn.q_proj.weight": "model-00004-of-00004.safetensors",
+    "model.layers.31.self_attn.v_proj.weight": "model-00004-of-00004.safetensors",
+    "model.layers.4.input_layernorm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.4.mlp.down_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.4.mlp.gate_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.4.mlp.up_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.4.post_attention_layernorm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.4.self_attn.k_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.4.self_attn.o_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.4.self_attn.q_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.4.self_attn.v_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.5.input_layernorm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.5.mlp.down_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.5.mlp.gate_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.5.mlp.up_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.5.post_attention_layernorm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.5.self_attn.k_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.5.self_attn.o_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.5.self_attn.q_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.5.self_attn.v_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.6.input_layernorm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.6.mlp.down_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.6.mlp.gate_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.6.mlp.up_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.6.post_attention_layernorm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.6.self_attn.k_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.6.self_attn.o_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.6.self_attn.q_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.6.self_attn.v_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.7.input_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.7.mlp.down_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.7.mlp.gate_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.7.mlp.up_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.7.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.7.self_attn.k_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.7.self_attn.o_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.7.self_attn.q_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.7.self_attn.v_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.8.input_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.8.mlp.down_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.8.mlp.gate_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.8.mlp.up_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.8.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.8.self_attn.k_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.8.self_attn.o_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.8.self_attn.q_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.8.self_attn.v_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.9.input_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.9.mlp.down_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.9.mlp.gate_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.9.mlp.up_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.9.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.9.self_attn.k_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.9.self_attn.o_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.9.self_attn.q_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.9.self_attn.v_proj.weight": "model-00002-of-00004.safetensors",
+    "model.norm.weight": "model-00004-of-00004.safetensors"
+  }
+}

rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0020cac19127fc67d6d5960e15fda43dbcb8fb80e14be25ac40ff69dfb58144c
+size 15984

rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:143f2bf9d901ff901e9e2154e08982661143a02b60290b2ce5922def4b4ba3d6
+size 15984

rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:751b067f191bcd682bf2fee60c8557dd60e9eaab4d77cd95921737f386fdb4e6
+size 15984

rng_state_3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d3f7105e072e76c4955155ddbec3394988e98ef3304c216e70c188559608cc06
+size 15984

rng_state_4.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3d6902f393a7113d4eab84561276dd26e0fe79dba1e4a516fa1564e5b92d272e
+size 15984

rng_state_5.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:04193a132d9e2c1feb69b5b27806970167a2247b1136fcf0099dc3be9a10d31d
+size 15984

rng_state_6.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:277ab7a7aa8282c34b4c5f04929859adbbd61df39dd62dd9a9a841f3b8bb888b
+size 15984

rng_state_7.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:823437db304bb4cf1bd59d5bb2e1eca8d99f15af66dd939c8756d68d5a12f879
+size 15984

scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cf41411111ad42f2c42c34fec153aaf1de551d4b135b9cfa6a9021e2b34de0eb
+size 1064

trainer_state.json

@@ -0,0 +1,1761 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.089829781147189,
+  "eval_steps": 400,
+  "global_step": 10800,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.014304105278214848,
+      "grad_norm": 1.6012784676437102,
+      "learning_rate": 1.6666666666666667e-06,
+      "loss": 4.4038,
+      "step": 50
+    },
+    {
+      "epoch": 0.028608210556429696,
+      "grad_norm": 0.81484251583879,
+      "learning_rate": 3.3333333333333333e-06,
+      "loss": 3.6036,
+      "step": 100
+    },
+    {
+      "epoch": 0.04291231583464454,
+      "grad_norm": 3.9762696099154904,
+      "learning_rate": 5e-06,
+      "loss": 3.0207,
+      "step": 150
+    },
+    {
+      "epoch": 0.05721642111285939,
+      "grad_norm": 3.057952660211588,
+      "learning_rate": 6.666666666666667e-06,
+      "loss": 2.4324,
+      "step": 200
+    },
+    {
+      "epoch": 0.07152052639107424,
+      "grad_norm": 2.092719622855296,
+      "learning_rate": 8.333333333333334e-06,
+      "loss": 2.222,
+      "step": 250
+    },
+    {
+      "epoch": 0.08582463166928908,
+      "grad_norm": 6.08825143706115,
+      "learning_rate": 1e-05,
+      "loss": 2.1021,
+      "step": 300
+    },
+    {
+      "epoch": 0.10012873694750393,
+      "grad_norm": 2.493878945601314,
+      "learning_rate": 9.999953760295448e-06,
+      "loss": 1.9831,
+      "step": 350
+    },
+    {
+      "epoch": 0.11443284222571878,
+      "grad_norm": 4.462960292469778,
+      "learning_rate": 9.999815042132062e-06,
+      "loss": 1.917,
+      "step": 400
+    },
+    {
+      "epoch": 0.11443284222571878,
+      "eval_loss": 1.808639645576477,
+      "eval_runtime": 14.2096,
+      "eval_samples_per_second": 70.375,
+      "eval_steps_per_second": 2.252,
+      "step": 400
+    },
+    {
+      "epoch": 0.12873694750393364,
+      "grad_norm": 2.038795534490349,
+      "learning_rate": 9.999583848360633e-06,
+      "loss": 1.8614,
+      "step": 450
+    },
+    {
+      "epoch": 0.1430410527821485,
+      "grad_norm": 2.259377386606669,
+      "learning_rate": 9.999260183732424e-06,
+      "loss": 1.8105,
+      "step": 500
+    },
+    {
+      "epoch": 0.1573451580603633,
+      "grad_norm": 1.6457423711505388,
+      "learning_rate": 9.998844054899058e-06,
+      "loss": 1.7759,
+      "step": 550
+    },
+    {
+      "epoch": 0.17164926333857816,
+      "grad_norm": 2.6198123977173555,
+      "learning_rate": 9.998335470412393e-06,
+      "loss": 1.7508,
+      "step": 600
+    },
+    {
+      "epoch": 0.185953368616793,
+      "grad_norm": 1.6377415784196128,
+      "learning_rate": 9.997734440724333e-06,
+      "loss": 1.7156,
+      "step": 650
+    },
+    {
+      "epoch": 0.20025747389500786,
+      "grad_norm": 3.5293148754159285,
+      "learning_rate": 9.997040978186633e-06,
+      "loss": 1.7015,
+      "step": 700
+    },
+    {
+      "epoch": 0.21456157917322272,
+      "grad_norm": 2.3013282525925263,
+      "learning_rate": 9.996255097050624e-06,
+      "loss": 1.6782,
+      "step": 750
+    },
+    {
+      "epoch": 0.22886568445143757,
+      "grad_norm": 2.428974082500653,
+      "learning_rate": 9.995376813466934e-06,
+      "loss": 1.66,
+      "step": 800
+    },
+    {
+      "epoch": 0.22886568445143757,
+      "eval_loss": 1.5992412567138672,
+      "eval_runtime": 14.0538,
+      "eval_samples_per_second": 71.155,
+      "eval_steps_per_second": 2.277,
+      "step": 800
+    },
+    {
+      "epoch": 0.24316978972965242,
+      "grad_norm": 2.9094373795416506,
+      "learning_rate": 9.994406145485151e-06,
+      "loss": 1.6399,
+      "step": 850
+    },
+    {
+      "epoch": 0.25747389500786727,
+      "grad_norm": 1.5133813561921106,
+      "learning_rate": 9.993343113053454e-06,
+      "loss": 1.626,
+      "step": 900
+    },
+    {
+      "epoch": 0.2717780002860821,
+      "grad_norm": 1.3663105185649191,
+      "learning_rate": 9.992187738018203e-06,
+      "loss": 1.6099,
+      "step": 950
+    },
+    {
+      "epoch": 0.286082105564297,
+      "grad_norm": 1.3144291853877879,
+      "learning_rate": 9.99094004412348e-06,
+      "loss": 1.5968,
+      "step": 1000
+    },
+    {
+      "epoch": 0.3003862108425118,
+      "grad_norm": 1.8770146895064077,
+      "learning_rate": 9.989600057010625e-06,
+      "loss": 1.5754,
+      "step": 1050
+    },
+    {
+      "epoch": 0.3146903161207266,
+      "grad_norm": 1.8478210167954083,
+      "learning_rate": 9.988167804217682e-06,
+      "loss": 1.5711,
+      "step": 1100
+    },
+    {
+      "epoch": 0.3289944213989415,
+      "grad_norm": 1.5949372088951037,
+      "learning_rate": 9.986643315178848e-06,
+      "loss": 1.5557,
+      "step": 1150
+    },
+    {
+      "epoch": 0.3432985266771563,
+      "grad_norm": 1.8431659408457755,
+      "learning_rate": 9.98502662122387e-06,
+      "loss": 1.5572,
+      "step": 1200
+    },
+    {
+      "epoch": 0.3432985266771563,
+      "eval_loss": 1.50032639503479,
+      "eval_runtime": 14.0776,
+      "eval_samples_per_second": 71.035,
+      "eval_steps_per_second": 2.273,
+      "step": 1200
+    },
+    {
+      "epoch": 0.3576026319553712,
+      "grad_norm": 1.3869607567913713,
+      "learning_rate": 9.983317755577392e-06,
+      "loss": 1.5363,
+      "step": 1250
+    },
+    {
+      "epoch": 0.371906737233586,
+      "grad_norm": 1.4514189742887267,
+      "learning_rate": 9.981516753358274e-06,
+      "loss": 1.5358,
+      "step": 1300
+    },
+    {
+      "epoch": 0.3862108425118009,
+      "grad_norm": 1.4333267526235296,
+      "learning_rate": 9.979623651578881e-06,
+      "loss": 1.5141,
+      "step": 1350
+    },
+    {
+      "epoch": 0.40051494779001573,
+      "grad_norm": 0.8580367772458624,
+      "learning_rate": 9.977638489144308e-06,
+      "loss": 1.523,
+      "step": 1400
+    },
+    {
+      "epoch": 0.4148190530682306,
+      "grad_norm": 0.9460440332154582,
+      "learning_rate": 9.975561306851585e-06,
+      "loss": 1.5175,
+      "step": 1450
+    },
+    {
+      "epoch": 0.42912315834644543,
+      "grad_norm": 1.376203229447874,
+      "learning_rate": 9.973392147388847e-06,
+      "loss": 1.5126,
+      "step": 1500
+    },
+    {
+      "epoch": 0.4434272636246603,
+      "grad_norm": 1.5041770784794857,
+      "learning_rate": 9.971131055334445e-06,
+      "loss": 1.4977,
+      "step": 1550
+    },
+    {
+      "epoch": 0.45773136890287514,
+      "grad_norm": 1.095703863839786,
+      "learning_rate": 9.968778077156035e-06,
+      "loss": 1.4877,
+      "step": 1600
+    },
+    {
+      "epoch": 0.45773136890287514,
+      "eval_loss": 1.4638383388519287,
+      "eval_runtime": 14.0468,
+      "eval_samples_per_second": 71.191,
+      "eval_steps_per_second": 2.278,
+      "step": 1600
+    },
+    {
+      "epoch": 0.47203547418109,
+      "grad_norm": 1.1967770971190828,
+      "learning_rate": 9.966333261209625e-06,
+      "loss": 1.4941,
+      "step": 1650
+    },
+    {
+      "epoch": 0.48633957945930484,
+      "grad_norm": 0.7764934991914475,
+      "learning_rate": 9.96379665773858e-06,
+      "loss": 1.4943,
+      "step": 1700
+    },
+    {
+      "epoch": 0.5006436847375196,
+      "grad_norm": 1.1957507140564159,
+      "learning_rate": 9.961168318872583e-06,
+      "loss": 1.4834,
+      "step": 1750
+    },
+    {
+      "epoch": 0.5149477900157345,
+      "grad_norm": 0.891291786132535,
+      "learning_rate": 9.958448298626576e-06,
+      "loss": 1.4766,
+      "step": 1800
+    },
+    {
+      "epoch": 0.5292518952939493,
+      "grad_norm": 0.9430107046686556,
+      "learning_rate": 9.95563665289964e-06,
+      "loss": 1.4659,
+      "step": 1850
+    },
+    {
+      "epoch": 0.5435560005721642,
+      "grad_norm": 1.3583446842191815,
+      "learning_rate": 9.952733439473847e-06,
+      "loss": 1.4681,
+      "step": 1900
+    },
+    {
+      "epoch": 0.557860105850379,
+      "grad_norm": 1.010261006024344,
+      "learning_rate": 9.94973871801308e-06,
+      "loss": 1.4667,
+      "step": 1950
+    },
+    {
+      "epoch": 0.572164211128594,
+      "grad_norm": 0.8494941104833196,
+      "learning_rate": 9.946652550061798e-06,
+      "loss": 1.4453,
+      "step": 2000
+    },
+    {
+      "epoch": 0.572164211128594,
+      "eval_loss": 1.4287511110305786,
+      "eval_runtime": 14.0255,
+      "eval_samples_per_second": 71.299,
+      "eval_steps_per_second": 2.282,
+      "step": 2000
+    },
+    {
+      "epoch": 0.5864683164068087,
+      "grad_norm": 0.7812469708103134,
+      "learning_rate": 9.943474999043775e-06,
+      "loss": 1.4496,
+      "step": 2050
+    },
+    {
+      "epoch": 0.6007724216850236,
+      "grad_norm": 0.7254104161544093,
+      "learning_rate": 9.9402061302608e-06,
+      "loss": 1.4462,
+      "step": 2100
+    },
+    {
+      "epoch": 0.6150765269632384,
+      "grad_norm": 1.1402597738223317,
+      "learning_rate": 9.93684601089133e-06,
+      "loss": 1.4402,
+      "step": 2150
+    },
+    {
+      "epoch": 0.6293806322414532,
+      "grad_norm": 1.0636750138637265,
+      "learning_rate": 9.933394709989109e-06,
+      "loss": 1.4514,
+      "step": 2200
+    },
+    {
+      "epoch": 0.6436847375196681,
+      "grad_norm": 0.6340325583537392,
+      "learning_rate": 9.92985229848175e-06,
+      "loss": 1.4376,
+      "step": 2250
+    },
+    {
+      "epoch": 0.657988842797883,
+      "grad_norm": 1.3226650510062645,
+      "learning_rate": 9.926218849169284e-06,
+      "loss": 1.4404,
+      "step": 2300
+    },
+    {
+      "epoch": 0.6722929480760979,
+      "grad_norm": 0.9023729708460776,
+      "learning_rate": 9.922494436722653e-06,
+      "loss": 1.435,
+      "step": 2350
+    },
+    {
+      "epoch": 0.6865970533543126,
+      "grad_norm": 1.1170660045757717,
+      "learning_rate": 9.91867913768218e-06,
+      "loss": 1.4275,
+      "step": 2400
+    },
+    {
+      "epoch": 0.6865970533543126,
+      "eval_loss": 1.4157905578613281,
+      "eval_runtime": 14.0561,
+      "eval_samples_per_second": 71.143,
+      "eval_steps_per_second": 2.277,
+      "step": 2400
+    },
+    {
+      "epoch": 0.7009011586325276,
+      "grad_norm": 1.164925228192199,
+      "learning_rate": 9.914773030456001e-06,
+      "loss": 1.4238,
+      "step": 2450
+    },
+    {
+      "epoch": 0.7152052639107424,
+      "grad_norm": 0.8519530167823217,
+      "learning_rate": 9.910776195318448e-06,
+      "loss": 1.4347,
+      "step": 2500
+    },
+    {
+      "epoch": 0.7295093691889573,
+      "grad_norm": 0.7139589978182425,
+      "learning_rate": 9.906688714408396e-06,
+      "loss": 1.4306,
+      "step": 2550
+    },
+    {
+      "epoch": 0.743813474467172,
+      "grad_norm": 0.8653282057170465,
+      "learning_rate": 9.902510671727583e-06,
+      "loss": 1.4229,
+      "step": 2600
+    },
+    {
+      "epoch": 0.758117579745387,
+      "grad_norm": 0.8247347491114752,
+      "learning_rate": 9.898242153138882e-06,
+      "loss": 1.4118,
+      "step": 2650
+    },
+    {
+      "epoch": 0.7724216850236018,
+      "grad_norm": 1.0924147996236788,
+      "learning_rate": 9.89388324636453e-06,
+      "loss": 1.4322,
+      "step": 2700
+    },
+    {
+      "epoch": 0.7867257903018167,
+      "grad_norm": 0.842516122122594,
+      "learning_rate": 9.889434040984333e-06,
+      "loss": 1.4101,
+      "step": 2750
+    },
+    {
+      "epoch": 0.8010298955800315,
+      "grad_norm": 0.8063486362804477,
+      "learning_rate": 9.88489462843382e-06,
+      "loss": 1.4191,
+      "step": 2800
+    },
+    {
+      "epoch": 0.8010298955800315,
+      "eval_loss": 1.4116355180740356,
+      "eval_runtime": 13.9779,
+      "eval_samples_per_second": 71.542,
+      "eval_steps_per_second": 2.289,
+      "step": 2800
+    },
+    {
+      "epoch": 0.8153340008582464,
+      "grad_norm": 0.6258848452847008,
+      "learning_rate": 9.880265102002369e-06,
+      "loss": 1.4001,
+      "step": 2850
+    },
+    {
+      "epoch": 0.8296381061364612,
+      "grad_norm": 0.726517642303323,
+      "learning_rate": 9.875545556831283e-06,
+      "loss": 1.4086,
+      "step": 2900
+    },
+    {
+      "epoch": 0.843942211414676,
+      "grad_norm": 0.6713970013254277,
+      "learning_rate": 9.870736089911836e-06,
+      "loss": 1.4073,
+      "step": 2950
+    },
+    {
+      "epoch": 0.8582463166928909,
+      "grad_norm": 0.6148598667666052,
+      "learning_rate": 9.865836800083291e-06,
+      "loss": 1.4093,
+      "step": 3000
+    },
+    {
+      "epoch": 0.8725504219711057,
+      "grad_norm": 0.5359562950631023,
+      "learning_rate": 9.860847788030852e-06,
+      "loss": 1.4017,
+      "step": 3050
+    },
+    {
+      "epoch": 0.8868545272493206,
+      "grad_norm": 0.6194549549607876,
+      "learning_rate": 9.855769156283604e-06,
+      "loss": 1.4196,
+      "step": 3100
+    },
+    {
+      "epoch": 0.9011586325275354,
+      "grad_norm": 0.7870838887793197,
+      "learning_rate": 9.850601009212408e-06,
+      "loss": 1.4039,
+      "step": 3150
+    },
+    {
+      "epoch": 0.9154627378057503,
+      "grad_norm": 0.8348797495331252,
+      "learning_rate": 9.845343453027747e-06,
+      "loss": 1.4092,
+      "step": 3200
+    },
+    {
+      "epoch": 0.9154627378057503,
+      "eval_loss": 1.3961894512176514,
+      "eval_runtime": 14.0237,
+      "eval_samples_per_second": 71.308,
+      "eval_steps_per_second": 2.282,
+      "step": 3200
+    },
+    {
+      "epoch": 0.9297668430839651,
+      "grad_norm": 0.8890086654120082,
+      "learning_rate": 9.839996595777552e-06,
+      "loss": 1.3991,
+      "step": 3250
+    },
+    {
+      "epoch": 0.94407094836218,
+      "grad_norm": 0.8338244522175184,
+      "learning_rate": 9.83456054734498e-06,
+      "loss": 1.3939,
+      "step": 3300
+    },
+    {
+      "epoch": 0.9583750536403948,
+      "grad_norm": 0.667534745389414,
+      "learning_rate": 9.829035419446156e-06,
+      "loss": 1.4052,
+      "step": 3350
+    },
+    {
+      "epoch": 0.9726791589186097,
+      "grad_norm": 0.830996338803645,
+      "learning_rate": 9.823421325627865e-06,
+      "loss": 1.408,
+      "step": 3400
+    },
+    {
+      "epoch": 0.9869832641968245,
+      "grad_norm": 0.752895350030203,
+      "learning_rate": 9.81771838126524e-06,
+      "loss": 1.3927,
+      "step": 3450
+    },
+    {
+      "epoch": 1.0012873694750393,
+      "grad_norm": 0.6022807633216317,
+      "learning_rate": 9.811926703559374e-06,
+      "loss": 1.3947,
+      "step": 3500
+    },
+    {
+      "epoch": 1.0155914747532542,
+      "grad_norm": 0.7757999852306153,
+      "learning_rate": 9.806046411534916e-06,
+      "loss": 1.3613,
+      "step": 3550
+    },
+    {
+      "epoch": 1.029895580031469,
+      "grad_norm": 0.6991186658573486,
+      "learning_rate": 9.800077626037633e-06,
+      "loss": 1.3805,
+      "step": 3600
+    },
+    {
+      "epoch": 1.029895580031469,
+      "eval_loss": 1.386795163154602,
+      "eval_runtime": 13.9668,
+      "eval_samples_per_second": 71.598,
+      "eval_steps_per_second": 2.291,
+      "step": 3600
+    },
+    {
+      "epoch": 1.044199685309684,
+      "grad_norm": 0.6304272914508194,
+      "learning_rate": 9.794020469731915e-06,
+      "loss": 1.3772,
+      "step": 3650
+    },
+    {
+      "epoch": 1.0585037905878987,
+      "grad_norm": 0.6127596406721845,
+      "learning_rate": 9.787875067098257e-06,
+      "loss": 1.3695,
+      "step": 3700
+    },
+    {
+      "epoch": 1.0728078958661136,
+      "grad_norm": 0.5752396229133312,
+      "learning_rate": 9.781641544430703e-06,
+      "loss": 1.3737,
+      "step": 3750
+    },
+    {
+      "epoch": 1.0871120011443285,
+      "grad_norm": 0.8167932197181069,
+      "learning_rate": 9.775320029834255e-06,
+      "loss": 1.3679,
+      "step": 3800
+    },
+    {
+      "epoch": 1.1014161064225432,
+      "grad_norm": 0.7493986062078165,
+      "learning_rate": 9.76891065322223e-06,
+      "loss": 1.3686,
+      "step": 3850
+    },
+    {
+      "epoch": 1.115720211700758,
+      "grad_norm": 0.6896574555563986,
+      "learning_rate": 9.762413546313597e-06,
+      "loss": 1.3688,
+      "step": 3900
+    },
+    {
+      "epoch": 1.130024316978973,
+      "grad_norm": 0.54479225381951,
+      "learning_rate": 9.755828842630269e-06,
+      "loss": 1.3577,
+      "step": 3950
+    },
+    {
+      "epoch": 1.144328422257188,
+      "grad_norm": 0.8631407967474234,
+      "learning_rate": 9.749156677494357e-06,
+      "loss": 1.3791,
+      "step": 4000
+    },
+    {
+      "epoch": 1.144328422257188,
+      "eval_loss": 1.3818904161453247,
+      "eval_runtime": 14.0228,
+      "eval_samples_per_second": 71.312,
+      "eval_steps_per_second": 2.282,
+      "step": 4000
+    },
+    {
+      "epoch": 1.1586325275354026,
+      "grad_norm": 0.6269351505110898,
+      "learning_rate": 9.742397188025394e-06,
+      "loss": 1.3672,
+      "step": 4050
+    },
+    {
+      "epoch": 1.1729366328136175,
+      "grad_norm": 0.5964977170501943,
+      "learning_rate": 9.735550513137513e-06,
+      "loss": 1.3579,
+      "step": 4100
+    },
+    {
+      "epoch": 1.1872407380918324,
+      "grad_norm": 0.6696040499572795,
+      "learning_rate": 9.728616793536588e-06,
+      "loss": 1.3704,
+      "step": 4150
+    },
+    {
+      "epoch": 1.2015448433700473,
+      "grad_norm": 0.7153959218092929,
+      "learning_rate": 9.721596171717352e-06,
+      "loss": 1.3631,
+      "step": 4200
+    },
+    {
+      "epoch": 1.215848948648262,
+      "grad_norm": 0.8228253318299735,
+      "learning_rate": 9.714488791960463e-06,
+      "loss": 1.3643,
+      "step": 4250
+    },
+    {
+      "epoch": 1.230153053926477,
+      "grad_norm": 0.6427955816989828,
+      "learning_rate": 9.707294800329536e-06,
+      "loss": 1.3608,
+      "step": 4300
+    },
+    {
+      "epoch": 1.2444571592046918,
+      "grad_norm": 0.6438118616712295,
+      "learning_rate": 9.700014344668152e-06,
+      "loss": 1.3564,
+      "step": 4350
+    },
+    {
+      "epoch": 1.2587612644829065,
+      "grad_norm": 0.5732058961632965,
+      "learning_rate": 9.692647574596803e-06,
+      "loss": 1.3623,
+      "step": 4400
+    },
+    {
+      "epoch": 1.2587612644829065,
+      "eval_loss": 1.3667371273040771,
+      "eval_runtime": 14.0711,
+      "eval_samples_per_second": 71.068,
+      "eval_steps_per_second": 2.274,
+      "step": 4400
+    },
+    {
+      "epoch": 1.2732084108139037,
+      "grad_norm": 0.5434874117890776,
+      "learning_rate": 9.685194641509837e-06,
+      "loss": 1.3592,
+      "step": 4450
+    },
+    {
+      "epoch": 1.2875125160921184,
+      "grad_norm": 0.8067573948854371,
+      "learning_rate": 9.677655698572326e-06,
+      "loss": 1.3571,
+      "step": 4500
+    },
+    {
+      "epoch": 1.3018166213703333,
+      "grad_norm": 0.6211280738341731,
+      "learning_rate": 9.670030900716941e-06,
+      "loss": 1.3577,
+      "step": 4550
+    },
+    {
+      "epoch": 1.3161207266485482,
+      "grad_norm": 0.47127980997402974,
+      "learning_rate": 9.662320404640743e-06,
+      "loss": 1.3497,
+      "step": 4600
+    },
+    {
+      "epoch": 1.3304248319267629,
+      "grad_norm": 0.6437090365289073,
+      "learning_rate": 9.654524368801982e-06,
+      "loss": 1.3611,
+      "step": 4650
+    },
+    {
+      "epoch": 1.3447289372049778,
+      "grad_norm": 0.4706214878937702,
+      "learning_rate": 9.646642953416835e-06,
+      "loss": 1.3596,
+      "step": 4700
+    },
+    {
+      "epoch": 1.3590330424831927,
+      "grad_norm": 0.4433218616654087,
+      "learning_rate": 9.638676320456109e-06,
+      "loss": 1.3612,
+      "step": 4750
+    },
+    {
+      "epoch": 1.3733371477614076,
+      "grad_norm": 0.6227834199361844,
+      "learning_rate": 9.630624633641918e-06,
+      "loss": 1.3487,
+      "step": 4800
+    },
+    {
+      "epoch": 1.3733371477614076,
+      "eval_loss": 1.3724805116653442,
+      "eval_runtime": 13.958,
+      "eval_samples_per_second": 71.643,
+      "eval_steps_per_second": 2.293,
+      "step": 4800
+    },
+    {
+      "epoch": 1.3876412530396225,
+      "grad_norm": 0.5615209752207829,
+      "learning_rate": 9.622488058444313e-06,
+      "loss": 1.3416,
+      "step": 4850
+    },
+    {
+      "epoch": 1.4019453583178372,
+      "grad_norm": 0.4593448830072353,
+      "learning_rate": 9.614266762077891e-06,
+      "loss": 1.3509,
+      "step": 4900
+    },
+    {
+      "epoch": 1.416249463596052,
+      "grad_norm": 0.5260361200473717,
+      "learning_rate": 9.605960913498342e-06,
+      "loss": 1.3504,
+      "step": 4950
+    },
+    {
+      "epoch": 1.430553568874267,
+      "grad_norm": 0.4949775762320425,
+      "learning_rate": 9.597570683398996e-06,
+      "loss": 1.3608,
+      "step": 5000
+    },
+    {
+      "epoch": 1.4448576741524817,
+      "grad_norm": 0.7134992551375888,
+      "learning_rate": 9.5890962442073e-06,
+      "loss": 1.3456,
+      "step": 5050
+    },
+    {
+      "epoch": 1.4591617794306966,
+      "grad_norm": 0.749997828555375,
+      "learning_rate": 9.580537770081285e-06,
+      "loss": 1.3413,
+      "step": 5100
+    },
+    {
+      "epoch": 1.4734658847089115,
+      "grad_norm": 0.5312330906616294,
+      "learning_rate": 9.57189543690598e-06,
+      "loss": 1.3507,
+      "step": 5150
+    },
+    {
+      "epoch": 1.4877699899871262,
+      "grad_norm": 0.5913338284525619,
+      "learning_rate": 9.563169422289798e-06,
+      "loss": 1.3386,
+      "step": 5200
+    },
+    {
+      "epoch": 1.4877699899871262,
+      "eval_loss": 1.359579086303711,
+      "eval_runtime": 14.046,
+      "eval_samples_per_second": 71.195,
+      "eval_steps_per_second": 2.278,
+      "step": 5200
+    },
+    {
+      "epoch": 1.5020740952653413,
+      "grad_norm": 0.63516444597305,
+      "learning_rate": 9.554359905560887e-06,
+      "loss": 1.3412,
+      "step": 5250
+    },
+    {
+      "epoch": 1.516378200543556,
+      "grad_norm": 0.4411581484928778,
+      "learning_rate": 9.54546706776345e-06,
+      "loss": 1.3505,
+      "step": 5300
+    },
+    {
+      "epoch": 1.530682305821771,
+      "grad_norm": 0.403266190389094,
+      "learning_rate": 9.536491091654018e-06,
+      "loss": 1.3418,
+      "step": 5350
+    },
+    {
+      "epoch": 1.5449864110999858,
+      "grad_norm": 0.4887790997121695,
+      "learning_rate": 9.527432161697696e-06,
+      "loss": 1.352,
+      "step": 5400
+    },
+    {
+      "epoch": 1.5592905163782005,
+      "grad_norm": 0.43803734390526294,
+      "learning_rate": 9.518290464064365e-06,
+      "loss": 1.3374,
+      "step": 5450
+    },
+    {
+      "epoch": 1.5735946216564154,
+      "grad_norm": 0.4477296911829739,
+      "learning_rate": 9.509066186624872e-06,
+      "loss": 1.3362,
+      "step": 5500
+    },
+    {
+      "epoch": 1.5878987269346303,
+      "grad_norm": 0.4849220779673394,
+      "learning_rate": 9.499759518947156e-06,
+      "loss": 1.3463,
+      "step": 5550
+    },
+    {
+      "epoch": 1.602202832212845,
+      "grad_norm": 0.43453154893881496,
+      "learning_rate": 9.490370652292357e-06,
+      "loss": 1.3342,
+      "step": 5600
+    },
+    {
+      "epoch": 1.602202832212845,
+      "eval_loss": 1.3611611127853394,
+      "eval_runtime": 13.9617,
+      "eval_samples_per_second": 71.625,
+      "eval_steps_per_second": 2.292,
+      "step": 5600
+    },
+    {
+      "epoch": 1.61650693749106,
+      "grad_norm": 0.4973975633500145,
+      "learning_rate": 9.480899779610883e-06,
+      "loss": 1.3557,
+      "step": 5650
+    },
+    {
+      "epoch": 1.6308110427692748,
+      "grad_norm": 0.8646218397904073,
+      "learning_rate": 9.471347095538448e-06,
+      "loss": 1.332,
+      "step": 5700
+    },
+    {
+      "epoch": 1.6451151480474895,
+      "grad_norm": 0.4766662524894494,
+      "learning_rate": 9.461712796392067e-06,
+      "loss": 1.3425,
+      "step": 5750
+    },
+    {
+      "epoch": 1.6594192533257046,
+      "grad_norm": 0.43492118267166,
+      "learning_rate": 9.45199708016603e-06,
+      "loss": 1.3366,
+      "step": 5800
+    },
+    {
+      "epoch": 1.6737233586039193,
+      "grad_norm": 0.7281191349195701,
+      "learning_rate": 9.442200146527824e-06,
+      "loss": 1.3405,
+      "step": 5850
+    },
+    {
+      "epoch": 1.6880274638821342,
+      "grad_norm": 0.5059870049803485,
+      "learning_rate": 9.432322196814032e-06,
+      "loss": 1.336,
+      "step": 5900
+    },
+    {
+      "epoch": 1.7023315691603491,
+      "grad_norm": 0.48815713123329457,
+      "learning_rate": 9.422363434026205e-06,
+      "loss": 1.3331,
+      "step": 5950
+    },
+    {
+      "epoch": 1.7166356744385638,
+      "grad_norm": 0.4825656212310282,
+      "learning_rate": 9.41232406282667e-06,
+      "loss": 1.3382,
+      "step": 6000
+    },
+    {
+      "epoch": 1.7166356744385638,
+      "eval_loss": 1.356214165687561,
+      "eval_runtime": 13.9939,
+      "eval_samples_per_second": 71.46,
+      "eval_steps_per_second": 2.287,
+      "step": 6000
+    },
+    {
+      "epoch": 1.7309397797167787,
+      "grad_norm": 0.7522246864779827,
+      "learning_rate": 9.402204289534344e-06,
+      "loss": 1.3239,
+      "step": 6050
+    },
+    {
+      "epoch": 1.7452438849949936,
+      "grad_norm": 0.48984350066891824,
+      "learning_rate": 9.392004322120484e-06,
+      "loss": 1.3237,
+      "step": 6100
+    },
+    {
+      "epoch": 1.7595479902732083,
+      "grad_norm": 0.544930574118496,
+      "learning_rate": 9.381724370204414e-06,
+      "loss": 1.3241,
+      "step": 6150
+    },
+    {
+      "epoch": 1.7738520955514234,
+      "grad_norm": 0.5482222598847393,
+      "learning_rate": 9.371364645049216e-06,
+      "loss": 1.3313,
+      "step": 6200
+    },
+    {
+      "epoch": 1.788156200829638,
+      "grad_norm": 0.46339705172698076,
+      "learning_rate": 9.360925359557397e-06,
+      "loss": 1.3256,
+      "step": 6250
+    },
+    {
+      "epoch": 1.8024603061078528,
+      "grad_norm": 0.5277875338001611,
+      "learning_rate": 9.3504067282665e-06,
+      "loss": 1.3503,
+      "step": 6300
+    },
+    {
+      "epoch": 1.816764411386068,
+      "grad_norm": 0.5539059109504075,
+      "learning_rate": 9.339808967344701e-06,
+      "loss": 1.3368,
+      "step": 6350
+    },
+    {
+      "epoch": 1.8310685166642826,
+      "grad_norm": 0.5119187022621997,
+      "learning_rate": 9.329132294586374e-06,
+      "loss": 1.3257,
+      "step": 6400
+    },
+    {
+      "epoch": 1.8310685166642826,
+      "eval_loss": 1.348954200744629,
+      "eval_runtime": 14.1165,
+      "eval_samples_per_second": 70.839,
+      "eval_steps_per_second": 2.267,
+      "step": 6400
+    },
+    {
+      "epoch": 1.8453726219424975,
+      "grad_norm": 0.4572643729622861,
+      "learning_rate": 9.318376929407606e-06,
+      "loss": 1.3296,
+      "step": 6450
+    },
+    {
+      "epoch": 1.8596767272207124,
+      "grad_norm": 0.41441721606603,
+      "learning_rate": 9.307543092841688e-06,
+      "loss": 1.3306,
+      "step": 6500
+    },
+    {
+      "epoch": 1.873980832498927,
+      "grad_norm": 0.4437842388580668,
+      "learning_rate": 9.296631007534576e-06,
+      "loss": 1.3219,
+      "step": 6550
+    },
+    {
+      "epoch": 1.888284937777142,
+      "grad_norm": 0.668469538481535,
+      "learning_rate": 9.285640897740316e-06,
+      "loss": 1.3201,
+      "step": 6600
+    },
+    {
+      "epoch": 1.902589043055357,
+      "grad_norm": 0.4476992280694945,
+      "learning_rate": 9.27457298931643e-06,
+      "loss": 1.3279,
+      "step": 6650
+    },
+    {
+      "epoch": 1.9168931483335716,
+      "grad_norm": 0.8609307931818154,
+      "learning_rate": 9.263427509719287e-06,
+      "loss": 1.3248,
+      "step": 6700
+    },
+    {
+      "epoch": 1.9311972536117867,
+      "grad_norm": 0.48764755574202223,
+      "learning_rate": 9.252204687999401e-06,
+      "loss": 1.3293,
+      "step": 6750
+    },
+    {
+      "epoch": 1.9455013588900014,
+      "grad_norm": 0.7588730534632143,
+      "learning_rate": 9.240904754796767e-06,
+      "loss": 1.3338,
+      "step": 6800
+    },
+    {
+      "epoch": 1.9455013588900014,
+      "eval_loss": 1.3457790613174438,
+      "eval_runtime": 14.0391,
+      "eval_samples_per_second": 71.229,
+      "eval_steps_per_second": 2.279,
+      "step": 6800
+    },
+    {
+      "epoch": 1.9598054641682163,
+      "grad_norm": 0.47728013357161364,
+      "learning_rate": 9.22952794233608e-06,
+      "loss": 1.328,
+      "step": 6850
+    },
+    {
+      "epoch": 1.9741095694464312,
+      "grad_norm": 0.4865065014657903,
+      "learning_rate": 9.218074484421977e-06,
+      "loss": 1.3329,
+      "step": 6900
+    },
+    {
+      "epoch": 1.988413674724646,
+      "grad_norm": 0.46233352981690246,
+      "learning_rate": 9.206544616434249e-06,
+      "loss": 1.3193,
+      "step": 6950
+    },
+    {
+      "epoch": 2.0027177800028606,
+      "grad_norm": 0.4748345037256569,
+      "learning_rate": 9.194938575322973e-06,
+      "loss": 1.3137,
+      "step": 7000
+    },
+    {
+      "epoch": 2.0170218852810757,
+      "grad_norm": 0.3961349395717629,
+      "learning_rate": 9.183256599603672e-06,
+      "loss": 1.2981,
+      "step": 7050
+    },
+    {
+      "epoch": 2.0313259905592904,
+      "grad_norm": 0.6284979836068443,
+      "learning_rate": 9.171498929352388e-06,
+      "loss": 1.2961,
+      "step": 7100
+    },
+    {
+      "epoch": 2.0456300958375055,
+      "grad_norm": 0.6558610249594138,
+      "learning_rate": 9.159665806200766e-06,
+      "loss": 1.2913,
+      "step": 7150
+    },
+    {
+      "epoch": 2.0599342011157202,
+      "grad_norm": 0.45514976033924853,
+      "learning_rate": 9.147757473331082e-06,
+      "loss": 1.2906,
+      "step": 7200
+    },
+    {
+      "epoch": 2.0599342011157202,
+      "eval_loss": 1.3430439233779907,
+      "eval_runtime": 14.0262,
+      "eval_samples_per_second": 71.295,
+      "eval_steps_per_second": 2.281,
+      "step": 7200
+    },
+    {
+      "epoch": 2.074238306393935,
+      "grad_norm": 0.5426461545993814,
+      "learning_rate": 9.135774175471244e-06,
+      "loss": 1.3004,
+      "step": 7250
+    },
+    {
+      "epoch": 2.08854241167215,
+      "grad_norm": 0.6005516516830625,
+      "learning_rate": 9.123716158889765e-06,
+      "loss": 1.292,
+      "step": 7300
+    },
+    {
+      "epoch": 2.1028465169503647,
+      "grad_norm": 0.9639752009743953,
+      "learning_rate": 9.111583671390697e-06,
+      "loss": 1.2862,
+      "step": 7350
+    },
+    {
+      "epoch": 2.1171506222285794,
+      "grad_norm": 0.4488649957289315,
+      "learning_rate": 9.09937696230855e-06,
+      "loss": 1.3036,
+      "step": 7400
+    },
+    {
+      "epoch": 2.1314547275067945,
+      "grad_norm": 0.7721978784000721,
+      "learning_rate": 9.087096282503152e-06,
+      "loss": 1.2901,
+      "step": 7450
+    },
+    {
+      "epoch": 2.1457588327850092,
+      "grad_norm": 0.4782857255612778,
+      "learning_rate": 9.074741884354507e-06,
+      "loss": 1.2946,
+      "step": 7500
+    },
+    {
+      "epoch": 2.1600629380632244,
+      "grad_norm": 0.43220427000612477,
+      "learning_rate": 9.062314021757603e-06,
+      "loss": 1.2921,
+      "step": 7550
+    },
+    {
+      "epoch": 2.174367043341439,
+      "grad_norm": 0.5795623059587878,
+      "learning_rate": 9.049812950117191e-06,
+      "loss": 1.279,
+      "step": 7600
+    },
+    {
+      "epoch": 2.174367043341439,
+      "eval_loss": 1.3394057750701904,
+      "eval_runtime": 14.0446,
+      "eval_samples_per_second": 71.202,
+      "eval_steps_per_second": 2.278,
+      "step": 7600
+    },
+    {
+      "epoch": 2.1886711486196537,
+      "grad_norm": 0.5713295331254999,
+      "learning_rate": 9.037238926342544e-06,
+      "loss": 1.2909,
+      "step": 7650
+    },
+    {
+      "epoch": 2.202975253897869,
+      "grad_norm": 0.45758770778160607,
+      "learning_rate": 9.02459220884217e-06,
+      "loss": 1.3009,
+      "step": 7700
+    },
+    {
+      "epoch": 2.2172793591760835,
+      "grad_norm": 0.4138476142224768,
+      "learning_rate": 9.011873057518503e-06,
+      "loss": 1.2901,
+      "step": 7750
+    },
+    {
+      "epoch": 2.2315834644542982,
+      "grad_norm": 0.5401623167342202,
+      "learning_rate": 8.999081733762568e-06,
+      "loss": 1.2883,
+      "step": 7800
+    },
+    {
+      "epoch": 2.2458875697325134,
+      "grad_norm": 0.4225832679092138,
+      "learning_rate": 8.986218500448598e-06,
+      "loss": 1.2986,
+      "step": 7850
+    },
+    {
+      "epoch": 2.260191675010728,
+      "grad_norm": 0.578769239923742,
+      "learning_rate": 8.973283621928644e-06,
+      "loss": 1.2932,
+      "step": 7900
+    },
+    {
+      "epoch": 2.2744957802889427,
+      "grad_norm": 0.42471537710995716,
+      "learning_rate": 8.96027736402713e-06,
+      "loss": 1.2911,
+      "step": 7950
+    },
+    {
+      "epoch": 2.288799885567158,
+      "grad_norm": 0.45640421971129197,
+      "learning_rate": 8.947199994035402e-06,
+      "loss": 1.2795,
+      "step": 8000
+    },
+    {
+      "epoch": 2.288799885567158,
+      "eval_loss": 1.3331786394119263,
+      "eval_runtime": 13.9979,
+      "eval_samples_per_second": 71.439,
+      "eval_steps_per_second": 2.286,
+      "step": 8000
+    },
+    {
+      "epoch": 2.3031039908453725,
+      "grad_norm": 0.5262528524865082,
+      "learning_rate": 8.934051780706226e-06,
+      "loss": 1.2847,
+      "step": 8050
+    },
+    {
+      "epoch": 2.3174080961235877,
+      "grad_norm": 0.4308615143171633,
+      "learning_rate": 8.920832994248268e-06,
+      "loss": 1.2942,
+      "step": 8100
+    },
+    {
+      "epoch": 2.3317122014018024,
+      "grad_norm": 0.46124798716185816,
+      "learning_rate": 8.907543906320542e-06,
+      "loss": 1.297,
+      "step": 8150
+    },
+    {
+      "epoch": 2.346016306680017,
+      "grad_norm": 0.4538526984132291,
+      "learning_rate": 8.894184790026823e-06,
+      "loss": 1.2832,
+      "step": 8200
+    },
+    {
+      "epoch": 2.360320411958232,
+      "grad_norm": 0.4645888620271419,
+      "learning_rate": 8.880755919910048e-06,
+      "loss": 1.2891,
+      "step": 8250
+    },
+    {
+      "epoch": 2.374624517236447,
+      "grad_norm": 0.5676282155239492,
+      "learning_rate": 8.867257571946646e-06,
+      "loss": 1.295,
+      "step": 8300
+    },
+    {
+      "epoch": 2.3889286225146615,
+      "grad_norm": 0.429927163826217,
+      "learning_rate": 8.853690023540898e-06,
+      "loss": 1.2917,
+      "step": 8350
+    },
+    {
+      "epoch": 2.4032327277928767,
+      "grad_norm": 0.4224712416764881,
+      "learning_rate": 8.840053553519216e-06,
+      "loss": 1.2793,
+      "step": 8400
+    },
+    {
+      "epoch": 2.4032327277928767,
+      "eval_loss": 1.3279030323028564,
+      "eval_runtime": 14.0803,
+      "eval_samples_per_second": 71.021,
+      "eval_steps_per_second": 2.273,
+      "step": 8400
+    },
+    {
+      "epoch": 2.4175368330710914,
+      "grad_norm": 0.3947030765297477,
+      "learning_rate": 8.82634844212442e-06,
+      "loss": 1.288,
+      "step": 8450
+    },
+    {
+      "epoch": 2.431840938349306,
+      "grad_norm": 0.4497937878369028,
+      "learning_rate": 8.81257497100998e-06,
+      "loss": 1.2949,
+      "step": 8500
+    },
+    {
+      "epoch": 2.446145043627521,
+      "grad_norm": 0.4948619624780139,
+      "learning_rate": 8.79873342323422e-06,
+      "loss": 1.2879,
+      "step": 8550
+    },
+    {
+      "epoch": 2.460449148905736,
+      "grad_norm": 0.8841779211631144,
+      "learning_rate": 8.78482408325451e-06,
+      "loss": 1.2842,
+      "step": 8600
+    },
+    {
+      "epoch": 2.474753254183951,
+      "grad_norm": 0.44783586114307045,
+      "learning_rate": 8.770847236921412e-06,
+      "loss": 1.2868,
+      "step": 8650
+    },
+    {
+      "epoch": 2.4890573594621657,
+      "grad_norm": 0.6387382536339177,
+      "learning_rate": 8.756803171472817e-06,
+      "loss": 1.2821,
+      "step": 8700
+    },
+    {
+      "epoch": 2.5033614647403803,
+      "grad_norm": 0.4704200568795867,
+      "learning_rate": 8.742692175528027e-06,
+      "loss": 1.2854,
+      "step": 8750
+    },
+    {
+      "epoch": 2.5176655700185955,
+      "grad_norm": 0.4776364379876357,
+      "learning_rate": 8.728514539081837e-06,
+      "loss": 1.2814,
+      "step": 8800
+    },
+    {
+      "epoch": 2.5176655700185955,
+      "eval_loss": 1.3429194688796997,
+      "eval_runtime": 13.9117,
+      "eval_samples_per_second": 71.882,
+      "eval_steps_per_second": 2.3,
+      "step": 8800
+    },
+    {
+      "epoch": 2.53196967529681,
+      "grad_norm": 0.7082070517295844,
+      "learning_rate": 8.714270553498567e-06,
+      "loss": 1.2851,
+      "step": 8850
+    },
+    {
+      "epoch": 2.5462737805750253,
+      "grad_norm": 0.4514295526886292,
+      "learning_rate": 8.699960511506077e-06,
+      "loss": 1.2809,
+      "step": 8900
+    },
+    {
+      "epoch": 2.56057788585324,
+      "grad_norm": 0.6853925555348788,
+      "learning_rate": 8.685584707189749e-06,
+      "loss": 1.2961,
+      "step": 8950
+    },
+    {
+      "epoch": 2.5748819911314547,
+      "grad_norm": 0.4538248869842651,
+      "learning_rate": 8.671143435986447e-06,
+      "loss": 1.2893,
+      "step": 9000
+    },
+    {
+      "epoch": 2.5891860964096693,
+      "grad_norm": 0.45631276178983216,
+      "learning_rate": 8.656636994678447e-06,
+      "loss": 1.2921,
+      "step": 9050
+    },
+    {
+      "epoch": 2.6034902016878845,
+      "grad_norm": 0.4181402292311998,
+      "learning_rate": 8.642065681387329e-06,
+      "loss": 1.2849,
+      "step": 9100
+    },
+    {
+      "epoch": 2.617794306966099,
+      "grad_norm": 0.4679963507707488,
+      "learning_rate": 8.627429795567858e-06,
+      "loss": 1.2789,
+      "step": 9150
+    },
+    {
+      "epoch": 2.6320984122443143,
+      "grad_norm": 0.4065327115468989,
+      "learning_rate": 8.61272963800183e-06,
+      "loss": 1.2805,
+      "step": 9200
+    },
+    {
+      "epoch": 2.6320984122443143,
+      "eval_loss": 1.3250114917755127,
+      "eval_runtime": 14.1045,
+      "eval_samples_per_second": 70.899,
+      "eval_steps_per_second": 2.269,
+      "step": 9200
+    },
+    {
+      "epoch": 2.646402517522529,
+      "grad_norm": 0.442868220510357,
+      "learning_rate": 8.597965510791883e-06,
+      "loss": 1.2878,
+      "step": 9250
+    },
+    {
+      "epoch": 2.6607066228007437,
+      "grad_norm": 0.4167482981358102,
+      "learning_rate": 8.5831377173553e-06,
+      "loss": 1.2812,
+      "step": 9300
+    },
+    {
+      "epoch": 2.675010728078959,
+      "grad_norm": 0.4090769340560565,
+      "learning_rate": 8.568246562417762e-06,
+      "loss": 1.2933,
+      "step": 9350
+    },
+    {
+      "epoch": 2.6893148333571735,
+      "grad_norm": 0.42518490969522255,
+      "learning_rate": 8.553292352007096e-06,
+      "loss": 1.2864,
+      "step": 9400
+    },
+    {
+      "epoch": 2.7036189386353886,
+      "grad_norm": 0.4463014716471431,
+      "learning_rate": 8.538275393446976e-06,
+      "loss": 1.2857,
+      "step": 9450
+    },
+    {
+      "epoch": 2.7179230439136033,
+      "grad_norm": 0.45596948523932324,
+      "learning_rate": 8.523195995350613e-06,
+      "loss": 1.2835,
+      "step": 9500
+    },
+    {
+      "epoch": 2.732227149191818,
+      "grad_norm": 0.4205155827535561,
+      "learning_rate": 8.508054467614417e-06,
+      "loss": 1.2849,
+      "step": 9550
+    },
+    {
+      "epoch": 2.7465312544700327,
+      "grad_norm": 0.48430008888282355,
+      "learning_rate": 8.492851121411614e-06,
+      "loss": 1.2789,
+      "step": 9600
+    },
+    {
+      "epoch": 2.7465312544700327,
+      "eval_loss": 1.3283616304397583,
+      "eval_runtime": 14.0066,
+      "eval_samples_per_second": 71.395,
+      "eval_steps_per_second": 2.285,
+      "step": 9600
+    },
+    {
+      "epoch": 2.760835359748248,
+      "grad_norm": 0.5759994995680412,
+      "learning_rate": 8.477586269185868e-06,
+      "loss": 1.2807,
+      "step": 9650
+    },
+    {
+      "epoch": 2.7751394650264625,
+      "grad_norm": 0.4062177321040095,
+      "learning_rate": 8.462260224644848e-06,
+      "loss": 1.2786,
+      "step": 9700
+    },
+    {
+      "epoch": 2.7894435703046776,
+      "grad_norm": 0.40744982615324904,
+      "learning_rate": 8.446873302753783e-06,
+      "loss": 1.288,
+      "step": 9750
+    },
+    {
+      "epoch": 2.8037476755828923,
+      "grad_norm": 0.4351554021842912,
+      "learning_rate": 8.431425819728998e-06,
+      "loss": 1.2809,
+      "step": 9800
+    },
+    {
+      "epoch": 2.818051780861107,
+      "grad_norm": 0.4565206220601423,
+      "learning_rate": 8.415918093031403e-06,
+      "loss": 1.2761,
+      "step": 9850
+    },
+    {
+      "epoch": 2.832355886139322,
+      "grad_norm": 0.4286148896345825,
+      "learning_rate": 8.400350441359976e-06,
+      "loss": 1.2738,
+      "step": 9900
+    },
+    {
+      "epoch": 2.846659991417537,
+      "grad_norm": 0.4091019318117471,
+      "learning_rate": 8.384723184645211e-06,
+      "loss": 1.2756,
+      "step": 9950
+    },
+    {
+      "epoch": 2.860964096695752,
+      "grad_norm": 0.5366072380832926,
+      "learning_rate": 8.369036644042546e-06,
+      "loss": 1.264,
+      "step": 10000
+    },
+    {
+      "epoch": 2.860964096695752,
+      "eval_loss": 1.319417953491211,
+      "eval_runtime": 14.0197,
+      "eval_samples_per_second": 71.328,
+      "eval_steps_per_second": 2.283,
+      "step": 10000
+    },
+    {
+      "epoch": 2.8752682019739666,
+      "grad_norm": 0.39891877892139094,
+      "learning_rate": 8.353291141925763e-06,
+      "loss": 1.2714,
+      "step": 10050
+    },
+    {
+      "epoch": 2.8895723072521813,
+      "grad_norm": 0.43116855479870975,
+      "learning_rate": 8.337487001880353e-06,
+      "loss": 1.276,
+      "step": 10100
+    },
+    {
+      "epoch": 2.903876412530396,
+      "grad_norm": 0.43311934645181527,
+      "learning_rate": 8.32162454869688e-06,
+      "loss": 1.2733,
+      "step": 10150
+    },
+    {
+      "epoch": 2.918180517808611,
+      "grad_norm": 0.4236540903742665,
+      "learning_rate": 8.305704108364301e-06,
+      "loss": 1.2758,
+      "step": 10200
+    },
+    {
+      "epoch": 2.932484623086826,
+      "grad_norm": 0.4815023613318688,
+      "learning_rate": 8.289726008063265e-06,
+      "loss": 1.275,
+      "step": 10250
+    },
+    {
+      "epoch": 2.946788728365041,
+      "grad_norm": 0.43681054268020525,
+      "learning_rate": 8.273690576159383e-06,
+      "loss": 1.2789,
+      "step": 10300
+    },
+    {
+      "epoch": 2.9610928336432556,
+      "grad_norm": 0.4370480894359291,
+      "learning_rate": 8.257598142196496e-06,
+      "loss": 1.267,
+      "step": 10350
+    },
+    {
+      "epoch": 2.9753969389214703,
+      "grad_norm": 0.4461842695375769,
+      "learning_rate": 8.241449036889892e-06,
+      "loss": 1.2734,
+      "step": 10400
+    },
+    {
+      "epoch": 2.9753969389214703,
+      "eval_loss": 1.3316634893417358,
+      "eval_runtime": 13.9113,
+      "eval_samples_per_second": 71.884,
+      "eval_steps_per_second": 2.3,
+      "step": 10400
+    },
+    {
+      "epoch": 2.9897010441996854,
+      "grad_norm": 0.44034804073477984,
+      "learning_rate": 8.225243592119501e-06,
+      "loss": 1.2736,
+      "step": 10450
+    },
+    {
+      "epoch": 3.0040051494779,
+      "grad_norm": 0.4720256474307512,
+      "learning_rate": 8.208982140923095e-06,
+      "loss": 1.2694,
+      "step": 10500
+    },
+    {
+      "epoch": 3.0183092547561152,
+      "grad_norm": 0.6347562232882346,
+      "learning_rate": 8.192665017489431e-06,
+      "loss": 1.2336,
+      "step": 10550
+    },
+    {
+      "epoch": 3.03261336003433,
+      "grad_norm": 0.37981139577002,
+      "learning_rate": 8.17629255715138e-06,
+      "loss": 1.2494,
+      "step": 10600
+    },
+    {
+      "epoch": 3.0469174653125446,
+      "grad_norm": 0.7105885207992017,
+      "learning_rate": 8.159865096379046e-06,
+      "loss": 1.2397,
+      "step": 10650
+    },
+    {
+      "epoch": 3.0612215705907597,
+      "grad_norm": 0.43006752774126733,
+      "learning_rate": 8.14338297277284e-06,
+      "loss": 1.2384,
+      "step": 10700
+    },
+    {
+      "epoch": 3.0755256758689744,
+      "grad_norm": 0.4261194480956777,
+      "learning_rate": 8.126846525056555e-06,
+      "loss": 1.2436,
+      "step": 10750
+    },
+    {
+      "epoch": 3.089829781147189,
+      "grad_norm": 0.45249834468920586,
+      "learning_rate": 8.110256093070393e-06,
+      "loss": 1.252,
+      "step": 10800
+    },
+    {
+      "epoch": 3.089829781147189,
+      "eval_loss": 1.324701189994812,
+      "eval_runtime": 14.0291,
+      "eval_samples_per_second": 71.28,
+      "eval_steps_per_second": 2.281,
+      "step": 10800
+    }
+  ],
+  "logging_steps": 50,
+  "max_steps": 34950,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 400,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.0328768862224384e+16,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}

training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d70f6b7223186f679f09a0f4404fc3f434bfeac6c1e1629dfecf969286e228b1
+size 7224

zero_to_fp32.py

@@ -0,0 +1,604 @@
+#!/usr/bin/env python
+
+# Copyright (c) Microsoft Corporation.
+# SPDX-License-Identifier: Apache-2.0
+
+# DeepSpeed Team
+
+# This script extracts fp32 consolidated weights from a zero 1, 2 and 3 DeepSpeed checkpoints. It gets
+# copied into the top level checkpoint dir, so the user can easily do the conversion at any point in
+# the future. Once extracted, the weights don't require DeepSpeed and can be used in any
+# application.
+#
+# example: python zero_to_fp32.py . pytorch_model.bin
+
+import argparse
+import torch
+import glob
+import math
+import os
+import re
+from collections import OrderedDict
+from dataclasses import dataclass
+
+# while this script doesn't use deepspeed to recover data, since the checkpoints are pickled with
+# DeepSpeed data structures it has to be available in the current python environment.
+from deepspeed.utils import logger
+from deepspeed.checkpoint.constants import (DS_VERSION, OPTIMIZER_STATE_DICT, SINGLE_PARTITION_OF_FP32_GROUPS,
+                                            FP32_FLAT_GROUPS, ZERO_STAGE, PARTITION_COUNT, PARAM_SHAPES, BUFFER_NAMES,
+                                            FROZEN_PARAM_SHAPES, FROZEN_PARAM_FRAGMENTS)
+
+
+@dataclass
+class zero_model_state:
+    buffers: dict()
+    param_shapes: dict()
+    shared_params: list
+    ds_version: int
+    frozen_param_shapes: dict()
+    frozen_param_fragments: dict()
+
+
+debug = 0
+
+# load to cpu
+device = torch.device('cpu')
+
+
+def atoi(text):
+    return int(text) if text.isdigit() else text
+
+
+def natural_keys(text):
+    '''
+    alist.sort(key=natural_keys) sorts in human order
+    http://nedbatchelder.com/blog/200712/human_sorting.html
+    (See Toothy's implementation in the comments)
+    '''
+    return [atoi(c) for c in re.split(r'(\d+)', text)]
+
+
+def get_model_state_file(checkpoint_dir, zero_stage):
+    if not os.path.isdir(checkpoint_dir):
+        raise FileNotFoundError(f"Directory '{checkpoint_dir}' doesn't exist")
+
+    # there should be only one file
+    if zero_stage <= 2:
+        file = os.path.join(checkpoint_dir, "mp_rank_00_model_states.pt")
+    elif zero_stage == 3:
+        file = os.path.join(checkpoint_dir, "zero_pp_rank_0_mp_rank_00_model_states.pt")
+
+    if not os.path.exists(file):
+        raise FileNotFoundError(f"can't find model states file at '{file}'")
+
+    return file
+
+
+def get_checkpoint_files(checkpoint_dir, glob_pattern):
+    # XXX: need to test that this simple glob rule works for multi-node setup too
+    ckpt_files = sorted(glob.glob(os.path.join(checkpoint_dir, glob_pattern)), key=natural_keys)
+
+    if len(ckpt_files) == 0:
+        raise FileNotFoundError(f"can't find {glob_pattern} files in directory '{checkpoint_dir}'")
+
+    return ckpt_files
+
+
+def get_optim_files(checkpoint_dir):
+    return get_checkpoint_files(checkpoint_dir, "*_optim_states.pt")
+
+
+def get_model_state_files(checkpoint_dir):
+    return get_checkpoint_files(checkpoint_dir, "*_model_states.pt")
+
+
+def parse_model_states(files):
+    zero_model_states = []
+    for file in files:
+        state_dict = torch.load(file, map_location=device)
+
+        if BUFFER_NAMES not in state_dict:
+            raise ValueError(f"{file} is not a model state checkpoint")
+        buffer_names = state_dict[BUFFER_NAMES]
+        if debug:
+            print("Found buffers:", buffer_names)
+
+        # recover just the buffers while restoring them to fp32 if they were saved in fp16
+        buffers = {k: v.float() for k, v in state_dict["module"].items() if k in buffer_names}
+        param_shapes = state_dict[PARAM_SHAPES]
+
+        # collect parameters that are included in param_shapes
+        param_names = []
+        for s in param_shapes:
+            for name in s.keys():
+                param_names.append(name)
+
+        # update with frozen parameters
+        frozen_param_shapes = state_dict.get(FROZEN_PARAM_SHAPES, None)
+        if frozen_param_shapes is not None:
+            if debug:
+                print(f"Found frozen_param_shapes: {frozen_param_shapes}")
+            param_names += list(frozen_param_shapes.keys())
+
+        # handle shared params
+        shared_params = [[k, v] for k, v in state_dict["shared_params"].items()]
+
+        ds_version = state_dict.get(DS_VERSION, None)
+
+        frozen_param_fragments = state_dict.get(FROZEN_PARAM_FRAGMENTS, None)
+
+        z_model_state = zero_model_state(buffers=buffers,
+                                         param_shapes=param_shapes,
+                                         shared_params=shared_params,
+                                         ds_version=ds_version,
+                                         frozen_param_shapes=frozen_param_shapes,
+                                         frozen_param_fragments=frozen_param_fragments)
+        zero_model_states.append(z_model_state)
+
+    return zero_model_states
+
+
+def parse_optim_states(files, ds_checkpoint_dir):
+
+    total_files = len(files)
+    state_dicts = []
+    for f in files:
+        state_dict = torch.load(f, map_location=device)
+        # immediately discard the potentially huge 2 optimizer states as we only care for fp32 master weights
+        # and also handle the case where it was already removed by another helper script
+        state_dict["optimizer_state_dict"].pop("optimizer_state_dict", None)
+        state_dicts.append(state_dict)
+
+    if not ZERO_STAGE in state_dicts[0][OPTIMIZER_STATE_DICT]:
+        raise ValueError(f"{files[0]} is not a zero checkpoint")
+    zero_stage = state_dicts[0][OPTIMIZER_STATE_DICT][ZERO_STAGE]
+    world_size = state_dicts[0][OPTIMIZER_STATE_DICT][PARTITION_COUNT]
+
+    # For ZeRO-2 each param group can have different partition_count as data parallelism for expert
+    # parameters can be different from data parallelism for non-expert parameters. So we can just
+    # use the max of the partition_count to get the dp world_size.
+
+    if type(world_size) is list:
+        world_size = max(world_size)
+
+    if world_size != total_files:
+        raise ValueError(
+            f"Expected {world_size} of '*_optim_states.pt' under '{ds_checkpoint_dir}' but found {total_files} files. "
+            "Possibly due to an overwrite of an old checkpoint, or a checkpoint didn't get saved by one or more processes."
+        )
+
+    # the groups are named differently in each stage
+    if zero_stage <= 2:
+        fp32_groups_key = SINGLE_PARTITION_OF_FP32_GROUPS
+    elif zero_stage == 3:
+        fp32_groups_key = FP32_FLAT_GROUPS
+    else:
+        raise ValueError(f"unknown zero stage {zero_stage}")
+
+    if zero_stage <= 2:
+        fp32_flat_groups = [state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key] for i in range(len(state_dicts))]
+    elif zero_stage == 3:
+        # if there is more than one param group, there will be multiple flattened tensors - one
+        # flattened tensor per group - for simplicity merge them into a single tensor
+        #
+        # XXX: could make the script more memory efficient for when there are multiple groups - it
+        # will require matching the sub-lists of param_shapes for each param group flattened tensor
+
+        fp32_flat_groups = [
+            torch.cat(state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key], 0) for i in range(len(state_dicts))
+        ]
+
+    return zero_stage, world_size, fp32_flat_groups
+
+
+def _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters):
+    """
+    Returns fp32 state_dict reconstructed from ds checkpoint
+
+    Args:
+        - ``ds_checkpoint_dir``: path to the deepspeed checkpoint folder (where the optimizer files are)
+
+    """
+    print(f"Processing zero checkpoint '{ds_checkpoint_dir}'")
+
+    optim_files = get_optim_files(ds_checkpoint_dir)
+    zero_stage, world_size, fp32_flat_groups = parse_optim_states(optim_files, ds_checkpoint_dir)
+    print(f"Detected checkpoint of type zero stage {zero_stage}, world_size: {world_size}")
+
+    model_files = get_model_state_files(ds_checkpoint_dir)
+
+    zero_model_states = parse_model_states(model_files)
+    print(f'Parsing checkpoint created by deepspeed=={zero_model_states[0].ds_version}')
+
+    if zero_stage <= 2:
+        return _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                                          exclude_frozen_parameters)
+    elif zero_stage == 3:
+        return _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                                          exclude_frozen_parameters)
+
+
+def _zero2_merge_frozen_params(state_dict, zero_model_states):
+    if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
+        return
+
+    frozen_param_shapes = zero_model_states[0].frozen_param_shapes
+    frozen_param_fragments = zero_model_states[0].frozen_param_fragments
+
+    if debug:
+        num_elem = sum(s.numel() for s in frozen_param_shapes.values())
+        print(f'rank 0: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
+
+        wanted_params = len(frozen_param_shapes)
+        wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
+        avail_numel = sum([p.numel() for p in frozen_param_fragments.values()])
+        print(f'Frozen params: Have {avail_numel} numels to process.')
+        print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
+
+    total_params = 0
+    total_numel = 0
+    for name, shape in frozen_param_shapes.items():
+        total_params += 1
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+
+        state_dict[name] = frozen_param_fragments[name]
+
+        if debug:
+            print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
+
+    print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _has_callable(obj, fn):
+    attr = getattr(obj, fn, None)
+    return callable(attr)
+
+
+def _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
+    param_shapes = zero_model_states[0].param_shapes
+
+    # Reconstruction protocol:
+    #
+    # XXX: document this
+
+    if debug:
+        for i in range(world_size):
+            for j in range(len(fp32_flat_groups[0])):
+                print(f"{FP32_FLAT_GROUPS}[{i}][{j}].shape={fp32_flat_groups[i][j].shape}")
+
+    # XXX: memory usage doubles here (zero2)
+    num_param_groups = len(fp32_flat_groups[0])
+    merged_single_partition_of_fp32_groups = []
+    for i in range(num_param_groups):
+        merged_partitions = [sd[i] for sd in fp32_flat_groups]
+        full_single_fp32_vector = torch.cat(merged_partitions, 0)
+        merged_single_partition_of_fp32_groups.append(full_single_fp32_vector)
+    avail_numel = sum(
+        [full_single_fp32_vector.numel() for full_single_fp32_vector in merged_single_partition_of_fp32_groups])
+
+    if debug:
+        wanted_params = sum([len(shapes) for shapes in param_shapes])
+        wanted_numel = sum([sum(shape.numel() for shape in shapes.values()) for shapes in param_shapes])
+        # not asserting if there is a mismatch due to possible padding
+        print(f"Have {avail_numel} numels to process.")
+        print(f"Need {wanted_numel} numels in {wanted_params} params.")
+
+    # params
+    # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
+    # out-of-core computing solution
+    total_numel = 0
+    total_params = 0
+    for shapes, full_single_fp32_vector in zip(param_shapes, merged_single_partition_of_fp32_groups):
+        offset = 0
+        avail_numel = full_single_fp32_vector.numel()
+        for name, shape in shapes.items():
+
+            unpartitioned_numel = shape.numel() if _has_callable(shape, 'numel') else math.prod(shape)
+            total_numel += unpartitioned_numel
+            total_params += 1
+
+            if debug:
+                print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
+            state_dict[name] = full_single_fp32_vector.narrow(0, offset, unpartitioned_numel).view(shape)
+            offset += unpartitioned_numel
+
+        # Z2 started to align to 2*world_size to improve nccl performance. Therefore both offset and
+        # avail_numel can differ by anywhere between 0..2*world_size. Due to two unrelated complex
+        # paddings performed in the code it's almost impossible to predict the exact numbers w/o the
+        # live optimizer object, so we are checking that the numbers are within the right range
+        align_to = 2 * world_size
+
+        def zero2_align(x):
+            return align_to * math.ceil(x / align_to)
+
+        if debug:
+            print(f"original offset={offset}, avail_numel={avail_numel}")
+
+        offset = zero2_align(offset)
+        avail_numel = zero2_align(avail_numel)
+
+        if debug:
+            print(f"aligned  offset={offset}, avail_numel={avail_numel}")
+
+        # Sanity check
+        if offset != avail_numel:
+            raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
+
+    print(f"Reconstructed fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                               exclude_frozen_parameters):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    if not exclude_frozen_parameters:
+        _zero2_merge_frozen_params(state_dict, zero_model_states)
+
+    _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
+
+    # recover shared parameters
+    for pair in zero_model_states[0].shared_params:
+        if pair[1] in state_dict:
+            state_dict[pair[0]] = state_dict[pair[1]]
+
+    return state_dict
+
+
+def zero3_partitioned_param_info(unpartitioned_numel, world_size):
+    remainder = unpartitioned_numel % world_size
+    padding_numel = (world_size - remainder) if remainder else 0
+    partitioned_numel = math.ceil(unpartitioned_numel / world_size)
+    return partitioned_numel, padding_numel
+
+
+def _zero3_merge_frozen_params(state_dict, world_size, zero_model_states):
+    if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
+        return
+
+    if debug:
+        for i in range(world_size):
+            num_elem = sum(s.numel() for s in zero_model_states[i].frozen_param_fragments.values())
+            print(f'rank {i}: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
+
+        frozen_param_shapes = zero_model_states[0].frozen_param_shapes
+        wanted_params = len(frozen_param_shapes)
+        wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
+        avail_numel = sum([p.numel() for p in zero_model_states[0].frozen_param_fragments.values()]) * world_size
+        print(f'Frozen params: Have {avail_numel} numels to process.')
+        print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
+
+    total_params = 0
+    total_numel = 0
+    for name, shape in zero_model_states[0].frozen_param_shapes.items():
+        total_params += 1
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+
+        param_frags = tuple(model_state.frozen_param_fragments[name] for model_state in zero_model_states)
+        state_dict[name] = torch.cat(param_frags, 0).narrow(0, 0, unpartitioned_numel).view(shape)
+
+        partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
+
+        if debug:
+            print(
+                f"Frozen params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
+            )
+
+    print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
+    param_shapes = zero_model_states[0].param_shapes
+    avail_numel = fp32_flat_groups[0].numel() * world_size
+    # Reconstruction protocol: For zero3 we need to zip the partitions together at boundary of each
+    # param, re-consolidating each param, while dealing with padding if any
+
+    # merge list of dicts, preserving order
+    param_shapes = {k: v for d in param_shapes for k, v in d.items()}
+
+    if debug:
+        for i in range(world_size):
+            print(f"{FP32_FLAT_GROUPS}[{i}].shape={fp32_flat_groups[i].shape}")
+
+        wanted_params = len(param_shapes)
+        wanted_numel = sum(shape.numel() for shape in param_shapes.values())
+        # not asserting if there is a mismatch due to possible padding
+        avail_numel = fp32_flat_groups[0].numel() * world_size
+        print(f"Trainable params: Have {avail_numel} numels to process.")
+        print(f"Trainable params: Need {wanted_numel} numels in {wanted_params} params.")
+
+    # params
+    # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
+    # out-of-core computing solution
+    offset = 0
+    total_numel = 0
+    total_params = 0
+    for name, shape in param_shapes.items():
+
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+        total_params += 1
+
+        partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
+
+        if debug:
+            print(
+                f"Trainable params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
+            )
+
+        # XXX: memory usage doubles here
+        state_dict[name] = torch.cat(
+            tuple(fp32_flat_groups[i].narrow(0, offset, partitioned_numel) for i in range(world_size)),
+            0).narrow(0, 0, unpartitioned_numel).view(shape)
+        offset += partitioned_numel
+
+    offset *= world_size
+
+    # Sanity check
+    if offset != avail_numel:
+        raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
+
+    print(f"Reconstructed Trainable fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                               exclude_frozen_parameters):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    if not exclude_frozen_parameters:
+        _zero3_merge_frozen_params(state_dict, world_size, zero_model_states)
+
+    _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
+
+    # recover shared parameters
+    for pair in zero_model_states[0].shared_params:
+        if pair[1] in state_dict:
+            state_dict[pair[0]] = state_dict[pair[1]]
+
+    return state_dict
+
+
+def get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag=None, exclude_frozen_parameters=False):
+    """
+    Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated state_dict that can be loaded with
+    ``load_state_dict()`` and used for training without DeepSpeed or shared with others, for example
+    via a model hub.
+
+    Args:
+        - ``checkpoint_dir``: path to the desired checkpoint folder
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in 'latest' file. e.g., ``global_step14``
+        - ``exclude_frozen_parameters``: exclude frozen parameters
+
+    Returns:
+        - pytorch ``state_dict``
+
+    Note: this approach may not work if your application doesn't have sufficient free CPU memory and
+    you may need to use the offline approach using the ``zero_to_fp32.py`` script that is saved with
+    the checkpoint.
+
+    A typical usage might be ::
+
+        from deepspeed.utils.zero_to_fp32 import get_fp32_state_dict_from_zero_checkpoint
+        # do the training and checkpoint saving
+        state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir) # already on cpu
+        model = model.cpu() # move to cpu
+        model.load_state_dict(state_dict)
+        # submit to model hub or save the model to share with others
+
+    In this example the ``model`` will no longer be usable in the deepspeed context of the same
+    application. i.e. you will need to re-initialize the deepspeed engine, since
+    ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
+
+    If you want it all done for you, use ``load_state_dict_from_zero_checkpoint`` instead.
+
+    """
+    if tag is None:
+        latest_path = os.path.join(checkpoint_dir, 'latest')
+        if os.path.isfile(latest_path):
+            with open(latest_path, 'r') as fd:
+                tag = fd.read().strip()
+        else:
+            raise ValueError(f"Unable to find 'latest' file at {latest_path}")
+
+    ds_checkpoint_dir = os.path.join(checkpoint_dir, tag)
+
+    if not os.path.isdir(ds_checkpoint_dir):
+        raise FileNotFoundError(f"Directory '{ds_checkpoint_dir}' doesn't exist")
+
+    return _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters)
+
+
+def convert_zero_checkpoint_to_fp32_state_dict(checkpoint_dir, output_file, tag=None, exclude_frozen_parameters=False):
+    """
+    Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict`` file that can be
+    loaded with ``torch.load(file)`` + ``load_state_dict()`` and used for training without DeepSpeed.
+
+    Args:
+        - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
+        - ``output_file``: path to the pytorch fp32 state_dict output file (e.g. path/pytorch_model.bin)
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
+        - ``exclude_frozen_parameters``: exclude frozen parameters
+    """
+
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag, exclude_frozen_parameters)
+    print(f"Saving fp32 state dict to {output_file}")
+    torch.save(state_dict, output_file)
+
+
+def load_state_dict_from_zero_checkpoint(model, checkpoint_dir, tag=None):
+    """
+    1. Put the provided model to cpu
+    2. Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict``
+    3. Load it into the provided model
+
+    Args:
+        - ``model``: the model object to update
+        - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
+
+    Returns:
+        - ``model`: modified model
+
+    Make sure you have plenty of CPU memory available before you call this function. If you don't
+    have enough use the ``zero_to_fp32.py`` utility to do the conversion. You will find it
+    conveniently placed for you in the checkpoint folder.
+
+    A typical usage might be ::
+
+        from deepspeed.utils.zero_to_fp32 import load_state_dict_from_zero_checkpoint
+        model = load_state_dict_from_zero_checkpoint(trainer.model, checkpoint_dir)
+        # submit to model hub or save the model to share with others
+
+    Note, that once this was run, the ``model`` will no longer be usable in the deepspeed context
+    of the same application. i.e. you will need to re-initialize the deepspeed engine, since
+    ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
+
+    """
+    logger.info(f"Extracting fp32 weights")
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag)
+
+    logger.info(f"Overwriting model with fp32 weights")
+    model = model.cpu()
+    model.load_state_dict(state_dict, strict=False)
+
+    return model
+
+
+if __name__ == "__main__":
+
+    parser = argparse.ArgumentParser()
+    parser.add_argument("checkpoint_dir",
+                        type=str,
+                        help="path to the desired checkpoint folder, e.g., path/checkpoint-12")
+    parser.add_argument(
+        "output_file",
+        type=str,
+        help="path to the pytorch fp32 state_dict output file (e.g. path/checkpoint-12/pytorch_model.bin)")
+    parser.add_argument("-t",
+                        "--tag",
+                        type=str,
+                        default=None,
+                        help="checkpoint tag used as a unique identifier for checkpoint. e.g., global_step1")
+    parser.add_argument("--exclude_frozen_parameters", action='store_true', help="exclude frozen parameters")
+    parser.add_argument("-d", "--debug", action='store_true', help="enable debug")
+    args = parser.parse_args()
+
+    debug = args.debug
+
+    convert_zero_checkpoint_to_fp32_state_dict(args.checkpoint_dir,
+                                               args.output_file,
+                                               tag=args.tag,
+                                               exclude_frozen_parameters=args.exclude_frozen_parameters)