Training in progress, step 400, checkpoint

.gitattributes

@@ -37,3 +37,4 @@ tokenizer.json filter=lfs diff=lfs merge=lfs -text
 checkpoint-100/tokenizer.json filter=lfs diff=lfs merge=lfs -text
 checkpoint-200/tokenizer.json filter=lfs diff=lfs merge=lfs -text
 checkpoint-300/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+checkpoint-400/tokenizer.json filter=lfs diff=lfs merge=lfs -text

checkpoint-400/added_tokens.json

@@ -0,0 +1,24 @@
+{
+  "</tool_call>": 151658,
+  "<tool_call>": 151657,
+  "<|box_end|>": 151649,
+  "<|box_start|>": 151648,
+  "<|endoftext|>": 151643,
+  "<|file_sep|>": 151664,
+  "<|fim_middle|>": 151660,
+  "<|fim_pad|>": 151662,
+  "<|fim_prefix|>": 151659,
+  "<|fim_suffix|>": 151661,
+  "<|im_end|>": 151645,
+  "<|im_start|>": 151644,
+  "<|image_pad|>": 151655,
+  "<|object_ref_end|>": 151647,
+  "<|object_ref_start|>": 151646,
+  "<|quad_end|>": 151651,
+  "<|quad_start|>": 151650,
+  "<|repo_name|>": 151663,
+  "<|video_pad|>": 151656,
+  "<|vision_end|>": 151653,
+  "<|vision_pad|>": 151654,
+  "<|vision_start|>": 151652
+}

checkpoint-400/config.json

@@ -0,0 +1,29 @@
+{
+  "_name_or_path": "Qwen/Qwen2.5-1.5B-Instruct",
+  "architectures": [
+    "Qwen2ForCausalLM"
+  ],
+  "attention_dropout": 0.0,
+  "bos_token_id": 151643,
+  "eos_token_id": 151645,
+  "hidden_act": "silu",
+  "hidden_size": 1536,
+  "initializer_range": 0.02,
+  "intermediate_size": 8960,
+  "max_position_embeddings": 32768,
+  "max_window_layers": 21,
+  "model_type": "qwen2",
+  "num_attention_heads": 12,
+  "num_hidden_layers": 28,
+  "num_key_value_heads": 2,
+  "rms_norm_eps": 1e-06,
+  "rope_scaling": null,
+  "rope_theta": 1000000.0,
+  "sliding_window": null,
+  "tie_word_embeddings": true,
+  "torch_dtype": "bfloat16",
+  "transformers_version": "4.46.1",
+  "use_cache": false,
+  "use_sliding_window": false,
+  "vocab_size": 151936
+}

checkpoint-400/generation_config.json

@@ -0,0 +1,14 @@
+{
+  "bos_token_id": 151643,
+  "do_sample": true,
+  "eos_token_id": [
+    151645,
+    151643
+  ],
+  "pad_token_id": 151643,
+  "repetition_penalty": 1.1,
+  "temperature": 0.7,
+  "top_k": 20,
+  "top_p": 0.8,
+  "transformers_version": "4.46.1"
+}

checkpoint-400/global_step400/bf16_zero_pp_rank_0_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6208a8e142d6685dca5a23b45a2227c2ae5cd3c997f0a63bea69f01426ad61f8
+size 9262302412

checkpoint-400/global_step400/bf16_zero_pp_rank_1_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8b7c380ee0b8e55dfcce06f472e461e530340d5e55e5cde5924447128eb30fa0
+size 9262305996

checkpoint-400/global_step400/mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a03e8f694188d20fdae77caa5c0f31839a595fadf4a1849dc046622d108b93ed
+size 3087520696

checkpoint-400/latest

@@ -0,0 +1 @@
+global_step400

checkpoint-400/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6f21907e035258574759a6aa0803b047f7c8a267a5e422e13113a885f04fb7e6
+size 3554214752

checkpoint-400/rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c8d6a959372d5e0c2ea025dd26c9d0ad2046fce19352056cae8074dcbd0a6fd4
+size 14512

checkpoint-400/rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0f68a37892a1b445d21bb35cc10bf7a058a6f9ec8c363f5ed156ff4f49d90fb6
+size 14512

checkpoint-400/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e14d6c259c1bb094d5270917de24d1375ae6c91713e3b553409a7fee3242a4bb
+size 1064

checkpoint-400/special_tokens_map.json

@@ -0,0 +1,31 @@
+{
+  "additional_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "eos_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

checkpoint-400/tokenizer.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9c5ae00e602b8860cbd784ba82a8aa14e8feecec692e7076590d014d7b7fdafa
+size 11421896

checkpoint-400/tokenizer_config.json

@@ -0,0 +1,208 @@
+{
+  "add_bos_token": false,
+  "add_prefix_space": false,
+  "added_tokens_decoder": {
+    "151643": {
+      "content": "<|endoftext|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151644": {
+      "content": "<|im_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151645": {
+      "content": "<|im_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151646": {
+      "content": "<|object_ref_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151647": {
+      "content": "<|object_ref_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151648": {
+      "content": "<|box_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151649": {
+      "content": "<|box_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151650": {
+      "content": "<|quad_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151651": {
+      "content": "<|quad_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151652": {
+      "content": "<|vision_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151653": {
+      "content": "<|vision_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151654": {
+      "content": "<|vision_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151655": {
+      "content": "<|image_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151656": {
+      "content": "<|video_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151657": {
+      "content": "<tool_call>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151658": {
+      "content": "</tool_call>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151659": {
+      "content": "<|fim_prefix|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151660": {
+      "content": "<|fim_middle|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151661": {
+      "content": "<|fim_suffix|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151662": {
+      "content": "<|fim_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151663": {
+      "content": "<|repo_name|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151664": {
+      "content": "<|file_sep|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    }
+  },
+  "additional_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "bos_token": null,
+  "chat_template": "{%- if tools %}\n    {{- '<|im_start|>system\\n' }}\n    {%- if messages[0]['role'] == 'system' %}\n        {{- messages[0]['content'] }}\n    {%- else %}\n        {{- 'You are Qwen, created by Alibaba Cloud. You are a helpful assistant.' }}\n    {%- endif %}\n    {{- \"\\n\\n# Tools\\n\\nYou may call one or more functions to assist with the user query.\\n\\nYou are provided with function signatures within <tools></tools> XML tags:\\n<tools>\" }}\n    {%- for tool in tools %}\n        {{- \"\\n\" }}\n        {{- tool | tojson }}\n    {%- endfor %}\n    {{- \"\\n</tools>\\n\\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\\n<tool_call>\\n{\\\"name\\\": <function-name>, \\\"arguments\\\": <args-json-object>}\\n</tool_call><|im_end|>\\n\" }}\n{%- else %}\n    {%- if messages[0]['role'] == 'system' %}\n        {{- '<|im_start|>system\\n' + messages[0]['content'] + '<|im_end|>\\n' }}\n    {%- else %}\n        {{- '<|im_start|>system\\nYou are Qwen, created by Alibaba Cloud. You are a helpful assistant.<|im_end|>\\n' }}\n    {%- endif %}\n{%- endif %}\n{%- for message in messages %}\n    {%- if (message.role == \"user\") or (message.role == \"system\" and not loop.first) or (message.role == \"assistant\" and not message.tool_calls) %}\n        {{- '<|im_start|>' + message.role + '\\n' + message.content + '<|im_end|>' + '\\n' }}\n    {%- elif message.role == \"assistant\" %}\n        {{- '<|im_start|>' + message.role }}\n        {%- if message.content %}\n            {{- '\\n' + message.content }}\n        {%- endif %}\n        {%- for tool_call in message.tool_calls %}\n            {%- if tool_call.function is defined %}\n                {%- set tool_call = tool_call.function %}\n            {%- endif %}\n            {{- '\\n<tool_call>\\n{\"name\": \"' }}\n            {{- tool_call.name }}\n            {{- '\", \"arguments\": ' }}\n            {{- tool_call.arguments | tojson }}\n            {{- '}\\n</tool_call>' }}\n        {%- endfor %}\n        {{- '<|im_end|>\\n' }}\n    {%- elif message.role == \"tool\" %}\n        {%- if (loop.index0 == 0) or (messages[loop.index0 - 1].role != \"tool\") %}\n            {{- '<|im_start|>user' }}\n        {%- endif %}\n        {{- '\\n<tool_response>\\n' }}\n        {{- message.content }}\n        {{- '\\n</tool_response>' }}\n        {%- if loop.last or (messages[loop.index0 + 1].role != \"tool\") %}\n            {{- '<|im_end|>\\n' }}\n        {%- endif %}\n    {%- endif %}\n{%- endfor %}\n{%- if add_generation_prompt %}\n    {{- '<|im_start|>assistant\\n' }}\n{%- endif %}\n",
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "padding_side": "right",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}

checkpoint-400/trainer_state.json

@@ -0,0 +1,2833 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.8983717012914093,
+  "eval_steps": 500,
+  "global_step": 400,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0022459292532285235,
+      "grad_norm": 1.9819730520248413,
+      "learning_rate": 1.1940298507462686e-06,
+      "loss": 0.8684,
+      "step": 1
+    },
+    {
+      "epoch": 0.004491858506457047,
+      "grad_norm": 2.0259573459625244,
+      "learning_rate": 2.3880597014925373e-06,
+      "loss": 0.8821,
+      "step": 2
+    },
+    {
+      "epoch": 0.00673778775968557,
+      "grad_norm": 2.085845947265625,
+      "learning_rate": 3.582089552238806e-06,
+      "loss": 0.9416,
+      "step": 3
+    },
+    {
+      "epoch": 0.008983717012914094,
+      "grad_norm": 1.9758139848709106,
+      "learning_rate": 4.7761194029850745e-06,
+      "loss": 0.9074,
+      "step": 4
+    },
+    {
+      "epoch": 0.011229646266142616,
+      "grad_norm": 1.7997628450393677,
+      "learning_rate": 5.970149253731343e-06,
+      "loss": 0.9069,
+      "step": 5
+    },
+    {
+      "epoch": 0.01347557551937114,
+      "grad_norm": 1.4710899591445923,
+      "learning_rate": 7.164179104477612e-06,
+      "loss": 0.8754,
+      "step": 6
+    },
+    {
+      "epoch": 0.015721504772599662,
+      "grad_norm": 1.4286643266677856,
+      "learning_rate": 8.35820895522388e-06,
+      "loss": 0.8896,
+      "step": 7
+    },
+    {
+      "epoch": 0.017967434025828188,
+      "grad_norm": 1.2477623224258423,
+      "learning_rate": 9.552238805970149e-06,
+      "loss": 0.8545,
+      "step": 8
+    },
+    {
+      "epoch": 0.02021336327905671,
+      "grad_norm": 1.2966960668563843,
+      "learning_rate": 1.074626865671642e-05,
+      "loss": 0.839,
+      "step": 9
+    },
+    {
+      "epoch": 0.022459292532285232,
+      "grad_norm": 1.4356369972229004,
+      "learning_rate": 1.1940298507462686e-05,
+      "loss": 0.8639,
+      "step": 10
+    },
+    {
+      "epoch": 0.024705221785513758,
+      "grad_norm": 1.1653496026992798,
+      "learning_rate": 1.3134328358208957e-05,
+      "loss": 0.8142,
+      "step": 11
+    },
+    {
+      "epoch": 0.02695115103874228,
+      "grad_norm": 0.9282035231590271,
+      "learning_rate": 1.4328358208955224e-05,
+      "loss": 0.8022,
+      "step": 12
+    },
+    {
+      "epoch": 0.029197080291970802,
+      "grad_norm": 1.086950421333313,
+      "learning_rate": 1.5522388059701494e-05,
+      "loss": 0.7908,
+      "step": 13
+    },
+    {
+      "epoch": 0.031443009545199324,
+      "grad_norm": 0.7987905144691467,
+      "learning_rate": 1.671641791044776e-05,
+      "loss": 0.7838,
+      "step": 14
+    },
+    {
+      "epoch": 0.033688938798427846,
+      "grad_norm": 0.7030066847801208,
+      "learning_rate": 1.791044776119403e-05,
+      "loss": 0.7856,
+      "step": 15
+    },
+    {
+      "epoch": 0.035934868051656375,
+      "grad_norm": 0.7216812372207642,
+      "learning_rate": 1.9104477611940298e-05,
+      "loss": 0.7666,
+      "step": 16
+    },
+    {
+      "epoch": 0.0381807973048849,
+      "grad_norm": 0.7004393935203552,
+      "learning_rate": 2.029850746268657e-05,
+      "loss": 0.7602,
+      "step": 17
+    },
+    {
+      "epoch": 0.04042672655811342,
+      "grad_norm": 0.5651209950447083,
+      "learning_rate": 2.149253731343284e-05,
+      "loss": 0.7637,
+      "step": 18
+    },
+    {
+      "epoch": 0.04267265581134194,
+      "grad_norm": 0.5799914598464966,
+      "learning_rate": 2.2686567164179106e-05,
+      "loss": 0.7357,
+      "step": 19
+    },
+    {
+      "epoch": 0.044918585064570464,
+      "grad_norm": 0.531233549118042,
+      "learning_rate": 2.3880597014925373e-05,
+      "loss": 0.7552,
+      "step": 20
+    },
+    {
+      "epoch": 0.047164514317798986,
+      "grad_norm": 0.5684418678283691,
+      "learning_rate": 2.5074626865671646e-05,
+      "loss": 0.7671,
+      "step": 21
+    },
+    {
+      "epoch": 0.049410443571027515,
+      "grad_norm": 0.4873054623603821,
+      "learning_rate": 2.6268656716417913e-05,
+      "loss": 0.737,
+      "step": 22
+    },
+    {
+      "epoch": 0.05165637282425604,
+      "grad_norm": 0.49275314807891846,
+      "learning_rate": 2.746268656716418e-05,
+      "loss": 0.7362,
+      "step": 23
+    },
+    {
+      "epoch": 0.05390230207748456,
+      "grad_norm": 0.47075843811035156,
+      "learning_rate": 2.8656716417910447e-05,
+      "loss": 0.7234,
+      "step": 24
+    },
+    {
+      "epoch": 0.05614823133071308,
+      "grad_norm": 0.3865251839160919,
+      "learning_rate": 2.985074626865672e-05,
+      "loss": 0.7169,
+      "step": 25
+    },
+    {
+      "epoch": 0.058394160583941604,
+      "grad_norm": 0.4154004156589508,
+      "learning_rate": 3.104477611940299e-05,
+      "loss": 0.7119,
+      "step": 26
+    },
+    {
+      "epoch": 0.060640089837170126,
+      "grad_norm": 0.37092125415802,
+      "learning_rate": 3.2238805970149255e-05,
+      "loss": 0.7138,
+      "step": 27
+    },
+    {
+      "epoch": 0.06288601909039865,
+      "grad_norm": 0.3488411605358124,
+      "learning_rate": 3.343283582089552e-05,
+      "loss": 0.7216,
+      "step": 28
+    },
+    {
+      "epoch": 0.06513194834362718,
+      "grad_norm": 0.32693928480148315,
+      "learning_rate": 3.462686567164179e-05,
+      "loss": 0.6925,
+      "step": 29
+    },
+    {
+      "epoch": 0.06737787759685569,
+      "grad_norm": 0.34904295206069946,
+      "learning_rate": 3.582089552238806e-05,
+      "loss": 0.7105,
+      "step": 30
+    },
+    {
+      "epoch": 0.06962380685008422,
+      "grad_norm": 0.32673367857933044,
+      "learning_rate": 3.701492537313433e-05,
+      "loss": 0.696,
+      "step": 31
+    },
+    {
+      "epoch": 0.07186973610331275,
+      "grad_norm": 0.32177790999412537,
+      "learning_rate": 3.8208955223880596e-05,
+      "loss": 0.7064,
+      "step": 32
+    },
+    {
+      "epoch": 0.07411566535654127,
+      "grad_norm": 0.3286134600639343,
+      "learning_rate": 3.940298507462687e-05,
+      "loss": 0.7091,
+      "step": 33
+    },
+    {
+      "epoch": 0.0763615946097698,
+      "grad_norm": 0.3438747525215149,
+      "learning_rate": 4.059701492537314e-05,
+      "loss": 0.7149,
+      "step": 34
+    },
+    {
+      "epoch": 0.07860752386299831,
+      "grad_norm": 0.29362648725509644,
+      "learning_rate": 4.1791044776119404e-05,
+      "loss": 0.685,
+      "step": 35
+    },
+    {
+      "epoch": 0.08085345311622684,
+      "grad_norm": 0.30074256658554077,
+      "learning_rate": 4.298507462686568e-05,
+      "loss": 0.7011,
+      "step": 36
+    },
+    {
+      "epoch": 0.08309938236945537,
+      "grad_norm": 0.3120618462562561,
+      "learning_rate": 4.4179104477611944e-05,
+      "loss": 0.684,
+      "step": 37
+    },
+    {
+      "epoch": 0.08534531162268388,
+      "grad_norm": 0.2569892406463623,
+      "learning_rate": 4.537313432835821e-05,
+      "loss": 0.684,
+      "step": 38
+    },
+    {
+      "epoch": 0.08759124087591241,
+      "grad_norm": 0.28327882289886475,
+      "learning_rate": 4.6567164179104485e-05,
+      "loss": 0.6968,
+      "step": 39
+    },
+    {
+      "epoch": 0.08983717012914093,
+      "grad_norm": 0.26424843072891235,
+      "learning_rate": 4.7761194029850745e-05,
+      "loss": 0.6915,
+      "step": 40
+    },
+    {
+      "epoch": 0.09208309938236946,
+      "grad_norm": 0.2620261609554291,
+      "learning_rate": 4.895522388059702e-05,
+      "loss": 0.6744,
+      "step": 41
+    },
+    {
+      "epoch": 0.09432902863559797,
+      "grad_norm": 0.32121092081069946,
+      "learning_rate": 5.014925373134329e-05,
+      "loss": 0.6746,
+      "step": 42
+    },
+    {
+      "epoch": 0.0965749578888265,
+      "grad_norm": 0.3997937738895416,
+      "learning_rate": 5.134328358208955e-05,
+      "loss": 0.6806,
+      "step": 43
+    },
+    {
+      "epoch": 0.09882088714205503,
+      "grad_norm": 0.3264799416065216,
+      "learning_rate": 5.2537313432835826e-05,
+      "loss": 0.6729,
+      "step": 44
+    },
+    {
+      "epoch": 0.10106681639528355,
+      "grad_norm": 0.33052176237106323,
+      "learning_rate": 5.37313432835821e-05,
+      "loss": 0.6758,
+      "step": 45
+    },
+    {
+      "epoch": 0.10331274564851207,
+      "grad_norm": 0.43345314264297485,
+      "learning_rate": 5.492537313432836e-05,
+      "loss": 0.6767,
+      "step": 46
+    },
+    {
+      "epoch": 0.10555867490174059,
+      "grad_norm": 0.37080681324005127,
+      "learning_rate": 5.6119402985074634e-05,
+      "loss": 0.6526,
+      "step": 47
+    },
+    {
+      "epoch": 0.10780460415496912,
+      "grad_norm": 0.381356418132782,
+      "learning_rate": 5.7313432835820894e-05,
+      "loss": 0.6739,
+      "step": 48
+    },
+    {
+      "epoch": 0.11005053340819765,
+      "grad_norm": 0.36677348613739014,
+      "learning_rate": 5.850746268656717e-05,
+      "loss": 0.6782,
+      "step": 49
+    },
+    {
+      "epoch": 0.11229646266142616,
+      "grad_norm": 0.40393349528312683,
+      "learning_rate": 5.970149253731344e-05,
+      "loss": 0.6528,
+      "step": 50
+    },
+    {
+      "epoch": 0.11454239191465469,
+      "grad_norm": 0.5078914165496826,
+      "learning_rate": 6.08955223880597e-05,
+      "loss": 0.6697,
+      "step": 51
+    },
+    {
+      "epoch": 0.11678832116788321,
+      "grad_norm": 0.742857813835144,
+      "learning_rate": 6.208955223880598e-05,
+      "loss": 0.6341,
+      "step": 52
+    },
+    {
+      "epoch": 0.11903425042111174,
+      "grad_norm": 0.8604367971420288,
+      "learning_rate": 6.328358208955224e-05,
+      "loss": 0.662,
+      "step": 53
+    },
+    {
+      "epoch": 0.12128017967434025,
+      "grad_norm": 0.7391287684440613,
+      "learning_rate": 6.447761194029851e-05,
+      "loss": 0.6696,
+      "step": 54
+    },
+    {
+      "epoch": 0.12352610892756878,
+      "grad_norm": 0.5703966617584229,
+      "learning_rate": 6.567164179104479e-05,
+      "loss": 0.6619,
+      "step": 55
+    },
+    {
+      "epoch": 0.1257720381807973,
+      "grad_norm": 0.7210264801979065,
+      "learning_rate": 6.686567164179104e-05,
+      "loss": 0.6647,
+      "step": 56
+    },
+    {
+      "epoch": 0.12801796743402583,
+      "grad_norm": 0.8133912682533264,
+      "learning_rate": 6.805970149253732e-05,
+      "loss": 0.658,
+      "step": 57
+    },
+    {
+      "epoch": 0.13026389668725435,
+      "grad_norm": 0.9062953591346741,
+      "learning_rate": 6.925373134328358e-05,
+      "loss": 0.6732,
+      "step": 58
+    },
+    {
+      "epoch": 0.13250982594048288,
+      "grad_norm": 0.9497516751289368,
+      "learning_rate": 7.044776119402986e-05,
+      "loss": 0.6743,
+      "step": 59
+    },
+    {
+      "epoch": 0.13475575519371139,
+      "grad_norm": 0.5923281908035278,
+      "learning_rate": 7.164179104477612e-05,
+      "loss": 0.6609,
+      "step": 60
+    },
+    {
+      "epoch": 0.13700168444693991,
+      "grad_norm": 0.839241087436676,
+      "learning_rate": 7.283582089552239e-05,
+      "loss": 0.6673,
+      "step": 61
+    },
+    {
+      "epoch": 0.13924761370016844,
+      "grad_norm": 0.9110313653945923,
+      "learning_rate": 7.402985074626866e-05,
+      "loss": 0.6795,
+      "step": 62
+    },
+    {
+      "epoch": 0.14149354295339697,
+      "grad_norm": 0.6465680599212646,
+      "learning_rate": 7.522388059701494e-05,
+      "loss": 0.6634,
+      "step": 63
+    },
+    {
+      "epoch": 0.1437394722066255,
+      "grad_norm": 0.5419987440109253,
+      "learning_rate": 7.641791044776119e-05,
+      "loss": 0.6489,
+      "step": 64
+    },
+    {
+      "epoch": 0.145985401459854,
+      "grad_norm": 0.6124593019485474,
+      "learning_rate": 7.761194029850747e-05,
+      "loss": 0.6617,
+      "step": 65
+    },
+    {
+      "epoch": 0.14823133071308253,
+      "grad_norm": 0.5836852788925171,
+      "learning_rate": 7.880597014925374e-05,
+      "loss": 0.6319,
+      "step": 66
+    },
+    {
+      "epoch": 0.15047725996631106,
+      "grad_norm": 0.6319289207458496,
+      "learning_rate": 8e-05,
+      "loss": 0.6504,
+      "step": 67
+    },
+    {
+      "epoch": 0.1527231892195396,
+      "grad_norm": 0.6081493496894836,
+      "learning_rate": 8.119402985074627e-05,
+      "loss": 0.65,
+      "step": 68
+    },
+    {
+      "epoch": 0.15496911847276812,
+      "grad_norm": 0.5973412394523621,
+      "learning_rate": 8.238805970149255e-05,
+      "loss": 0.6449,
+      "step": 69
+    },
+    {
+      "epoch": 0.15721504772599662,
+      "grad_norm": 0.6423139572143555,
+      "learning_rate": 8.358208955223881e-05,
+      "loss": 0.6584,
+      "step": 70
+    },
+    {
+      "epoch": 0.15946097697922515,
+      "grad_norm": 0.7579260468482971,
+      "learning_rate": 8.477611940298507e-05,
+      "loss": 0.6472,
+      "step": 71
+    },
+    {
+      "epoch": 0.16170690623245368,
+      "grad_norm": 0.8475743532180786,
+      "learning_rate": 8.597014925373135e-05,
+      "loss": 0.6405,
+      "step": 72
+    },
+    {
+      "epoch": 0.1639528354856822,
+      "grad_norm": 0.5964512228965759,
+      "learning_rate": 8.716417910447762e-05,
+      "loss": 0.633,
+      "step": 73
+    },
+    {
+      "epoch": 0.16619876473891074,
+      "grad_norm": 0.47265729308128357,
+      "learning_rate": 8.835820895522389e-05,
+      "loss": 0.6453,
+      "step": 74
+    },
+    {
+      "epoch": 0.16844469399213924,
+      "grad_norm": 0.7188097238540649,
+      "learning_rate": 8.955223880597014e-05,
+      "loss": 0.6603,
+      "step": 75
+    },
+    {
+      "epoch": 0.17069062324536777,
+      "grad_norm": 0.49939826130867004,
+      "learning_rate": 9.074626865671642e-05,
+      "loss": 0.6339,
+      "step": 76
+    },
+    {
+      "epoch": 0.1729365524985963,
+      "grad_norm": 0.5468081831932068,
+      "learning_rate": 9.194029850746269e-05,
+      "loss": 0.639,
+      "step": 77
+    },
+    {
+      "epoch": 0.17518248175182483,
+      "grad_norm": 0.6105530858039856,
+      "learning_rate": 9.313432835820897e-05,
+      "loss": 0.6537,
+      "step": 78
+    },
+    {
+      "epoch": 0.17742841100505333,
+      "grad_norm": 0.48114606738090515,
+      "learning_rate": 9.432835820895524e-05,
+      "loss": 0.6579,
+      "step": 79
+    },
+    {
+      "epoch": 0.17967434025828186,
+      "grad_norm": 0.6263488531112671,
+      "learning_rate": 9.552238805970149e-05,
+      "loss": 0.6292,
+      "step": 80
+    },
+    {
+      "epoch": 0.18192026951151039,
+      "grad_norm": 0.5369325280189514,
+      "learning_rate": 9.671641791044777e-05,
+      "loss": 0.6608,
+      "step": 81
+    },
+    {
+      "epoch": 0.18416619876473891,
+      "grad_norm": 0.7140039801597595,
+      "learning_rate": 9.791044776119404e-05,
+      "loss": 0.6339,
+      "step": 82
+    },
+    {
+      "epoch": 0.18641212801796744,
+      "grad_norm": 0.9011125564575195,
+      "learning_rate": 9.91044776119403e-05,
+      "loss": 0.6342,
+      "step": 83
+    },
+    {
+      "epoch": 0.18865805727119594,
+      "grad_norm": 1.1369616985321045,
+      "learning_rate": 0.00010029850746268659,
+      "loss": 0.6442,
+      "step": 84
+    },
+    {
+      "epoch": 0.19090398652442447,
+      "grad_norm": 1.0306285619735718,
+      "learning_rate": 0.00010149253731343285,
+      "loss": 0.6419,
+      "step": 85
+    },
+    {
+      "epoch": 0.193149915777653,
+      "grad_norm": 0.8979660272598267,
+      "learning_rate": 0.0001026865671641791,
+      "loss": 0.632,
+      "step": 86
+    },
+    {
+      "epoch": 0.19539584503088153,
+      "grad_norm": 0.6676183342933655,
+      "learning_rate": 0.00010388059701492539,
+      "loss": 0.6386,
+      "step": 87
+    },
+    {
+      "epoch": 0.19764177428411006,
+      "grad_norm": 0.7217721939086914,
+      "learning_rate": 0.00010507462686567165,
+      "loss": 0.6546,
+      "step": 88
+    },
+    {
+      "epoch": 0.19988770353733856,
+      "grad_norm": 0.7290446162223816,
+      "learning_rate": 0.00010626865671641792,
+      "loss": 0.6328,
+      "step": 89
+    },
+    {
+      "epoch": 0.2021336327905671,
+      "grad_norm": 0.8381432890892029,
+      "learning_rate": 0.0001074626865671642,
+      "loss": 0.6311,
+      "step": 90
+    },
+    {
+      "epoch": 0.20437956204379562,
+      "grad_norm": 1.0938982963562012,
+      "learning_rate": 0.00010865671641791045,
+      "loss": 0.6559,
+      "step": 91
+    },
+    {
+      "epoch": 0.20662549129702415,
+      "grad_norm": 0.8039063215255737,
+      "learning_rate": 0.00010985074626865672,
+      "loss": 0.636,
+      "step": 92
+    },
+    {
+      "epoch": 0.20887142055025268,
+      "grad_norm": 0.7171061635017395,
+      "learning_rate": 0.000111044776119403,
+      "loss": 0.6456,
+      "step": 93
+    },
+    {
+      "epoch": 0.21111734980348118,
+      "grad_norm": 0.7186174988746643,
+      "learning_rate": 0.00011223880597014927,
+      "loss": 0.6285,
+      "step": 94
+    },
+    {
+      "epoch": 0.2133632790567097,
+      "grad_norm": 0.6290779113769531,
+      "learning_rate": 0.00011343283582089553,
+      "loss": 0.6336,
+      "step": 95
+    },
+    {
+      "epoch": 0.21560920830993824,
+      "grad_norm": 0.7359249591827393,
+      "learning_rate": 0.00011462686567164179,
+      "loss": 0.6542,
+      "step": 96
+    },
+    {
+      "epoch": 0.21785513756316677,
+      "grad_norm": 0.775365948677063,
+      "learning_rate": 0.00011582089552238807,
+      "loss": 0.6369,
+      "step": 97
+    },
+    {
+      "epoch": 0.2201010668163953,
+      "grad_norm": 0.8260976076126099,
+      "learning_rate": 0.00011701492537313434,
+      "loss": 0.6142,
+      "step": 98
+    },
+    {
+      "epoch": 0.2223469960696238,
+      "grad_norm": 0.704872727394104,
+      "learning_rate": 0.00011820895522388062,
+      "loss": 0.6473,
+      "step": 99
+    },
+    {
+      "epoch": 0.22459292532285233,
+      "grad_norm": 0.5987293124198914,
+      "learning_rate": 0.00011940298507462688,
+      "loss": 0.6458,
+      "step": 100
+    },
+    {
+      "epoch": 0.22683885457608086,
+      "grad_norm": 0.7472802400588989,
+      "learning_rate": 0.00012059701492537314,
+      "loss": 0.6235,
+      "step": 101
+    },
+    {
+      "epoch": 0.22908478382930939,
+      "grad_norm": 0.7303177118301392,
+      "learning_rate": 0.0001217910447761194,
+      "loss": 0.6432,
+      "step": 102
+    },
+    {
+      "epoch": 0.2313307130825379,
+      "grad_norm": 0.5669957995414734,
+      "learning_rate": 0.00012298507462686568,
+      "loss": 0.6276,
+      "step": 103
+    },
+    {
+      "epoch": 0.23357664233576642,
+      "grad_norm": 0.47117286920547485,
+      "learning_rate": 0.00012417910447761195,
+      "loss": 0.6429,
+      "step": 104
+    },
+    {
+      "epoch": 0.23582257158899494,
+      "grad_norm": 0.6563988327980042,
+      "learning_rate": 0.00012537313432835822,
+      "loss": 0.6276,
+      "step": 105
+    },
+    {
+      "epoch": 0.23806850084222347,
+      "grad_norm": 0.5849066972732544,
+      "learning_rate": 0.00012656716417910448,
+      "loss": 0.6309,
+      "step": 106
+    },
+    {
+      "epoch": 0.240314430095452,
+      "grad_norm": 0.7347849607467651,
+      "learning_rate": 0.00012776119402985075,
+      "loss": 0.6382,
+      "step": 107
+    },
+    {
+      "epoch": 0.2425603593486805,
+      "grad_norm": 0.6520137190818787,
+      "learning_rate": 0.00012895522388059702,
+      "loss": 0.6386,
+      "step": 108
+    },
+    {
+      "epoch": 0.24480628860190903,
+      "grad_norm": 0.60540372133255,
+      "learning_rate": 0.00013014925373134329,
+      "loss": 0.613,
+      "step": 109
+    },
+    {
+      "epoch": 0.24705221785513756,
+      "grad_norm": 0.7710558176040649,
+      "learning_rate": 0.00013134328358208958,
+      "loss": 0.6104,
+      "step": 110
+    },
+    {
+      "epoch": 0.2492981471083661,
+      "grad_norm": 0.6582499742507935,
+      "learning_rate": 0.00013253731343283582,
+      "loss": 0.628,
+      "step": 111
+    },
+    {
+      "epoch": 0.2515440763615946,
+      "grad_norm": 0.6089588403701782,
+      "learning_rate": 0.00013373134328358209,
+      "loss": 0.6313,
+      "step": 112
+    },
+    {
+      "epoch": 0.2537900056148231,
+      "grad_norm": 0.5754179358482361,
+      "learning_rate": 0.00013492537313432838,
+      "loss": 0.6283,
+      "step": 113
+    },
+    {
+      "epoch": 0.25603593486805165,
+      "grad_norm": 0.617273211479187,
+      "learning_rate": 0.00013611940298507465,
+      "loss": 0.6187,
+      "step": 114
+    },
+    {
+      "epoch": 0.2582818641212802,
+      "grad_norm": 0.6104961037635803,
+      "learning_rate": 0.00013731343283582091,
+      "loss": 0.6267,
+      "step": 115
+    },
+    {
+      "epoch": 0.2605277933745087,
+      "grad_norm": 0.691856861114502,
+      "learning_rate": 0.00013850746268656715,
+      "loss": 0.6202,
+      "step": 116
+    },
+    {
+      "epoch": 0.26277372262773724,
+      "grad_norm": 0.8089864253997803,
+      "learning_rate": 0.00013970149253731345,
+      "loss": 0.635,
+      "step": 117
+    },
+    {
+      "epoch": 0.26501965188096577,
+      "grad_norm": 1.1346023082733154,
+      "learning_rate": 0.00014089552238805972,
+      "loss": 0.6462,
+      "step": 118
+    },
+    {
+      "epoch": 0.2672655811341943,
+      "grad_norm": 0.8319297432899475,
+      "learning_rate": 0.00014208955223880598,
+      "loss": 0.6179,
+      "step": 119
+    },
+    {
+      "epoch": 0.26951151038742277,
+      "grad_norm": 0.5904942154884338,
+      "learning_rate": 0.00014328358208955225,
+      "loss": 0.629,
+      "step": 120
+    },
+    {
+      "epoch": 0.2717574396406513,
+      "grad_norm": 0.5950160026550293,
+      "learning_rate": 0.00014447761194029852,
+      "loss": 0.6245,
+      "step": 121
+    },
+    {
+      "epoch": 0.27400336889387983,
+      "grad_norm": 0.6426451802253723,
+      "learning_rate": 0.00014567164179104478,
+      "loss": 0.614,
+      "step": 122
+    },
+    {
+      "epoch": 0.27624929814710836,
+      "grad_norm": 0.6028596758842468,
+      "learning_rate": 0.00014686567164179105,
+      "loss": 0.6127,
+      "step": 123
+    },
+    {
+      "epoch": 0.2784952274003369,
+      "grad_norm": 0.6075330972671509,
+      "learning_rate": 0.00014805970149253732,
+      "loss": 0.6283,
+      "step": 124
+    },
+    {
+      "epoch": 0.2807411566535654,
+      "grad_norm": 0.6084921360015869,
+      "learning_rate": 0.0001492537313432836,
+      "loss": 0.6351,
+      "step": 125
+    },
+    {
+      "epoch": 0.28298708590679394,
+      "grad_norm": 0.627112865447998,
+      "learning_rate": 0.00015044776119402988,
+      "loss": 0.6393,
+      "step": 126
+    },
+    {
+      "epoch": 0.2852330151600225,
+      "grad_norm": 0.6501988172531128,
+      "learning_rate": 0.00015164179104477612,
+      "loss": 0.6097,
+      "step": 127
+    },
+    {
+      "epoch": 0.287478944413251,
+      "grad_norm": 0.6280235648155212,
+      "learning_rate": 0.00015283582089552238,
+      "loss": 0.6281,
+      "step": 128
+    },
+    {
+      "epoch": 0.28972487366647953,
+      "grad_norm": 0.49232786893844604,
+      "learning_rate": 0.00015402985074626868,
+      "loss": 0.6341,
+      "step": 129
+    },
+    {
+      "epoch": 0.291970802919708,
+      "grad_norm": 0.5303974747657776,
+      "learning_rate": 0.00015522388059701495,
+      "loss": 0.6098,
+      "step": 130
+    },
+    {
+      "epoch": 0.29421673217293653,
+      "grad_norm": 0.5729207992553711,
+      "learning_rate": 0.0001564179104477612,
+      "loss": 0.617,
+      "step": 131
+    },
+    {
+      "epoch": 0.29646266142616506,
+      "grad_norm": 0.6265519857406616,
+      "learning_rate": 0.00015761194029850748,
+      "loss": 0.5968,
+      "step": 132
+    },
+    {
+      "epoch": 0.2987085906793936,
+      "grad_norm": 0.6463232636451721,
+      "learning_rate": 0.00015880597014925375,
+      "loss": 0.6391,
+      "step": 133
+    },
+    {
+      "epoch": 0.3009545199326221,
+      "grad_norm": 0.593257486820221,
+      "learning_rate": 0.00016,
+      "loss": 0.6189,
+      "step": 134
+    },
+    {
+      "epoch": 0.30320044918585065,
+      "grad_norm": 0.5925970077514648,
+      "learning_rate": 0.00015999972630083387,
+      "loss": 0.6139,
+      "step": 135
+    },
+    {
+      "epoch": 0.3054463784390792,
+      "grad_norm": 0.6394967436790466,
+      "learning_rate": 0.00015999890520520824,
+      "loss": 0.6038,
+      "step": 136
+    },
+    {
+      "epoch": 0.3076923076923077,
+      "grad_norm": 0.649044394493103,
+      "learning_rate": 0.00015999753671874147,
+      "loss": 0.614,
+      "step": 137
+    },
+    {
+      "epoch": 0.30993823694553624,
+      "grad_norm": 0.6179019808769226,
+      "learning_rate": 0.00015999562085079733,
+      "loss": 0.6171,
+      "step": 138
+    },
+    {
+      "epoch": 0.3121841661987647,
+      "grad_norm": 0.5114040374755859,
+      "learning_rate": 0.0001599931576144852,
+      "loss": 0.6076,
+      "step": 139
+    },
+    {
+      "epoch": 0.31443009545199324,
+      "grad_norm": 0.5721436142921448,
+      "learning_rate": 0.00015999014702665964,
+      "loss": 0.6173,
+      "step": 140
+    },
+    {
+      "epoch": 0.31667602470522177,
+      "grad_norm": 0.7164266109466553,
+      "learning_rate": 0.00015998658910792058,
+      "loss": 0.611,
+      "step": 141
+    },
+    {
+      "epoch": 0.3189219539584503,
+      "grad_norm": 0.8134217858314514,
+      "learning_rate": 0.00015998248388261302,
+      "loss": 0.6296,
+      "step": 142
+    },
+    {
+      "epoch": 0.32116788321167883,
+      "grad_norm": 0.828131377696991,
+      "learning_rate": 0.00015997783137882682,
+      "loss": 0.6331,
+      "step": 143
+    },
+    {
+      "epoch": 0.32341381246490736,
+      "grad_norm": 0.7628505825996399,
+      "learning_rate": 0.00015997263162839667,
+      "loss": 0.6524,
+      "step": 144
+    },
+    {
+      "epoch": 0.3256597417181359,
+      "grad_norm": 0.6403250098228455,
+      "learning_rate": 0.0001599668846669018,
+      "loss": 0.6097,
+      "step": 145
+    },
+    {
+      "epoch": 0.3279056709713644,
+      "grad_norm": 0.5496403574943542,
+      "learning_rate": 0.00015996059053366562,
+      "loss": 0.6187,
+      "step": 146
+    },
+    {
+      "epoch": 0.33015160022459294,
+      "grad_norm": 0.6352928876876831,
+      "learning_rate": 0.0001599537492717556,
+      "loss": 0.619,
+      "step": 147
+    },
+    {
+      "epoch": 0.3323975294778215,
+      "grad_norm": 0.6073532104492188,
+      "learning_rate": 0.00015994636092798295,
+      "loss": 0.6218,
+      "step": 148
+    },
+    {
+      "epoch": 0.33464345873104995,
+      "grad_norm": 0.40914225578308105,
+      "learning_rate": 0.00015993842555290226,
+      "loss": 0.6161,
+      "step": 149
+    },
+    {
+      "epoch": 0.3368893879842785,
+      "grad_norm": 0.4364437758922577,
+      "learning_rate": 0.0001599299432008112,
+      "loss": 0.637,
+      "step": 150
+    },
+    {
+      "epoch": 0.339135317237507,
+      "grad_norm": 0.5311095118522644,
+      "learning_rate": 0.00015992091392975002,
+      "loss": 0.5972,
+      "step": 151
+    },
+    {
+      "epoch": 0.34138124649073553,
+      "grad_norm": 0.545671284198761,
+      "learning_rate": 0.00015991133780150136,
+      "loss": 0.6103,
+      "step": 152
+    },
+    {
+      "epoch": 0.34362717574396406,
+      "grad_norm": 0.4276280105113983,
+      "learning_rate": 0.00015990121488158968,
+      "loss": 0.6148,
+      "step": 153
+    },
+    {
+      "epoch": 0.3458731049971926,
+      "grad_norm": 0.4059518575668335,
+      "learning_rate": 0.00015989054523928085,
+      "loss": 0.6332,
+      "step": 154
+    },
+    {
+      "epoch": 0.3481190342504211,
+      "grad_norm": 0.42028188705444336,
+      "learning_rate": 0.00015987932894758164,
+      "loss": 0.5972,
+      "step": 155
+    },
+    {
+      "epoch": 0.35036496350364965,
+      "grad_norm": 0.44919151067733765,
+      "learning_rate": 0.00015986756608323932,
+      "loss": 0.6017,
+      "step": 156
+    },
+    {
+      "epoch": 0.3526108927568782,
+      "grad_norm": 0.3990235924720764,
+      "learning_rate": 0.00015985525672674103,
+      "loss": 0.6146,
+      "step": 157
+    },
+    {
+      "epoch": 0.35485682201010665,
+      "grad_norm": 0.42787206172943115,
+      "learning_rate": 0.0001598424009623133,
+      "loss": 0.6199,
+      "step": 158
+    },
+    {
+      "epoch": 0.3571027512633352,
+      "grad_norm": 0.44234439730644226,
+      "learning_rate": 0.00015982899887792145,
+      "loss": 0.6279,
+      "step": 159
+    },
+    {
+      "epoch": 0.3593486805165637,
+      "grad_norm": 0.4152110815048218,
+      "learning_rate": 0.00015981505056526893,
+      "loss": 0.6032,
+      "step": 160
+    },
+    {
+      "epoch": 0.36159460976979224,
+      "grad_norm": 0.36194872856140137,
+      "learning_rate": 0.0001598005561197968,
+      "loss": 0.6286,
+      "step": 161
+    },
+    {
+      "epoch": 0.36384053902302077,
+      "grad_norm": 0.4214819669723511,
+      "learning_rate": 0.00015978551564068295,
+      "loss": 0.6006,
+      "step": 162
+    },
+    {
+      "epoch": 0.3660864682762493,
+      "grad_norm": 0.41948559880256653,
+      "learning_rate": 0.00015976992923084161,
+      "loss": 0.615,
+      "step": 163
+    },
+    {
+      "epoch": 0.36833239752947783,
+      "grad_norm": 0.44141775369644165,
+      "learning_rate": 0.00015975379699692245,
+      "loss": 0.6236,
+      "step": 164
+    },
+    {
+      "epoch": 0.37057832678270636,
+      "grad_norm": 0.47903239727020264,
+      "learning_rate": 0.00015973711904930993,
+      "loss": 0.5979,
+      "step": 165
+    },
+    {
+      "epoch": 0.3728242560359349,
+      "grad_norm": 0.45099982619285583,
+      "learning_rate": 0.00015971989550212255,
+      "loss": 0.6229,
+      "step": 166
+    },
+    {
+      "epoch": 0.3750701852891634,
+      "grad_norm": 0.4214828312397003,
+      "learning_rate": 0.00015970212647321207,
+      "loss": 0.6146,
+      "step": 167
+    },
+    {
+      "epoch": 0.3773161145423919,
+      "grad_norm": 0.43835896253585815,
+      "learning_rate": 0.00015968381208416273,
+      "loss": 0.6162,
+      "step": 168
+    },
+    {
+      "epoch": 0.3795620437956204,
+      "grad_norm": 0.4372192621231079,
+      "learning_rate": 0.00015966495246029033,
+      "loss": 0.6152,
+      "step": 169
+    },
+    {
+      "epoch": 0.38180797304884895,
+      "grad_norm": 0.45570138096809387,
+      "learning_rate": 0.00015964554773064148,
+      "loss": 0.6107,
+      "step": 170
+    },
+    {
+      "epoch": 0.3840539023020775,
+      "grad_norm": 0.5014758706092834,
+      "learning_rate": 0.0001596255980279926,
+      "loss": 0.5861,
+      "step": 171
+    },
+    {
+      "epoch": 0.386299831555306,
+      "grad_norm": 0.46727222204208374,
+      "learning_rate": 0.00015960510348884914,
+      "loss": 0.6104,
+      "step": 172
+    },
+    {
+      "epoch": 0.38854576080853453,
+      "grad_norm": 0.5081140398979187,
+      "learning_rate": 0.00015958406425344455,
+      "loss": 0.5948,
+      "step": 173
+    },
+    {
+      "epoch": 0.39079169006176306,
+      "grad_norm": 0.4470350742340088,
+      "learning_rate": 0.00015956248046573938,
+      "loss": 0.5924,
+      "step": 174
+    },
+    {
+      "epoch": 0.3930376193149916,
+      "grad_norm": 0.36340662837028503,
+      "learning_rate": 0.00015954035227342019,
+      "loss": 0.5972,
+      "step": 175
+    },
+    {
+      "epoch": 0.3952835485682201,
+      "grad_norm": 0.34276771545410156,
+      "learning_rate": 0.00015951767982789875,
+      "loss": 0.5955,
+      "step": 176
+    },
+    {
+      "epoch": 0.39752947782144865,
+      "grad_norm": 0.35867977142333984,
+      "learning_rate": 0.00015949446328431075,
+      "loss": 0.611,
+      "step": 177
+    },
+    {
+      "epoch": 0.3997754070746771,
+      "grad_norm": 0.3728366792201996,
+      "learning_rate": 0.00015947070280151492,
+      "loss": 0.6117,
+      "step": 178
+    },
+    {
+      "epoch": 0.40202133632790565,
+      "grad_norm": 0.32302939891815186,
+      "learning_rate": 0.00015944639854209184,
+      "loss": 0.6225,
+      "step": 179
+    },
+    {
+      "epoch": 0.4042672655811342,
+      "grad_norm": 0.33579641580581665,
+      "learning_rate": 0.00015942155067234293,
+      "loss": 0.5915,
+      "step": 180
+    },
+    {
+      "epoch": 0.4065131948343627,
+      "grad_norm": 0.29243412613868713,
+      "learning_rate": 0.00015939615936228922,
+      "loss": 0.5915,
+      "step": 181
+    },
+    {
+      "epoch": 0.40875912408759124,
+      "grad_norm": 0.32980793714523315,
+      "learning_rate": 0.00015937022478567023,
+      "loss": 0.6172,
+      "step": 182
+    },
+    {
+      "epoch": 0.41100505334081977,
+      "grad_norm": 0.30575114488601685,
+      "learning_rate": 0.0001593437471199427,
+      "loss": 0.5958,
+      "step": 183
+    },
+    {
+      "epoch": 0.4132509825940483,
+      "grad_norm": 0.3298383951187134,
+      "learning_rate": 0.00015931672654627958,
+      "loss": 0.5949,
+      "step": 184
+    },
+    {
+      "epoch": 0.41549691184727683,
+      "grad_norm": 0.330642431974411,
+      "learning_rate": 0.00015928916324956855,
+      "loss": 0.5929,
+      "step": 185
+    },
+    {
+      "epoch": 0.41774284110050536,
+      "grad_norm": 0.33059626817703247,
+      "learning_rate": 0.00015926105741841088,
+      "loss": 0.609,
+      "step": 186
+    },
+    {
+      "epoch": 0.41998877035373383,
+      "grad_norm": 0.3915445804595947,
+      "learning_rate": 0.00015923240924512014,
+      "loss": 0.6045,
+      "step": 187
+    },
+    {
+      "epoch": 0.42223469960696236,
+      "grad_norm": 0.3589101731777191,
+      "learning_rate": 0.00015920321892572088,
+      "loss": 0.6175,
+      "step": 188
+    },
+    {
+      "epoch": 0.4244806288601909,
+      "grad_norm": 0.399964302778244,
+      "learning_rate": 0.00015917348665994723,
+      "loss": 0.6157,
+      "step": 189
+    },
+    {
+      "epoch": 0.4267265581134194,
+      "grad_norm": 0.3923908770084381,
+      "learning_rate": 0.0001591432126512416,
+      "loss": 0.6041,
+      "step": 190
+    },
+    {
+      "epoch": 0.42897248736664795,
+      "grad_norm": 0.38844165205955505,
+      "learning_rate": 0.0001591123971067533,
+      "loss": 0.5865,
+      "step": 191
+    },
+    {
+      "epoch": 0.4312184166198765,
+      "grad_norm": 0.41744178533554077,
+      "learning_rate": 0.00015908104023733697,
+      "loss": 0.5823,
+      "step": 192
+    },
+    {
+      "epoch": 0.433464345873105,
+      "grad_norm": 0.3478281795978546,
+      "learning_rate": 0.0001590491422575514,
+      "loss": 0.6064,
+      "step": 193
+    },
+    {
+      "epoch": 0.43571027512633353,
+      "grad_norm": 0.38580065965652466,
+      "learning_rate": 0.00015901670338565785,
+      "loss": 0.6119,
+      "step": 194
+    },
+    {
+      "epoch": 0.43795620437956206,
+      "grad_norm": 0.5283933877944946,
+      "learning_rate": 0.0001589837238436186,
+      "loss": 0.5945,
+      "step": 195
+    },
+    {
+      "epoch": 0.4402021336327906,
+      "grad_norm": 0.48087194561958313,
+      "learning_rate": 0.00015895020385709553,
+      "loss": 0.6058,
+      "step": 196
+    },
+    {
+      "epoch": 0.44244806288601907,
+      "grad_norm": 0.35071608424186707,
+      "learning_rate": 0.00015891614365544837,
+      "loss": 0.5672,
+      "step": 197
+    },
+    {
+      "epoch": 0.4446939921392476,
+      "grad_norm": 0.3820844888687134,
+      "learning_rate": 0.0001588815434717334,
+      "loss": 0.5898,
+      "step": 198
+    },
+    {
+      "epoch": 0.4469399213924761,
+      "grad_norm": 0.3622789680957794,
+      "learning_rate": 0.0001588464035427016,
+      "loss": 0.5842,
+      "step": 199
+    },
+    {
+      "epoch": 0.44918585064570465,
+      "grad_norm": 0.348568856716156,
+      "learning_rate": 0.00015881072410879726,
+      "loss": 0.6025,
+      "step": 200
+    },
+    {
+      "epoch": 0.4514317798989332,
+      "grad_norm": 0.36718496680259705,
+      "learning_rate": 0.00015877450541415615,
+      "loss": 0.5888,
+      "step": 201
+    },
+    {
+      "epoch": 0.4536777091521617,
+      "grad_norm": 0.39695170521736145,
+      "learning_rate": 0.0001587377477066039,
+      "loss": 0.6159,
+      "step": 202
+    },
+    {
+      "epoch": 0.45592363840539024,
+      "grad_norm": 0.4380107522010803,
+      "learning_rate": 0.0001587004512376544,
+      "loss": 0.6001,
+      "step": 203
+    },
+    {
+      "epoch": 0.45816956765861877,
+      "grad_norm": 0.40494075417518616,
+      "learning_rate": 0.00015866261626250794,
+      "loss": 0.6016,
+      "step": 204
+    },
+    {
+      "epoch": 0.4604154969118473,
+      "grad_norm": 0.3275372385978699,
+      "learning_rate": 0.00015862424304004954,
+      "loss": 0.5918,
+      "step": 205
+    },
+    {
+      "epoch": 0.4626614261650758,
+      "grad_norm": 0.3288284242153168,
+      "learning_rate": 0.00015858533183284718,
+      "loss": 0.608,
+      "step": 206
+    },
+    {
+      "epoch": 0.4649073554183043,
+      "grad_norm": 0.32171040773391724,
+      "learning_rate": 0.00015854588290714999,
+      "loss": 0.5816,
+      "step": 207
+    },
+    {
+      "epoch": 0.46715328467153283,
+      "grad_norm": 0.3992040157318115,
+      "learning_rate": 0.00015850589653288642,
+      "loss": 0.591,
+      "step": 208
+    },
+    {
+      "epoch": 0.46939921392476136,
+      "grad_norm": 0.38158226013183594,
+      "learning_rate": 0.00015846537298366242,
+      "loss": 0.5831,
+      "step": 209
+    },
+    {
+      "epoch": 0.4716451431779899,
+      "grad_norm": 0.32366326451301575,
+      "learning_rate": 0.0001584243125367595,
+      "loss": 0.5822,
+      "step": 210
+    },
+    {
+      "epoch": 0.4738910724312184,
+      "grad_norm": 0.41187676787376404,
+      "learning_rate": 0.00015838271547313293,
+      "loss": 0.6027,
+      "step": 211
+    },
+    {
+      "epoch": 0.47613700168444695,
+      "grad_norm": 0.48473531007766724,
+      "learning_rate": 0.00015834058207740974,
+      "loss": 0.5819,
+      "step": 212
+    },
+    {
+      "epoch": 0.4783829309376755,
+      "grad_norm": 0.3934939205646515,
+      "learning_rate": 0.00015829791263788682,
+      "loss": 0.6042,
+      "step": 213
+    },
+    {
+      "epoch": 0.480628860190904,
+      "grad_norm": 0.32344624400138855,
+      "learning_rate": 0.00015825470744652894,
+      "loss": 0.5717,
+      "step": 214
+    },
+    {
+      "epoch": 0.48287478944413254,
+      "grad_norm": 0.27189725637435913,
+      "learning_rate": 0.0001582109667989667,
+      "loss": 0.6015,
+      "step": 215
+    },
+    {
+      "epoch": 0.485120718697361,
+      "grad_norm": 0.349128395318985,
+      "learning_rate": 0.00015816669099449454,
+      "loss": 0.6037,
+      "step": 216
+    },
+    {
+      "epoch": 0.48736664795058954,
+      "grad_norm": 0.3456957936286926,
+      "learning_rate": 0.00015812188033606877,
+      "loss": 0.5974,
+      "step": 217
+    },
+    {
+      "epoch": 0.48961257720381807,
+      "grad_norm": 0.29926273226737976,
+      "learning_rate": 0.00015807653513030538,
+      "loss": 0.6,
+      "step": 218
+    },
+    {
+      "epoch": 0.4918585064570466,
+      "grad_norm": 0.3260749280452728,
+      "learning_rate": 0.00015803065568747798,
+      "loss": 0.5955,
+      "step": 219
+    },
+    {
+      "epoch": 0.4941044357102751,
+      "grad_norm": 0.4071785509586334,
+      "learning_rate": 0.00015798424232151573,
+      "loss": 0.5899,
+      "step": 220
+    },
+    {
+      "epoch": 0.49635036496350365,
+      "grad_norm": 0.37568220496177673,
+      "learning_rate": 0.00015793729535000108,
+      "loss": 0.6008,
+      "step": 221
+    },
+    {
+      "epoch": 0.4985962942167322,
+      "grad_norm": 0.4158768355846405,
+      "learning_rate": 0.00015788981509416773,
+      "loss": 0.5897,
+      "step": 222
+    },
+    {
+      "epoch": 0.5008422234699607,
+      "grad_norm": 0.44514065980911255,
+      "learning_rate": 0.00015784180187889833,
+      "loss": 0.5807,
+      "step": 223
+    },
+    {
+      "epoch": 0.5030881527231892,
+      "grad_norm": 0.37475013732910156,
+      "learning_rate": 0.00015779325603272232,
+      "loss": 0.586,
+      "step": 224
+    },
+    {
+      "epoch": 0.5053340819764177,
+      "grad_norm": 0.4093579649925232,
+      "learning_rate": 0.0001577441778878136,
+      "loss": 0.5966,
+      "step": 225
+    },
+    {
+      "epoch": 0.5075800112296462,
+      "grad_norm": 0.4048860967159271,
+      "learning_rate": 0.00015769456777998842,
+      "loss": 0.6107,
+      "step": 226
+    },
+    {
+      "epoch": 0.5098259404828748,
+      "grad_norm": 0.31557515263557434,
+      "learning_rate": 0.00015764442604870285,
+      "loss": 0.609,
+      "step": 227
+    },
+    {
+      "epoch": 0.5120718697361033,
+      "grad_norm": 0.33514130115509033,
+      "learning_rate": 0.0001575937530370507,
+      "loss": 0.5866,
+      "step": 228
+    },
+    {
+      "epoch": 0.5143177989893318,
+      "grad_norm": 0.3601367771625519,
+      "learning_rate": 0.0001575425490917609,
+      "loss": 0.586,
+      "step": 229
+    },
+    {
+      "epoch": 0.5165637282425604,
+      "grad_norm": 0.3701965808868408,
+      "learning_rate": 0.00015749081456319544,
+      "loss": 0.5755,
+      "step": 230
+    },
+    {
+      "epoch": 0.5188096574957889,
+      "grad_norm": 0.3042786419391632,
+      "learning_rate": 0.0001574385498053468,
+      "loss": 0.5978,
+      "step": 231
+    },
+    {
+      "epoch": 0.5210555867490174,
+      "grad_norm": 0.33692997694015503,
+      "learning_rate": 0.00015738575517583542,
+      "loss": 0.6078,
+      "step": 232
+    },
+    {
+      "epoch": 0.523301516002246,
+      "grad_norm": 0.36644524335861206,
+      "learning_rate": 0.00015733243103590748,
+      "loss": 0.575,
+      "step": 233
+    },
+    {
+      "epoch": 0.5255474452554745,
+      "grad_norm": 0.3802548050880432,
+      "learning_rate": 0.00015727857775043227,
+      "loss": 0.6041,
+      "step": 234
+    },
+    {
+      "epoch": 0.527793374508703,
+      "grad_norm": 0.37767213582992554,
+      "learning_rate": 0.00015722419568789983,
+      "loss": 0.591,
+      "step": 235
+    },
+    {
+      "epoch": 0.5300393037619315,
+      "grad_norm": 0.38524535298347473,
+      "learning_rate": 0.00015716928522041825,
+      "loss": 0.601,
+      "step": 236
+    },
+    {
+      "epoch": 0.5322852330151601,
+      "grad_norm": 0.4806888699531555,
+      "learning_rate": 0.00015711384672371126,
+      "loss": 0.5935,
+      "step": 237
+    },
+    {
+      "epoch": 0.5345311622683886,
+      "grad_norm": 0.3960248827934265,
+      "learning_rate": 0.0001570578805771156,
+      "loss": 0.5789,
+      "step": 238
+    },
+    {
+      "epoch": 0.5367770915216171,
+      "grad_norm": 0.29831525683403015,
+      "learning_rate": 0.00015700138716357852,
+      "loss": 0.5917,
+      "step": 239
+    },
+    {
+      "epoch": 0.5390230207748455,
+      "grad_norm": 0.30690401792526245,
+      "learning_rate": 0.00015694436686965497,
+      "loss": 0.5819,
+      "step": 240
+    },
+    {
+      "epoch": 0.5412689500280741,
+      "grad_norm": 0.30107825994491577,
+      "learning_rate": 0.00015688682008550514,
+      "loss": 0.5965,
+      "step": 241
+    },
+    {
+      "epoch": 0.5435148792813026,
+      "grad_norm": 0.30696406960487366,
+      "learning_rate": 0.0001568287472048917,
+      "loss": 0.6025,
+      "step": 242
+    },
+    {
+      "epoch": 0.5457608085345311,
+      "grad_norm": 0.32280731201171875,
+      "learning_rate": 0.00015677014862517714,
+      "loss": 0.5868,
+      "step": 243
+    },
+    {
+      "epoch": 0.5480067377877597,
+      "grad_norm": 0.31739377975463867,
+      "learning_rate": 0.000156711024747321,
+      "loss": 0.5898,
+      "step": 244
+    },
+    {
+      "epoch": 0.5502526670409882,
+      "grad_norm": 0.3620510995388031,
+      "learning_rate": 0.0001566513759758772,
+      "loss": 0.5621,
+      "step": 245
+    },
+    {
+      "epoch": 0.5524985962942167,
+      "grad_norm": 0.26646366715431213,
+      "learning_rate": 0.00015659120271899118,
+      "loss": 0.5731,
+      "step": 246
+    },
+    {
+      "epoch": 0.5547445255474452,
+      "grad_norm": 0.3814524710178375,
+      "learning_rate": 0.00015653050538839722,
+      "loss": 0.5947,
+      "step": 247
+    },
+    {
+      "epoch": 0.5569904548006738,
+      "grad_norm": 0.4031396210193634,
+      "learning_rate": 0.00015646928439941557,
+      "loss": 0.612,
+      "step": 248
+    },
+    {
+      "epoch": 0.5592363840539023,
+      "grad_norm": 0.38268253207206726,
+      "learning_rate": 0.00015640754017094954,
+      "loss": 0.5792,
+      "step": 249
+    },
+    {
+      "epoch": 0.5614823133071308,
+      "grad_norm": 0.37941139936447144,
+      "learning_rate": 0.0001563452731254827,
+      "loss": 0.6071,
+      "step": 250
+    },
+    {
+      "epoch": 0.5637282425603594,
+      "grad_norm": 0.3618276119232178,
+      "learning_rate": 0.00015628248368907603,
+      "loss": 0.5776,
+      "step": 251
+    },
+    {
+      "epoch": 0.5659741718135879,
+      "grad_norm": 0.3906313180923462,
+      "learning_rate": 0.000156219172291365,
+      "loss": 0.5732,
+      "step": 252
+    },
+    {
+      "epoch": 0.5682201010668164,
+      "grad_norm": 0.4234972894191742,
+      "learning_rate": 0.0001561553393655564,
+      "loss": 0.5674,
+      "step": 253
+    },
+    {
+      "epoch": 0.570466030320045,
+      "grad_norm": 0.4400922954082489,
+      "learning_rate": 0.00015609098534842582,
+      "loss": 0.5894,
+      "step": 254
+    },
+    {
+      "epoch": 0.5727119595732735,
+      "grad_norm": 0.38799750804901123,
+      "learning_rate": 0.0001560261106803142,
+      "loss": 0.5833,
+      "step": 255
+    },
+    {
+      "epoch": 0.574957888826502,
+      "grad_norm": 0.31524044275283813,
+      "learning_rate": 0.00015596071580512515,
+      "loss": 0.5841,
+      "step": 256
+    },
+    {
+      "epoch": 0.5772038180797305,
+      "grad_norm": 0.3451038599014282,
+      "learning_rate": 0.00015589480117032174,
+      "loss": 0.6003,
+      "step": 257
+    },
+    {
+      "epoch": 0.5794497473329591,
+      "grad_norm": 0.3648560047149658,
+      "learning_rate": 0.00015582836722692346,
+      "loss": 0.5787,
+      "step": 258
+    },
+    {
+      "epoch": 0.5816956765861875,
+      "grad_norm": 0.37476226687431335,
+      "learning_rate": 0.00015576141442950317,
+      "loss": 0.5719,
+      "step": 259
+    },
+    {
+      "epoch": 0.583941605839416,
+      "grad_norm": 0.33187106251716614,
+      "learning_rate": 0.00015569394323618403,
+      "loss": 0.5785,
+      "step": 260
+    },
+    {
+      "epoch": 0.5861875350926445,
+      "grad_norm": 0.36073818802833557,
+      "learning_rate": 0.00015562595410863626,
+      "loss": 0.5965,
+      "step": 261
+    },
+    {
+      "epoch": 0.5884334643458731,
+      "grad_norm": 0.3586486577987671,
+      "learning_rate": 0.00015555744751207404,
+      "loss": 0.5857,
+      "step": 262
+    },
+    {
+      "epoch": 0.5906793935991016,
+      "grad_norm": 0.44820883870124817,
+      "learning_rate": 0.0001554884239152523,
+      "loss": 0.5804,
+      "step": 263
+    },
+    {
+      "epoch": 0.5929253228523301,
+      "grad_norm": 0.43128344416618347,
+      "learning_rate": 0.00015541888379046366,
+      "loss": 0.5613,
+      "step": 264
+    },
+    {
+      "epoch": 0.5951712521055587,
+      "grad_norm": 0.38606396317481995,
+      "learning_rate": 0.0001553488276135349,
+      "loss": 0.5958,
+      "step": 265
+    },
+    {
+      "epoch": 0.5974171813587872,
+      "grad_norm": 0.36493563652038574,
+      "learning_rate": 0.0001552782558638239,
+      "loss": 0.5663,
+      "step": 266
+    },
+    {
+      "epoch": 0.5996631106120157,
+      "grad_norm": 0.40545809268951416,
+      "learning_rate": 0.00015520716902421648,
+      "loss": 0.5934,
+      "step": 267
+    },
+    {
+      "epoch": 0.6019090398652442,
+      "grad_norm": 0.42288488149642944,
+      "learning_rate": 0.00015513556758112282,
+      "loss": 0.5729,
+      "step": 268
+    },
+    {
+      "epoch": 0.6041549691184728,
+      "grad_norm": 0.2895568311214447,
+      "learning_rate": 0.00015506345202447432,
+      "loss": 0.6046,
+      "step": 269
+    },
+    {
+      "epoch": 0.6064008983717013,
+      "grad_norm": 0.3440837562084198,
+      "learning_rate": 0.00015499082284772017,
+      "loss": 0.5654,
+      "step": 270
+    },
+    {
+      "epoch": 0.6086468276249298,
+      "grad_norm": 0.36002352833747864,
+      "learning_rate": 0.00015491768054782395,
+      "loss": 0.5923,
+      "step": 271
+    },
+    {
+      "epoch": 0.6108927568781584,
+      "grad_norm": 0.28700196743011475,
+      "learning_rate": 0.00015484402562526036,
+      "loss": 0.5826,
+      "step": 272
+    },
+    {
+      "epoch": 0.6131386861313869,
+      "grad_norm": 0.32599133253097534,
+      "learning_rate": 0.0001547698585840117,
+      "loss": 0.5783,
+      "step": 273
+    },
+    {
+      "epoch": 0.6153846153846154,
+      "grad_norm": 0.37215182185173035,
+      "learning_rate": 0.00015469517993156435,
+      "loss": 0.583,
+      "step": 274
+    },
+    {
+      "epoch": 0.617630544637844,
+      "grad_norm": 0.3325370252132416,
+      "learning_rate": 0.0001546199901789055,
+      "loss": 0.582,
+      "step": 275
+    },
+    {
+      "epoch": 0.6198764738910725,
+      "grad_norm": 0.3477807939052582,
+      "learning_rate": 0.00015454428984051937,
+      "loss": 0.5726,
+      "step": 276
+    },
+    {
+      "epoch": 0.622122403144301,
+      "grad_norm": 0.37678489089012146,
+      "learning_rate": 0.000154468079434384,
+      "loss": 0.5786,
+      "step": 277
+    },
+    {
+      "epoch": 0.6243683323975294,
+      "grad_norm": 0.3045758008956909,
+      "learning_rate": 0.00015439135948196756,
+      "loss": 0.5829,
+      "step": 278
+    },
+    {
+      "epoch": 0.626614261650758,
+      "grad_norm": 0.3221797049045563,
+      "learning_rate": 0.0001543141305082246,
+      "loss": 0.5811,
+      "step": 279
+    },
+    {
+      "epoch": 0.6288601909039865,
+      "grad_norm": 0.35202842950820923,
+      "learning_rate": 0.00015423639304159288,
+      "loss": 0.5655,
+      "step": 280
+    },
+    {
+      "epoch": 0.631106120157215,
+      "grad_norm": 0.2838123142719269,
+      "learning_rate": 0.00015415814761398936,
+      "loss": 0.5991,
+      "step": 281
+    },
+    {
+      "epoch": 0.6333520494104435,
+      "grad_norm": 0.33944493532180786,
+      "learning_rate": 0.0001540793947608067,
+      "loss": 0.5764,
+      "step": 282
+    },
+    {
+      "epoch": 0.6355979786636721,
+      "grad_norm": 0.29667678475379944,
+      "learning_rate": 0.0001540001350209097,
+      "loss": 0.5745,
+      "step": 283
+    },
+    {
+      "epoch": 0.6378439079169006,
+      "grad_norm": 0.37716934084892273,
+      "learning_rate": 0.00015392036893663148,
+      "loss": 0.5739,
+      "step": 284
+    },
+    {
+      "epoch": 0.6400898371701291,
+      "grad_norm": 0.3955274522304535,
+      "learning_rate": 0.00015384009705376978,
+      "loss": 0.574,
+      "step": 285
+    },
+    {
+      "epoch": 0.6423357664233577,
+      "grad_norm": 0.29740408062934875,
+      "learning_rate": 0.00015375931992158331,
+      "loss": 0.567,
+      "step": 286
+    },
+    {
+      "epoch": 0.6445816956765862,
+      "grad_norm": 0.3198919892311096,
+      "learning_rate": 0.0001536780380927879,
+      "loss": 0.5672,
+      "step": 287
+    },
+    {
+      "epoch": 0.6468276249298147,
+      "grad_norm": 0.3355892598628998,
+      "learning_rate": 0.0001535962521235528,
+      "loss": 0.57,
+      "step": 288
+    },
+    {
+      "epoch": 0.6490735541830432,
+      "grad_norm": 0.32803425192832947,
+      "learning_rate": 0.00015351396257349675,
+      "loss": 0.5839,
+      "step": 289
+    },
+    {
+      "epoch": 0.6513194834362718,
+      "grad_norm": 0.3538999557495117,
+      "learning_rate": 0.00015343117000568432,
+      "loss": 0.5864,
+      "step": 290
+    },
+    {
+      "epoch": 0.6535654126895003,
+      "grad_norm": 0.3156984746456146,
+      "learning_rate": 0.00015334787498662192,
+      "loss": 0.5872,
+      "step": 291
+    },
+    {
+      "epoch": 0.6558113419427288,
+      "grad_norm": 0.336056113243103,
+      "learning_rate": 0.00015326407808625395,
+      "loss": 0.578,
+      "step": 292
+    },
+    {
+      "epoch": 0.6580572711959574,
+      "grad_norm": 0.3894708454608917,
+      "learning_rate": 0.00015317977987795898,
+      "loss": 0.5682,
+      "step": 293
+    },
+    {
+      "epoch": 0.6603032004491859,
+      "grad_norm": 0.3500683605670929,
+      "learning_rate": 0.00015309498093854577,
+      "loss": 0.5934,
+      "step": 294
+    },
+    {
+      "epoch": 0.6625491297024144,
+      "grad_norm": 0.331767201423645,
+      "learning_rate": 0.00015300968184824926,
+      "loss": 0.5781,
+      "step": 295
+    },
+    {
+      "epoch": 0.664795058955643,
+      "grad_norm": 0.4042721092700958,
+      "learning_rate": 0.0001529238831907267,
+      "loss": 0.5811,
+      "step": 296
+    },
+    {
+      "epoch": 0.6670409882088714,
+      "grad_norm": 0.2907451093196869,
+      "learning_rate": 0.00015283758555305362,
+      "loss": 0.5925,
+      "step": 297
+    },
+    {
+      "epoch": 0.6692869174620999,
+      "grad_norm": 0.28044381737709045,
+      "learning_rate": 0.0001527507895257198,
+      "loss": 0.5717,
+      "step": 298
+    },
+    {
+      "epoch": 0.6715328467153284,
+      "grad_norm": 0.2812747359275818,
+      "learning_rate": 0.00015266349570262528,
+      "loss": 0.5796,
+      "step": 299
+    },
+    {
+      "epoch": 0.673778775968557,
+      "grad_norm": 0.28039273619651794,
+      "learning_rate": 0.00015257570468107617,
+      "loss": 0.5682,
+      "step": 300
+    },
+    {
+      "epoch": 0.6760247052217855,
+      "grad_norm": 0.2821033000946045,
+      "learning_rate": 0.00015248741706178073,
+      "loss": 0.5939,
+      "step": 301
+    },
+    {
+      "epoch": 0.678270634475014,
+      "grad_norm": 0.31085771322250366,
+      "learning_rate": 0.0001523986334488452,
+      "loss": 0.5829,
+      "step": 302
+    },
+    {
+      "epoch": 0.6805165637282425,
+      "grad_norm": 0.31658798456192017,
+      "learning_rate": 0.00015230935444976955,
+      "loss": 0.6073,
+      "step": 303
+    },
+    {
+      "epoch": 0.6827624929814711,
+      "grad_norm": 0.28057488799095154,
+      "learning_rate": 0.00015221958067544348,
+      "loss": 0.5888,
+      "step": 304
+    },
+    {
+      "epoch": 0.6850084222346996,
+      "grad_norm": 0.29499179124832153,
+      "learning_rate": 0.00015212931274014214,
+      "loss": 0.5713,
+      "step": 305
+    },
+    {
+      "epoch": 0.6872543514879281,
+      "grad_norm": 0.31696656346321106,
+      "learning_rate": 0.00015203855126152204,
+      "loss": 0.5956,
+      "step": 306
+    },
+    {
+      "epoch": 0.6895002807411567,
+      "grad_norm": 0.2905656695365906,
+      "learning_rate": 0.00015194729686061672,
+      "loss": 0.56,
+      "step": 307
+    },
+    {
+      "epoch": 0.6917462099943852,
+      "grad_norm": 0.33711618185043335,
+      "learning_rate": 0.00015185555016183246,
+      "loss": 0.5816,
+      "step": 308
+    },
+    {
+      "epoch": 0.6939921392476137,
+      "grad_norm": 0.3962436616420746,
+      "learning_rate": 0.00015176331179294416,
+      "loss": 0.5933,
+      "step": 309
+    },
+    {
+      "epoch": 0.6962380685008422,
+      "grad_norm": 0.2827875316143036,
+      "learning_rate": 0.00015167058238509093,
+      "loss": 0.5529,
+      "step": 310
+    },
+    {
+      "epoch": 0.6984839977540708,
+      "grad_norm": 0.252986878156662,
+      "learning_rate": 0.00015157736257277182,
+      "loss": 0.5915,
+      "step": 311
+    },
+    {
+      "epoch": 0.7007299270072993,
+      "grad_norm": 0.28363773226737976,
+      "learning_rate": 0.00015148365299384145,
+      "loss": 0.5621,
+      "step": 312
+    },
+    {
+      "epoch": 0.7029758562605278,
+      "grad_norm": 0.26527139544487,
+      "learning_rate": 0.00015138945428950566,
+      "loss": 0.5791,
+      "step": 313
+    },
+    {
+      "epoch": 0.7052217855137564,
+      "grad_norm": 0.24393455684185028,
+      "learning_rate": 0.0001512947671043171,
+      "loss": 0.5549,
+      "step": 314
+    },
+    {
+      "epoch": 0.7074677147669849,
+      "grad_norm": 0.24904131889343262,
+      "learning_rate": 0.00015119959208617092,
+      "loss": 0.5627,
+      "step": 315
+    },
+    {
+      "epoch": 0.7097136440202133,
+      "grad_norm": 0.3018868863582611,
+      "learning_rate": 0.00015110392988630016,
+      "loss": 0.5802,
+      "step": 316
+    },
+    {
+      "epoch": 0.7119595732734418,
+      "grad_norm": 0.34517163038253784,
+      "learning_rate": 0.0001510077811592714,
+      "loss": 0.5831,
+      "step": 317
+    },
+    {
+      "epoch": 0.7142055025266704,
+      "grad_norm": 0.3295687437057495,
+      "learning_rate": 0.00015091114656298033,
+      "loss": 0.5978,
+      "step": 318
+    },
+    {
+      "epoch": 0.7164514317798989,
+      "grad_norm": 0.3116067945957184,
+      "learning_rate": 0.00015081402675864717,
+      "loss": 0.58,
+      "step": 319
+    },
+    {
+      "epoch": 0.7186973610331274,
+      "grad_norm": 0.2843012809753418,
+      "learning_rate": 0.00015071642241081212,
+      "loss": 0.5837,
+      "step": 320
+    },
+    {
+      "epoch": 0.720943290286356,
+      "grad_norm": 0.27185961604118347,
+      "learning_rate": 0.00015061833418733095,
+      "loss": 0.5746,
+      "step": 321
+    },
+    {
+      "epoch": 0.7231892195395845,
+      "grad_norm": 0.26890790462493896,
+      "learning_rate": 0.00015051976275937023,
+      "loss": 0.5642,
+      "step": 322
+    },
+    {
+      "epoch": 0.725435148792813,
+      "grad_norm": 0.29379114508628845,
+      "learning_rate": 0.00015042070880140292,
+      "loss": 0.5796,
+      "step": 323
+    },
+    {
+      "epoch": 0.7276810780460415,
+      "grad_norm": 0.2906297743320465,
+      "learning_rate": 0.0001503211729912037,
+      "loss": 0.5666,
+      "step": 324
+    },
+    {
+      "epoch": 0.7299270072992701,
+      "grad_norm": 0.2815559506416321,
+      "learning_rate": 0.00015022115600984423,
+      "loss": 0.5582,
+      "step": 325
+    },
+    {
+      "epoch": 0.7321729365524986,
+      "grad_norm": 0.3286380469799042,
+      "learning_rate": 0.0001501206585416886,
+      "loss": 0.5462,
+      "step": 326
+    },
+    {
+      "epoch": 0.7344188658057271,
+      "grad_norm": 0.3522341549396515,
+      "learning_rate": 0.00015001968127438872,
+      "loss": 0.5654,
+      "step": 327
+    },
+    {
+      "epoch": 0.7366647950589557,
+      "grad_norm": 0.33905208110809326,
+      "learning_rate": 0.00014991822489887938,
+      "loss": 0.5606,
+      "step": 328
+    },
+    {
+      "epoch": 0.7389107243121842,
+      "grad_norm": 0.29921072721481323,
+      "learning_rate": 0.00014981629010937372,
+      "loss": 0.5772,
+      "step": 329
+    },
+    {
+      "epoch": 0.7411566535654127,
+      "grad_norm": 0.2822812497615814,
+      "learning_rate": 0.00014971387760335841,
+      "loss": 0.5772,
+      "step": 330
+    },
+    {
+      "epoch": 0.7434025828186412,
+      "grad_norm": 0.3244154155254364,
+      "learning_rate": 0.0001496109880815889,
+      "loss": 0.5736,
+      "step": 331
+    },
+    {
+      "epoch": 0.7456485120718698,
+      "grad_norm": 0.3305480480194092,
+      "learning_rate": 0.0001495076222480846,
+      "loss": 0.586,
+      "step": 332
+    },
+    {
+      "epoch": 0.7478944413250983,
+      "grad_norm": 0.3018239140510559,
+      "learning_rate": 0.00014940378081012407,
+      "loss": 0.579,
+      "step": 333
+    },
+    {
+      "epoch": 0.7501403705783268,
+      "grad_norm": 0.3692958652973175,
+      "learning_rate": 0.00014929946447824014,
+      "loss": 0.5767,
+      "step": 334
+    },
+    {
+      "epoch": 0.7523862998315554,
+      "grad_norm": 0.3724178373813629,
+      "learning_rate": 0.00014919467396621523,
+      "loss": 0.5721,
+      "step": 335
+    },
+    {
+      "epoch": 0.7546322290847838,
+      "grad_norm": 0.3226647973060608,
+      "learning_rate": 0.00014908940999107615,
+      "loss": 0.553,
+      "step": 336
+    },
+    {
+      "epoch": 0.7568781583380123,
+      "grad_norm": 0.28518086671829224,
+      "learning_rate": 0.00014898367327308945,
+      "loss": 0.566,
+      "step": 337
+    },
+    {
+      "epoch": 0.7591240875912408,
+      "grad_norm": 0.2642190158367157,
+      "learning_rate": 0.0001488774645357565,
+      "loss": 0.5732,
+      "step": 338
+    },
+    {
+      "epoch": 0.7613700168444694,
+      "grad_norm": 0.2713199555873871,
+      "learning_rate": 0.0001487707845058083,
+      "loss": 0.5679,
+      "step": 339
+    },
+    {
+      "epoch": 0.7636159460976979,
+      "grad_norm": 0.28339532017707825,
+      "learning_rate": 0.00014866363391320076,
+      "loss": 0.5664,
+      "step": 340
+    },
+    {
+      "epoch": 0.7658618753509264,
+      "grad_norm": 0.26976078748703003,
+      "learning_rate": 0.0001485560134911096,
+      "loss": 0.5917,
+      "step": 341
+    },
+    {
+      "epoch": 0.768107804604155,
+      "grad_norm": 0.31055644154548645,
+      "learning_rate": 0.00014844792397592524,
+      "loss": 0.5609,
+      "step": 342
+    },
+    {
+      "epoch": 0.7703537338573835,
+      "grad_norm": 0.28089481592178345,
+      "learning_rate": 0.000148339366107248,
+      "loss": 0.5553,
+      "step": 343
+    },
+    {
+      "epoch": 0.772599663110612,
+      "grad_norm": 0.3059735894203186,
+      "learning_rate": 0.00014823034062788282,
+      "loss": 0.5827,
+      "step": 344
+    },
+    {
+      "epoch": 0.7748455923638405,
+      "grad_norm": 0.3540654480457306,
+      "learning_rate": 0.00014812084828383425,
+      "loss": 0.5417,
+      "step": 345
+    },
+    {
+      "epoch": 0.7770915216170691,
+      "grad_norm": 0.3125968277454376,
+      "learning_rate": 0.0001480108898243014,
+      "loss": 0.5676,
+      "step": 346
+    },
+    {
+      "epoch": 0.7793374508702976,
+      "grad_norm": 0.2534315884113312,
+      "learning_rate": 0.0001479004660016727,
+      "loss": 0.5724,
+      "step": 347
+    },
+    {
+      "epoch": 0.7815833801235261,
+      "grad_norm": 0.30985814332962036,
+      "learning_rate": 0.0001477895775715209,
+      "loss": 0.5682,
+      "step": 348
+    },
+    {
+      "epoch": 0.7838293093767547,
+      "grad_norm": 0.334831178188324,
+      "learning_rate": 0.00014767822529259772,
+      "loss": 0.5653,
+      "step": 349
+    },
+    {
+      "epoch": 0.7860752386299832,
+      "grad_norm": 0.29639920592308044,
+      "learning_rate": 0.00014756640992682883,
+      "loss": 0.5959,
+      "step": 350
+    },
+    {
+      "epoch": 0.7883211678832117,
+      "grad_norm": 0.33278346061706543,
+      "learning_rate": 0.00014745413223930858,
+      "loss": 0.57,
+      "step": 351
+    },
+    {
+      "epoch": 0.7905670971364402,
+      "grad_norm": 0.26434555649757385,
+      "learning_rate": 0.00014734139299829466,
+      "loss": 0.5847,
+      "step": 352
+    },
+    {
+      "epoch": 0.7928130263896688,
+      "grad_norm": 0.295564204454422,
+      "learning_rate": 0.00014722819297520296,
+      "loss": 0.5345,
+      "step": 353
+    },
+    {
+      "epoch": 0.7950589556428973,
+      "grad_norm": 0.32043787837028503,
+      "learning_rate": 0.00014711453294460235,
+      "loss": 0.5751,
+      "step": 354
+    },
+    {
+      "epoch": 0.7973048848961257,
+      "grad_norm": 0.35145339369773865,
+      "learning_rate": 0.00014700041368420914,
+      "loss": 0.5782,
+      "step": 355
+    },
+    {
+      "epoch": 0.7995508141493542,
+      "grad_norm": 0.2663813531398773,
+      "learning_rate": 0.00014688583597488204,
+      "loss": 0.5457,
+      "step": 356
+    },
+    {
+      "epoch": 0.8017967434025828,
+      "grad_norm": 0.3394940197467804,
+      "learning_rate": 0.00014677080060061662,
+      "loss": 0.5669,
+      "step": 357
+    },
+    {
+      "epoch": 0.8040426726558113,
+      "grad_norm": 0.28702473640441895,
+      "learning_rate": 0.00014665530834854002,
+      "loss": 0.5715,
+      "step": 358
+    },
+    {
+      "epoch": 0.8062886019090398,
+      "grad_norm": 0.3419654071331024,
+      "learning_rate": 0.0001465393600089056,
+      "loss": 0.5804,
+      "step": 359
+    },
+    {
+      "epoch": 0.8085345311622684,
+      "grad_norm": 0.35292762517929077,
+      "learning_rate": 0.00014642295637508742,
+      "loss": 0.5666,
+      "step": 360
+    },
+    {
+      "epoch": 0.8107804604154969,
+      "grad_norm": 0.31325843930244446,
+      "learning_rate": 0.00014630609824357494,
+      "loss": 0.5857,
+      "step": 361
+    },
+    {
+      "epoch": 0.8130263896687254,
+      "grad_norm": 0.27262774109840393,
+      "learning_rate": 0.00014618878641396748,
+      "loss": 0.5797,
+      "step": 362
+    },
+    {
+      "epoch": 0.815272318921954,
+      "grad_norm": 0.2780674397945404,
+      "learning_rate": 0.00014607102168896882,
+      "loss": 0.5552,
+      "step": 363
+    },
+    {
+      "epoch": 0.8175182481751825,
+      "grad_norm": 0.2732245922088623,
+      "learning_rate": 0.00014595280487438158,
+      "loss": 0.5716,
+      "step": 364
+    },
+    {
+      "epoch": 0.819764177428411,
+      "grad_norm": 0.33612555265426636,
+      "learning_rate": 0.0001458341367791019,
+      "loss": 0.5756,
+      "step": 365
+    },
+    {
+      "epoch": 0.8220101066816395,
+      "grad_norm": 0.267904669046402,
+      "learning_rate": 0.0001457150182151137,
+      "loss": 0.5694,
+      "step": 366
+    },
+    {
+      "epoch": 0.8242560359348681,
+      "grad_norm": 0.2547987401485443,
+      "learning_rate": 0.0001455954499974833,
+      "loss": 0.5718,
+      "step": 367
+    },
+    {
+      "epoch": 0.8265019651880966,
+      "grad_norm": 0.2813619375228882,
+      "learning_rate": 0.00014547543294435376,
+      "loss": 0.5521,
+      "step": 368
+    },
+    {
+      "epoch": 0.8287478944413251,
+      "grad_norm": 0.2692398428916931,
+      "learning_rate": 0.0001453549678769392,
+      "loss": 0.5644,
+      "step": 369
+    },
+    {
+      "epoch": 0.8309938236945537,
+      "grad_norm": 0.24875199794769287,
+      "learning_rate": 0.0001452340556195194,
+      "loss": 0.5563,
+      "step": 370
+    },
+    {
+      "epoch": 0.8332397529477822,
+      "grad_norm": 0.24863849580287933,
+      "learning_rate": 0.00014511269699943392,
+      "loss": 0.5479,
+      "step": 371
+    },
+    {
+      "epoch": 0.8354856822010107,
+      "grad_norm": 0.2492000311613083,
+      "learning_rate": 0.00014499089284707658,
+      "loss": 0.5742,
+      "step": 372
+    },
+    {
+      "epoch": 0.8377316114542392,
+      "grad_norm": 0.2373623251914978,
+      "learning_rate": 0.0001448686439958898,
+      "loss": 0.5688,
+      "step": 373
+    },
+    {
+      "epoch": 0.8399775407074677,
+      "grad_norm": 0.265248566865921,
+      "learning_rate": 0.00014474595128235876,
+      "loss": 0.5616,
+      "step": 374
+    },
+    {
+      "epoch": 0.8422234699606962,
+      "grad_norm": 0.2871013879776001,
+      "learning_rate": 0.00014462281554600577,
+      "loss": 0.556,
+      "step": 375
+    },
+    {
+      "epoch": 0.8444693992139247,
+      "grad_norm": 0.31418806314468384,
+      "learning_rate": 0.00014449923762938462,
+      "loss": 0.5644,
+      "step": 376
+    },
+    {
+      "epoch": 0.8467153284671532,
+      "grad_norm": 0.3332020044326782,
+      "learning_rate": 0.00014437521837807455,
+      "loss": 0.5611,
+      "step": 377
+    },
+    {
+      "epoch": 0.8489612577203818,
+      "grad_norm": 0.2672823965549469,
+      "learning_rate": 0.00014425075864067473,
+      "loss": 0.5575,
+      "step": 378
+    },
+    {
+      "epoch": 0.8512071869736103,
+      "grad_norm": 0.23632559180259705,
+      "learning_rate": 0.00014412585926879833,
+      "loss": 0.578,
+      "step": 379
+    },
+    {
+      "epoch": 0.8534531162268388,
+      "grad_norm": 0.31967830657958984,
+      "learning_rate": 0.00014400052111706668,
+      "loss": 0.5738,
+      "step": 380
+    },
+    {
+      "epoch": 0.8556990454800674,
+      "grad_norm": 0.3274000287055969,
+      "learning_rate": 0.0001438747450431035,
+      "loss": 0.5606,
+      "step": 381
+    },
+    {
+      "epoch": 0.8579449747332959,
+      "grad_norm": 0.32115650177001953,
+      "learning_rate": 0.00014374853190752892,
+      "loss": 0.601,
+      "step": 382
+    },
+    {
+      "epoch": 0.8601909039865244,
+      "grad_norm": 0.3195722997188568,
+      "learning_rate": 0.00014362188257395367,
+      "loss": 0.5794,
+      "step": 383
+    },
+    {
+      "epoch": 0.862436833239753,
+      "grad_norm": 0.32217174768447876,
+      "learning_rate": 0.00014349479790897325,
+      "loss": 0.5687,
+      "step": 384
+    },
+    {
+      "epoch": 0.8646827624929815,
+      "grad_norm": 0.3338417410850525,
+      "learning_rate": 0.00014336727878216178,
+      "loss": 0.5513,
+      "step": 385
+    },
+    {
+      "epoch": 0.86692869174621,
+      "grad_norm": 0.2939014732837677,
+      "learning_rate": 0.00014323932606606624,
+      "loss": 0.5845,
+      "step": 386
+    },
+    {
+      "epoch": 0.8691746209994385,
+      "grad_norm": 0.34269392490386963,
+      "learning_rate": 0.00014311094063620036,
+      "loss": 0.5721,
+      "step": 387
+    },
+    {
+      "epoch": 0.8714205502526671,
+      "grad_norm": 0.3684992492198944,
+      "learning_rate": 0.00014298212337103888,
+      "loss": 0.5924,
+      "step": 388
+    },
+    {
+      "epoch": 0.8736664795058956,
+      "grad_norm": 0.27671441435813904,
+      "learning_rate": 0.0001428528751520112,
+      "loss": 0.5536,
+      "step": 389
+    },
+    {
+      "epoch": 0.8759124087591241,
+      "grad_norm": 0.3508271276950836,
+      "learning_rate": 0.0001427231968634955,
+      "loss": 0.5499,
+      "step": 390
+    },
+    {
+      "epoch": 0.8781583380123527,
+      "grad_norm": 0.3735405504703522,
+      "learning_rate": 0.00014259308939281292,
+      "loss": 0.5472,
+      "step": 391
+    },
+    {
+      "epoch": 0.8804042672655812,
+      "grad_norm": 0.30313754081726074,
+      "learning_rate": 0.00014246255363022095,
+      "loss": 0.5598,
+      "step": 392
+    },
+    {
+      "epoch": 0.8826501965188096,
+      "grad_norm": 0.28613924980163574,
+      "learning_rate": 0.00014233159046890792,
+      "loss": 0.5589,
+      "step": 393
+    },
+    {
+      "epoch": 0.8848961257720381,
+      "grad_norm": 0.3396552503108978,
+      "learning_rate": 0.00014220020080498648,
+      "loss": 0.5722,
+      "step": 394
+    },
+    {
+      "epoch": 0.8871420550252667,
+      "grad_norm": 0.24562208354473114,
+      "learning_rate": 0.00014206838553748773,
+      "loss": 0.5617,
+      "step": 395
+    },
+    {
+      "epoch": 0.8893879842784952,
+      "grad_norm": 0.26731806993484497,
+      "learning_rate": 0.00014193614556835482,
+      "loss": 0.5876,
+      "step": 396
+    },
+    {
+      "epoch": 0.8916339135317237,
+      "grad_norm": 0.30024391412734985,
+      "learning_rate": 0.00014180348180243706,
+      "loss": 0.5457,
+      "step": 397
+    },
+    {
+      "epoch": 0.8938798427849522,
+      "grad_norm": 0.23074807226657867,
+      "learning_rate": 0.0001416703951474834,
+      "loss": 0.5767,
+      "step": 398
+    },
+    {
+      "epoch": 0.8961257720381808,
+      "grad_norm": 0.2882399260997772,
+      "learning_rate": 0.00014153688651413662,
+      "loss": 0.548,
+      "step": 399
+    },
+    {
+      "epoch": 0.8983717012914093,
+      "grad_norm": 0.30070793628692627,
+      "learning_rate": 0.00014140295681592667,
+      "loss": 0.5483,
+      "step": 400
+    }
+  ],
+  "logging_steps": 1,
+  "max_steps": 1335,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 3,
+  "save_steps": 100,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.2601562302777393e+19,
+  "train_batch_size": 16,
+  "trial_name": null,
+  "trial_params": null
+}

checkpoint-400/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:19636a092c5e7f0fefdaadef77b19806c9e92329af56356597f8083c85935b8f
+size 7288

checkpoint-400/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)