Upload folder using huggingface_hub

.gitattributes

@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+limo/full/checkpoint-515/tokenizer.json filter=lfs diff=lfs merge=lfs -text

limo/full/checkpoint-515/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
+}

limo/full/checkpoint-515/config.json

@@ -0,0 +1,29 @@
+{
+  "_name_or_path": "/mnt/data/lh/models/Qwen2.5-Math-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": 4096,
+  "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": 10000.0,
+  "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
+}

limo/full/checkpoint-515/generation_config.json

@@ -0,0 +1,9 @@
+{
+  "bos_token_id": 151643,
+  "eos_token_id": [
+    151645,
+    151643
+  ],
+  "pad_token_id": 151643,
+  "transformers_version": "4.46.1"
+}

limo/full/checkpoint-515/global_step515/bf16_zero_pp_rank_0_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:29a302c2725f0851f4eb31be81376bdf4ae54995efba31a29f9d9e70b5f24463
+size 2315576560

limo/full/checkpoint-515/global_step515/bf16_zero_pp_rank_1_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:357a92d278608d945ff299229f5c7d03bead7ebd0c197be60a95a8dd64a961d2
+size 2315576560

limo/full/checkpoint-515/global_step515/bf16_zero_pp_rank_2_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:73d27db6368186a0c5dfe82b9434386aabbf624a443c3755db79b025edd743ae
+size 2315576560

limo/full/checkpoint-515/global_step515/bf16_zero_pp_rank_3_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:19c80547307b589e3a5a8859ead58ea3acfc7005a1a1ab3f8878e12fc0f038db
+size 2315576560

limo/full/checkpoint-515/global_step515/bf16_zero_pp_rank_4_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:24277a1953a5caee47dedb6a56e83ef310c8b86529e777279397f84ccd81500a
+size 2315576560

limo/full/checkpoint-515/global_step515/bf16_zero_pp_rank_5_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c1e76fbd9d9c5f2b8cbc9feecd78d470de47c0604cd383676b6d63b5bdf7da30
+size 2315576560

limo/full/checkpoint-515/global_step515/bf16_zero_pp_rank_6_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:01a3045f70cfac382d7b6fd1430729276da62a06e7c1698b6d624748f426d601
+size 2315576560

limo/full/checkpoint-515/global_step515/bf16_zero_pp_rank_7_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:65734d35f371bc0eae5ed749a276acc569769e313dfcf515b62bf6104f5eb521
+size 2315576560

limo/full/checkpoint-515/global_step515/zero_pp_rank_0_mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:233d2c55f6a05790591d35e2c3182f2a7dd9b233cdd34e3be09d0dd00334edaa
+size 168056

limo/full/checkpoint-515/global_step515/zero_pp_rank_1_mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:67ecc3222ec51d4f0d5a6b5cf3eee4a27ff807448cf9bdbf8beca5c5a3af561e
+size 167992

limo/full/checkpoint-515/global_step515/zero_pp_rank_2_mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e713c3836ab9ebcf4e97b4d1c916d53cf7a96228dde30d3b2fb9bc087130a591
+size 167992

limo/full/checkpoint-515/global_step515/zero_pp_rank_3_mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:71826b62a7f809a2ef7dae3ce3bb667fbc86097815e6509a3b51597ce660ba53
+size 167992

limo/full/checkpoint-515/global_step515/zero_pp_rank_4_mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7ebfee55c35aab284b06910683f5bb75849632814b5f950da8c2939db3fa971e
+size 167992

limo/full/checkpoint-515/global_step515/zero_pp_rank_5_mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d637230497529e940ea2b60e16a856351f0b305cb781418fa50d752a36dfa6d6
+size 167992

limo/full/checkpoint-515/global_step515/zero_pp_rank_6_mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6d9429a755e61bf6d69b48109f01187b5627eda2d699054fc56a159a69a32bad
+size 167992

limo/full/checkpoint-515/global_step515/zero_pp_rank_7_mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:15351deb28f9010d42255092d71edc2198ba7c01c7f193f4d3a7fd8ff38f6ca3
+size 167992

limo/full/checkpoint-515/latest

@@ -0,0 +1 @@
+global_step515

limo/full/checkpoint-515/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9311ae95d5141d9d1768fc11ed07a687db96b7b6acbcbd4b5b4ff5f0b0c6dccf
+size 3087467144

limo/full/checkpoint-515/rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5a0ef6f96a48e59aa52c4b471312c2a62378c19acc7ebbae839612b03a7d775a
+size 15984

limo/full/checkpoint-515/rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ab11d533c0fdad46ea8b8e295ba5fdb705e078eeb88cc28f37d82913508766e9
+size 15984

limo/full/checkpoint-515/rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:615c168147e3465ce5bfab6da2ff4afc68566ce00ec0f0c6c9fc988038a58d0a
+size 15984

limo/full/checkpoint-515/rng_state_3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:79f71e8f8674ecaef9f8cdcbf7ac457a8b8ff15b12694ba2a2fffcb4b43f0f08
+size 15984

limo/full/checkpoint-515/rng_state_4.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:88cf6d674dab5545c300a55135f08ca935730a3d35e2c419fb0b333f19482c19
+size 15984

limo/full/checkpoint-515/rng_state_5.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2754f2cd8824702f027870d93748b3c0491b0ecd30f1e3d8e937116b2be6151f
+size 15984

limo/full/checkpoint-515/rng_state_6.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1385124ac55604598f45ea6e2d141f29456647d3e7c10d12ca64ec93d312be8d
+size 15984

limo/full/checkpoint-515/rng_state_7.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:416538efaec7391fa8fe782fb15146b83e5612d9e1961292c34c53e964806873
+size 15984

limo/full/checkpoint-515/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e5d0919066ee6a14427d7be219c870467c623faeab11aea9e226fd92a7435d45
+size 1064

limo/full/checkpoint-515/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": "<|im_end|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

limo/full/checkpoint-515/tokenizer.json

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

limo/full/checkpoint-515/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        {{- 'Please reason step by step, and put your final answer within \\\\boxed{}.' }}\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\\nPlease reason step by step, and put your final answer within \\\\boxed{}.<|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": "<|im_end|>",
+  "errors": "replace",
+  "model_max_length": 16384,
+  "pad_token": "<|endoftext|>",
+  "padding_side": "right",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}

limo/full/checkpoint-515/trainer_state.json

@@ -0,0 +1,3638 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 5.0,
+  "eval_steps": 500,
+  "global_step": 515,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.009708737864077669,
+      "grad_norm": 4.871730834654797,
+      "learning_rate": 4.999994831641374e-06,
+      "loss": 1.8478,
+      "step": 1
+    },
+    {
+      "epoch": 0.019417475728155338,
+      "grad_norm": 3.669459516332945,
+      "learning_rate": 4.9999793265868636e-06,
+      "loss": 1.4574,
+      "step": 2
+    },
+    {
+      "epoch": 0.02912621359223301,
+      "grad_norm": 3.321268843922114,
+      "learning_rate": 4.999953484900578e-06,
+      "loss": 1.5713,
+      "step": 3
+    },
+    {
+      "epoch": 0.038834951456310676,
+      "grad_norm": 4.757625498508145,
+      "learning_rate": 4.9999173066893655e-06,
+      "loss": 2.1849,
+      "step": 4
+    },
+    {
+      "epoch": 0.04854368932038835,
+      "grad_norm": 4.484235162199009,
+      "learning_rate": 4.9998707921028104e-06,
+      "loss": 2.538,
+      "step": 5
+    },
+    {
+      "epoch": 0.05825242718446602,
+      "grad_norm": 3.0105096661858055,
+      "learning_rate": 4.999813941333237e-06,
+      "loss": 1.5868,
+      "step": 6
+    },
+    {
+      "epoch": 0.06796116504854369,
+      "grad_norm": 2.8411927476391083,
+      "learning_rate": 4.999746754615704e-06,
+      "loss": 1.6472,
+      "step": 7
+    },
+    {
+      "epoch": 0.07766990291262135,
+      "grad_norm": 1.506880067190551,
+      "learning_rate": 4.9996692322280085e-06,
+      "loss": 1.0478,
+      "step": 8
+    },
+    {
+      "epoch": 0.08737864077669903,
+      "grad_norm": 4.0008731506353445,
+      "learning_rate": 4.999581374490681e-06,
+      "loss": 2.373,
+      "step": 9
+    },
+    {
+      "epoch": 0.0970873786407767,
+      "grad_norm": 2.4382442959213715,
+      "learning_rate": 4.999483181766986e-06,
+      "loss": 1.7706,
+      "step": 10
+    },
+    {
+      "epoch": 0.10679611650485436,
+      "grad_norm": 1.6052608591178483,
+      "learning_rate": 4.999374654462919e-06,
+      "loss": 1.3104,
+      "step": 11
+    },
+    {
+      "epoch": 0.11650485436893204,
+      "grad_norm": 2.7437905836581296,
+      "learning_rate": 4.999255793027207e-06,
+      "loss": 2.048,
+      "step": 12
+    },
+    {
+      "epoch": 0.1262135922330097,
+      "grad_norm": 1.3906335136930252,
+      "learning_rate": 4.999126597951305e-06,
+      "loss": 1.0066,
+      "step": 13
+    },
+    {
+      "epoch": 0.13592233009708737,
+      "grad_norm": 1.3340600696507865,
+      "learning_rate": 4.998987069769394e-06,
+      "loss": 1.0998,
+      "step": 14
+    },
+    {
+      "epoch": 0.14563106796116504,
+      "grad_norm": 1.7130835123840018,
+      "learning_rate": 4.998837209058379e-06,
+      "loss": 1.6654,
+      "step": 15
+    },
+    {
+      "epoch": 0.1553398058252427,
+      "grad_norm": 1.7208415743567615,
+      "learning_rate": 4.998677016437888e-06,
+      "loss": 1.3388,
+      "step": 16
+    },
+    {
+      "epoch": 0.1650485436893204,
+      "grad_norm": 2.4073143065161937,
+      "learning_rate": 4.998506492570266e-06,
+      "loss": 1.8365,
+      "step": 17
+    },
+    {
+      "epoch": 0.17475728155339806,
+      "grad_norm": 1.4448713906198571,
+      "learning_rate": 4.998325638160576e-06,
+      "loss": 1.4415,
+      "step": 18
+    },
+    {
+      "epoch": 0.18446601941747573,
+      "grad_norm": 1.5964597845558501,
+      "learning_rate": 4.998134453956596e-06,
+      "loss": 1.1629,
+      "step": 19
+    },
+    {
+      "epoch": 0.1941747572815534,
+      "grad_norm": 1.3355139686727637,
+      "learning_rate": 4.997932940748811e-06,
+      "loss": 1.2245,
+      "step": 20
+    },
+    {
+      "epoch": 0.20388349514563106,
+      "grad_norm": 1.2386749477462127,
+      "learning_rate": 4.997721099370416e-06,
+      "loss": 1.2827,
+      "step": 21
+    },
+    {
+      "epoch": 0.21359223300970873,
+      "grad_norm": 1.3520067333735615,
+      "learning_rate": 4.997498930697308e-06,
+      "loss": 1.3963,
+      "step": 22
+    },
+    {
+      "epoch": 0.22330097087378642,
+      "grad_norm": 1.7137484637690432,
+      "learning_rate": 4.997266435648086e-06,
+      "loss": 1.4022,
+      "step": 23
+    },
+    {
+      "epoch": 0.23300970873786409,
+      "grad_norm": 1.9868889107654797,
+      "learning_rate": 4.997023615184044e-06,
+      "loss": 1.7246,
+      "step": 24
+    },
+    {
+      "epoch": 0.24271844660194175,
+      "grad_norm": 1.6032606855000195,
+      "learning_rate": 4.996770470309167e-06,
+      "loss": 1.7975,
+      "step": 25
+    },
+    {
+      "epoch": 0.2524271844660194,
+      "grad_norm": 1.268066436948988,
+      "learning_rate": 4.996507002070131e-06,
+      "loss": 1.3028,
+      "step": 26
+    },
+    {
+      "epoch": 0.2621359223300971,
+      "grad_norm": 1.2303902485186446,
+      "learning_rate": 4.996233211556295e-06,
+      "loss": 1.3613,
+      "step": 27
+    },
+    {
+      "epoch": 0.27184466019417475,
+      "grad_norm": 0.9549363286063691,
+      "learning_rate": 4.9959490998996974e-06,
+      "loss": 0.9778,
+      "step": 28
+    },
+    {
+      "epoch": 0.2815533980582524,
+      "grad_norm": 1.2603575097056405,
+      "learning_rate": 4.995654668275049e-06,
+      "loss": 1.4077,
+      "step": 29
+    },
+    {
+      "epoch": 0.2912621359223301,
+      "grad_norm": 1.5075543470098034,
+      "learning_rate": 4.995349917899735e-06,
+      "loss": 1.5452,
+      "step": 30
+    },
+    {
+      "epoch": 0.30097087378640774,
+      "grad_norm": 0.9482722379956937,
+      "learning_rate": 4.9950348500338005e-06,
+      "loss": 1.0589,
+      "step": 31
+    },
+    {
+      "epoch": 0.3106796116504854,
+      "grad_norm": 0.8370896925167312,
+      "learning_rate": 4.994709465979954e-06,
+      "loss": 0.977,
+      "step": 32
+    },
+    {
+      "epoch": 0.32038834951456313,
+      "grad_norm": 1.204641294852316,
+      "learning_rate": 4.994373767083556e-06,
+      "loss": 1.2391,
+      "step": 33
+    },
+    {
+      "epoch": 0.3300970873786408,
+      "grad_norm": 1.1272285807079887,
+      "learning_rate": 4.994027754732616e-06,
+      "loss": 1.2678,
+      "step": 34
+    },
+    {
+      "epoch": 0.33980582524271846,
+      "grad_norm": 1.0240110484845462,
+      "learning_rate": 4.993671430357788e-06,
+      "loss": 1.1705,
+      "step": 35
+    },
+    {
+      "epoch": 0.34951456310679613,
+      "grad_norm": 1.3831448718921984,
+      "learning_rate": 4.99330479543236e-06,
+      "loss": 1.6517,
+      "step": 36
+    },
+    {
+      "epoch": 0.3592233009708738,
+      "grad_norm": 1.0046287010394588,
+      "learning_rate": 4.992927851472254e-06,
+      "loss": 1.2714,
+      "step": 37
+    },
+    {
+      "epoch": 0.36893203883495146,
+      "grad_norm": 1.3527812004094968,
+      "learning_rate": 4.992540600036014e-06,
+      "loss": 1.6523,
+      "step": 38
+    },
+    {
+      "epoch": 0.3786407766990291,
+      "grad_norm": 1.162751011119127,
+      "learning_rate": 4.992143042724805e-06,
+      "loss": 1.4164,
+      "step": 39
+    },
+    {
+      "epoch": 0.3883495145631068,
+      "grad_norm": 1.183212466699315,
+      "learning_rate": 4.991735181182401e-06,
+      "loss": 1.6037,
+      "step": 40
+    },
+    {
+      "epoch": 0.39805825242718446,
+      "grad_norm": 1.0671445151315921,
+      "learning_rate": 4.991317017095182e-06,
+      "loss": 1.2074,
+      "step": 41
+    },
+    {
+      "epoch": 0.4077669902912621,
+      "grad_norm": 1.1374295195195832,
+      "learning_rate": 4.990888552192126e-06,
+      "loss": 1.2052,
+      "step": 42
+    },
+    {
+      "epoch": 0.4174757281553398,
+      "grad_norm": 1.1838914278145132,
+      "learning_rate": 4.9904497882448004e-06,
+      "loss": 1.3772,
+      "step": 43
+    },
+    {
+      "epoch": 0.42718446601941745,
+      "grad_norm": 1.1898023150727208,
+      "learning_rate": 4.990000727067357e-06,
+      "loss": 1.4616,
+      "step": 44
+    },
+    {
+      "epoch": 0.4368932038834951,
+      "grad_norm": 1.3045045420389443,
+      "learning_rate": 4.989541370516523e-06,
+      "loss": 1.2714,
+      "step": 45
+    },
+    {
+      "epoch": 0.44660194174757284,
+      "grad_norm": 1.1641307331044313,
+      "learning_rate": 4.989071720491595e-06,
+      "loss": 1.2003,
+      "step": 46
+    },
+    {
+      "epoch": 0.4563106796116505,
+      "grad_norm": 0.6576132196355234,
+      "learning_rate": 4.988591778934428e-06,
+      "loss": 0.791,
+      "step": 47
+    },
+    {
+      "epoch": 0.46601941747572817,
+      "grad_norm": 0.8500656663270045,
+      "learning_rate": 4.9881015478294294e-06,
+      "loss": 0.9369,
+      "step": 48
+    },
+    {
+      "epoch": 0.47572815533980584,
+      "grad_norm": 1.4233952816867905,
+      "learning_rate": 4.987601029203553e-06,
+      "loss": 1.4027,
+      "step": 49
+    },
+    {
+      "epoch": 0.4854368932038835,
+      "grad_norm": 1.1258574635846645,
+      "learning_rate": 4.987090225126285e-06,
+      "loss": 1.3511,
+      "step": 50
+    },
+    {
+      "epoch": 0.49514563106796117,
+      "grad_norm": 1.1687633946081721,
+      "learning_rate": 4.98656913770964e-06,
+      "loss": 1.1697,
+      "step": 51
+    },
+    {
+      "epoch": 0.5048543689320388,
+      "grad_norm": 1.2317336138068717,
+      "learning_rate": 4.986037769108154e-06,
+      "loss": 1.4503,
+      "step": 52
+    },
+    {
+      "epoch": 0.5145631067961165,
+      "grad_norm": 0.7574877986414429,
+      "learning_rate": 4.9854961215188676e-06,
+      "loss": 1.1225,
+      "step": 53
+    },
+    {
+      "epoch": 0.5242718446601942,
+      "grad_norm": 0.8037978880160027,
+      "learning_rate": 4.984944197181324e-06,
+      "loss": 0.9828,
+      "step": 54
+    },
+    {
+      "epoch": 0.5339805825242718,
+      "grad_norm": 1.1494946115521347,
+      "learning_rate": 4.9843819983775575e-06,
+      "loss": 1.2705,
+      "step": 55
+    },
+    {
+      "epoch": 0.5436893203883495,
+      "grad_norm": 1.0404784765183215,
+      "learning_rate": 4.983809527432086e-06,
+      "loss": 1.1189,
+      "step": 56
+    },
+    {
+      "epoch": 0.5533980582524272,
+      "grad_norm": 1.0072279484535855,
+      "learning_rate": 4.983226786711895e-06,
+      "loss": 1.2205,
+      "step": 57
+    },
+    {
+      "epoch": 0.5631067961165048,
+      "grad_norm": 0.9065921594118113,
+      "learning_rate": 4.982633778626437e-06,
+      "loss": 1.07,
+      "step": 58
+    },
+    {
+      "epoch": 0.5728155339805825,
+      "grad_norm": 1.2370148909491399,
+      "learning_rate": 4.982030505627613e-06,
+      "loss": 1.535,
+      "step": 59
+    },
+    {
+      "epoch": 0.5825242718446602,
+      "grad_norm": 0.7862714572091344,
+      "learning_rate": 4.98141697020977e-06,
+      "loss": 1.0275,
+      "step": 60
+    },
+    {
+      "epoch": 0.5922330097087378,
+      "grad_norm": 1.0179109472315906,
+      "learning_rate": 4.9807931749096836e-06,
+      "loss": 1.1225,
+      "step": 61
+    },
+    {
+      "epoch": 0.6019417475728155,
+      "grad_norm": 0.6647718819463199,
+      "learning_rate": 4.980159122306551e-06,
+      "loss": 0.9871,
+      "step": 62
+    },
+    {
+      "epoch": 0.6116504854368932,
+      "grad_norm": 0.7487245734536395,
+      "learning_rate": 4.979514815021984e-06,
+      "loss": 1.2099,
+      "step": 63
+    },
+    {
+      "epoch": 0.6213592233009708,
+      "grad_norm": 1.1029648607720772,
+      "learning_rate": 4.978860255719989e-06,
+      "loss": 1.0207,
+      "step": 64
+    },
+    {
+      "epoch": 0.6310679611650486,
+      "grad_norm": 0.7621265063758594,
+      "learning_rate": 4.978195447106965e-06,
+      "loss": 1.1363,
+      "step": 65
+    },
+    {
+      "epoch": 0.6407766990291263,
+      "grad_norm": 2.313544656936807,
+      "learning_rate": 4.9775203919316864e-06,
+      "loss": 0.9453,
+      "step": 66
+    },
+    {
+      "epoch": 0.6504854368932039,
+      "grad_norm": 0.8280847399908062,
+      "learning_rate": 4.976835092985297e-06,
+      "loss": 1.1131,
+      "step": 67
+    },
+    {
+      "epoch": 0.6601941747572816,
+      "grad_norm": 0.8927638160250211,
+      "learning_rate": 4.976139553101291e-06,
+      "loss": 1.133,
+      "step": 68
+    },
+    {
+      "epoch": 0.6699029126213593,
+      "grad_norm": 0.9610176243493397,
+      "learning_rate": 4.975433775155509e-06,
+      "loss": 1.1768,
+      "step": 69
+    },
+    {
+      "epoch": 0.6796116504854369,
+      "grad_norm": 1.194498620463663,
+      "learning_rate": 4.974717762066123e-06,
+      "loss": 1.3615,
+      "step": 70
+    },
+    {
+      "epoch": 0.6893203883495146,
+      "grad_norm": 0.8347048742413943,
+      "learning_rate": 4.973991516793621e-06,
+      "loss": 1.2546,
+      "step": 71
+    },
+    {
+      "epoch": 0.6990291262135923,
+      "grad_norm": 0.9172122226894214,
+      "learning_rate": 4.973255042340801e-06,
+      "loss": 1.1607,
+      "step": 72
+    },
+    {
+      "epoch": 0.7087378640776699,
+      "grad_norm": 1.03327431376727,
+      "learning_rate": 4.972508341752754e-06,
+      "loss": 1.361,
+      "step": 73
+    },
+    {
+      "epoch": 0.7184466019417476,
+      "grad_norm": 1.3504753816166224,
+      "learning_rate": 4.9717514181168534e-06,
+      "loss": 0.9757,
+      "step": 74
+    },
+    {
+      "epoch": 0.7281553398058253,
+      "grad_norm": 1.002604662508283,
+      "learning_rate": 4.970984274562741e-06,
+      "loss": 1.145,
+      "step": 75
+    },
+    {
+      "epoch": 0.7378640776699029,
+      "grad_norm": 0.8521212437833093,
+      "learning_rate": 4.970206914262315e-06,
+      "loss": 1.0841,
+      "step": 76
+    },
+    {
+      "epoch": 0.7475728155339806,
+      "grad_norm": 0.7268667125506638,
+      "learning_rate": 4.969419340429717e-06,
+      "loss": 1.0124,
+      "step": 77
+    },
+    {
+      "epoch": 0.7572815533980582,
+      "grad_norm": 1.0230017430471454,
+      "learning_rate": 4.968621556321319e-06,
+      "loss": 1.4082,
+      "step": 78
+    },
+    {
+      "epoch": 0.7669902912621359,
+      "grad_norm": 0.8075925843214921,
+      "learning_rate": 4.967813565235708e-06,
+      "loss": 0.9341,
+      "step": 79
+    },
+    {
+      "epoch": 0.7766990291262136,
+      "grad_norm": 0.7417143914647092,
+      "learning_rate": 4.966995370513675e-06,
+      "loss": 0.9597,
+      "step": 80
+    },
+    {
+      "epoch": 0.7864077669902912,
+      "grad_norm": 0.9927625054735497,
+      "learning_rate": 4.966166975538197e-06,
+      "loss": 1.298,
+      "step": 81
+    },
+    {
+      "epoch": 0.7961165048543689,
+      "grad_norm": 0.6918994370607547,
+      "learning_rate": 4.965328383734429e-06,
+      "loss": 0.9559,
+      "step": 82
+    },
+    {
+      "epoch": 0.8058252427184466,
+      "grad_norm": 0.9084696462282987,
+      "learning_rate": 4.964479598569686e-06,
+      "loss": 1.3896,
+      "step": 83
+    },
+    {
+      "epoch": 0.8155339805825242,
+      "grad_norm": 0.9240522865451553,
+      "learning_rate": 4.963620623553428e-06,
+      "loss": 1.2391,
+      "step": 84
+    },
+    {
+      "epoch": 0.8252427184466019,
+      "grad_norm": 0.9801285757696232,
+      "learning_rate": 4.962751462237248e-06,
+      "loss": 0.9936,
+      "step": 85
+    },
+    {
+      "epoch": 0.8349514563106796,
+      "grad_norm": 0.7316677725651993,
+      "learning_rate": 4.9618721182148564e-06,
+      "loss": 0.8096,
+      "step": 86
+    },
+    {
+      "epoch": 0.8446601941747572,
+      "grad_norm": 0.7986070833021874,
+      "learning_rate": 4.960982595122064e-06,
+      "loss": 0.9936,
+      "step": 87
+    },
+    {
+      "epoch": 0.8543689320388349,
+      "grad_norm": 2.991170115347129,
+      "learning_rate": 4.960082896636773e-06,
+      "loss": 1.2219,
+      "step": 88
+    },
+    {
+      "epoch": 0.8640776699029126,
+      "grad_norm": 0.841666379855367,
+      "learning_rate": 4.959173026478952e-06,
+      "loss": 1.0418,
+      "step": 89
+    },
+    {
+      "epoch": 0.8737864077669902,
+      "grad_norm": 0.8276404717553167,
+      "learning_rate": 4.958252988410631e-06,
+      "loss": 0.9154,
+      "step": 90
+    },
+    {
+      "epoch": 0.883495145631068,
+      "grad_norm": 0.9617539349436555,
+      "learning_rate": 4.9573227862358794e-06,
+      "loss": 1.1546,
+      "step": 91
+    },
+    {
+      "epoch": 0.8932038834951457,
+      "grad_norm": 0.9410894322853203,
+      "learning_rate": 4.956382423800791e-06,
+      "loss": 1.2453,
+      "step": 92
+    },
+    {
+      "epoch": 0.9029126213592233,
+      "grad_norm": 0.7461574178337397,
+      "learning_rate": 4.955431904993471e-06,
+      "loss": 1.0039,
+      "step": 93
+    },
+    {
+      "epoch": 0.912621359223301,
+      "grad_norm": 0.7787771559636139,
+      "learning_rate": 4.954471233744015e-06,
+      "loss": 0.7834,
+      "step": 94
+    },
+    {
+      "epoch": 0.9223300970873787,
+      "grad_norm": 0.8319061238375747,
+      "learning_rate": 4.9535004140245005e-06,
+      "loss": 0.9597,
+      "step": 95
+    },
+    {
+      "epoch": 0.9320388349514563,
+      "grad_norm": 1.241881123747875,
+      "learning_rate": 4.952519449848962e-06,
+      "loss": 1.2155,
+      "step": 96
+    },
+    {
+      "epoch": 0.941747572815534,
+      "grad_norm": 0.9416478821490492,
+      "learning_rate": 4.951528345273379e-06,
+      "loss": 0.9302,
+      "step": 97
+    },
+    {
+      "epoch": 0.9514563106796117,
+      "grad_norm": 0.6462620680684062,
+      "learning_rate": 4.950527104395659e-06,
+      "loss": 0.9483,
+      "step": 98
+    },
+    {
+      "epoch": 0.9611650485436893,
+      "grad_norm": 1.3945318287905581,
+      "learning_rate": 4.9495157313556185e-06,
+      "loss": 0.884,
+      "step": 99
+    },
+    {
+      "epoch": 0.970873786407767,
+      "grad_norm": 0.6127133377291775,
+      "learning_rate": 4.94849423033497e-06,
+      "loss": 0.6754,
+      "step": 100
+    },
+    {
+      "epoch": 0.9805825242718447,
+      "grad_norm": 0.8204199917952882,
+      "learning_rate": 4.9474626055573e-06,
+      "loss": 1.1493,
+      "step": 101
+    },
+    {
+      "epoch": 0.9902912621359223,
+      "grad_norm": 0.6412217993750089,
+      "learning_rate": 4.946420861288051e-06,
+      "loss": 0.9205,
+      "step": 102
+    },
+    {
+      "epoch": 1.0,
+      "grad_norm": 0.7254619944280207,
+      "learning_rate": 4.9453690018345144e-06,
+      "loss": 0.9894,
+      "step": 103
+    },
+    {
+      "epoch": 1.0097087378640777,
+      "grad_norm": 0.8838225821340072,
+      "learning_rate": 4.944307031545797e-06,
+      "loss": 1.0139,
+      "step": 104
+    },
+    {
+      "epoch": 1.0194174757281553,
+      "grad_norm": 0.8681724674727661,
+      "learning_rate": 4.943234954812812e-06,
+      "loss": 1.2911,
+      "step": 105
+    },
+    {
+      "epoch": 1.029126213592233,
+      "grad_norm": 0.6707492146981981,
+      "learning_rate": 4.942152776068264e-06,
+      "loss": 0.8712,
+      "step": 106
+    },
+    {
+      "epoch": 1.0388349514563107,
+      "grad_norm": 0.8258075866719828,
+      "learning_rate": 4.941060499786622e-06,
+      "loss": 1.2535,
+      "step": 107
+    },
+    {
+      "epoch": 1.0485436893203883,
+      "grad_norm": 0.9343039880211657,
+      "learning_rate": 4.939958130484106e-06,
+      "loss": 1.0248,
+      "step": 108
+    },
+    {
+      "epoch": 1.058252427184466,
+      "grad_norm": 0.6215772104671746,
+      "learning_rate": 4.938845672718668e-06,
+      "loss": 0.899,
+      "step": 109
+    },
+    {
+      "epoch": 1.0679611650485437,
+      "grad_norm": 0.6106502770977033,
+      "learning_rate": 4.937723131089974e-06,
+      "loss": 0.8534,
+      "step": 110
+    },
+    {
+      "epoch": 1.0776699029126213,
+      "grad_norm": 0.8527770514681675,
+      "learning_rate": 4.93659051023938e-06,
+      "loss": 1.0798,
+      "step": 111
+    },
+    {
+      "epoch": 1.087378640776699,
+      "grad_norm": 0.6013648488199381,
+      "learning_rate": 4.93544781484992e-06,
+      "loss": 0.7126,
+      "step": 112
+    },
+    {
+      "epoch": 1.0970873786407767,
+      "grad_norm": 0.673264040200857,
+      "learning_rate": 4.9342950496462815e-06,
+      "loss": 1.0433,
+      "step": 113
+    },
+    {
+      "epoch": 1.1067961165048543,
+      "grad_norm": 0.6367668762270327,
+      "learning_rate": 4.933132219394786e-06,
+      "loss": 0.8586,
+      "step": 114
+    },
+    {
+      "epoch": 1.116504854368932,
+      "grad_norm": 0.7326471152183239,
+      "learning_rate": 4.931959328903376e-06,
+      "loss": 1.1661,
+      "step": 115
+    },
+    {
+      "epoch": 1.1262135922330097,
+      "grad_norm": 0.7905138029016624,
+      "learning_rate": 4.930776383021584e-06,
+      "loss": 0.9155,
+      "step": 116
+    },
+    {
+      "epoch": 1.1359223300970873,
+      "grad_norm": 0.7113840452485917,
+      "learning_rate": 4.92958338664052e-06,
+      "loss": 0.7101,
+      "step": 117
+    },
+    {
+      "epoch": 1.145631067961165,
+      "grad_norm": 0.7085441798426322,
+      "learning_rate": 4.928380344692853e-06,
+      "loss": 0.9499,
+      "step": 118
+    },
+    {
+      "epoch": 1.1553398058252426,
+      "grad_norm": 0.6654689711573367,
+      "learning_rate": 4.927167262152784e-06,
+      "loss": 0.9453,
+      "step": 119
+    },
+    {
+      "epoch": 1.1650485436893203,
+      "grad_norm": 0.7093345141752957,
+      "learning_rate": 4.925944144036027e-06,
+      "loss": 1.0183,
+      "step": 120
+    },
+    {
+      "epoch": 1.174757281553398,
+      "grad_norm": 0.7761162990313378,
+      "learning_rate": 4.924710995399796e-06,
+      "loss": 0.9277,
+      "step": 121
+    },
+    {
+      "epoch": 1.1844660194174756,
+      "grad_norm": 0.8690880735001523,
+      "learning_rate": 4.923467821342773e-06,
+      "loss": 1.1022,
+      "step": 122
+    },
+    {
+      "epoch": 1.1941747572815533,
+      "grad_norm": 0.8588202527080991,
+      "learning_rate": 4.922214627005092e-06,
+      "loss": 1.0451,
+      "step": 123
+    },
+    {
+      "epoch": 1.203883495145631,
+      "grad_norm": 0.6415200336267318,
+      "learning_rate": 4.920951417568323e-06,
+      "loss": 1.004,
+      "step": 124
+    },
+    {
+      "epoch": 1.2135922330097086,
+      "grad_norm": 0.796282956613202,
+      "learning_rate": 4.919678198255438e-06,
+      "loss": 0.9759,
+      "step": 125
+    },
+    {
+      "epoch": 1.2233009708737863,
+      "grad_norm": 0.7236053821892523,
+      "learning_rate": 4.918394974330801e-06,
+      "loss": 1.0669,
+      "step": 126
+    },
+    {
+      "epoch": 1.233009708737864,
+      "grad_norm": 0.7465165265947691,
+      "learning_rate": 4.917101751100142e-06,
+      "loss": 1.0007,
+      "step": 127
+    },
+    {
+      "epoch": 1.2427184466019416,
+      "grad_norm": 0.6913202728148214,
+      "learning_rate": 4.915798533910534e-06,
+      "loss": 1.037,
+      "step": 128
+    },
+    {
+      "epoch": 1.2524271844660193,
+      "grad_norm": 0.6269818244555071,
+      "learning_rate": 4.9144853281503715e-06,
+      "loss": 0.7153,
+      "step": 129
+    },
+    {
+      "epoch": 1.262135922330097,
+      "grad_norm": 0.7329901927771046,
+      "learning_rate": 4.91316213924935e-06,
+      "loss": 1.0037,
+      "step": 130
+    },
+    {
+      "epoch": 1.2718446601941746,
+      "grad_norm": 0.9047110334486224,
+      "learning_rate": 4.911828972678441e-06,
+      "loss": 0.9842,
+      "step": 131
+    },
+    {
+      "epoch": 1.2815533980582523,
+      "grad_norm": 0.7849365731256595,
+      "learning_rate": 4.91048583394987e-06,
+      "loss": 0.8823,
+      "step": 132
+    },
+    {
+      "epoch": 1.29126213592233,
+      "grad_norm": 0.8276108169878443,
+      "learning_rate": 4.909132728617095e-06,
+      "loss": 1.2258,
+      "step": 133
+    },
+    {
+      "epoch": 1.3009708737864076,
+      "grad_norm": 0.6818450934871862,
+      "learning_rate": 4.907769662274785e-06,
+      "loss": 0.9031,
+      "step": 134
+    },
+    {
+      "epoch": 1.3106796116504853,
+      "grad_norm": 0.573255811123994,
+      "learning_rate": 4.90639664055879e-06,
+      "loss": 0.7234,
+      "step": 135
+    },
+    {
+      "epoch": 1.3203883495145632,
+      "grad_norm": 0.6504319050897613,
+      "learning_rate": 4.905013669146127e-06,
+      "loss": 0.8108,
+      "step": 136
+    },
+    {
+      "epoch": 1.3300970873786409,
+      "grad_norm": 0.6160025172155659,
+      "learning_rate": 4.903620753754949e-06,
+      "loss": 0.951,
+      "step": 137
+    },
+    {
+      "epoch": 1.3398058252427185,
+      "grad_norm": 0.7489527548993854,
+      "learning_rate": 4.902217900144524e-06,
+      "loss": 1.1361,
+      "step": 138
+    },
+    {
+      "epoch": 1.3495145631067962,
+      "grad_norm": 0.7112858054270321,
+      "learning_rate": 4.900805114115214e-06,
+      "loss": 1.1212,
+      "step": 139
+    },
+    {
+      "epoch": 1.3592233009708738,
+      "grad_norm": 0.6142945035367823,
+      "learning_rate": 4.899382401508446e-06,
+      "loss": 0.6799,
+      "step": 140
+    },
+    {
+      "epoch": 1.3689320388349515,
+      "grad_norm": 0.7192927924067215,
+      "learning_rate": 4.8979497682066916e-06,
+      "loss": 0.9943,
+      "step": 141
+    },
+    {
+      "epoch": 1.3786407766990292,
+      "grad_norm": 0.5629568582132167,
+      "learning_rate": 4.89650722013344e-06,
+      "loss": 0.6788,
+      "step": 142
+    },
+    {
+      "epoch": 1.3883495145631068,
+      "grad_norm": 1.1483194850326925,
+      "learning_rate": 4.895054763253177e-06,
+      "loss": 1.0813,
+      "step": 143
+    },
+    {
+      "epoch": 1.3980582524271845,
+      "grad_norm": 0.7435969328117086,
+      "learning_rate": 4.8935924035713564e-06,
+      "loss": 0.8197,
+      "step": 144
+    },
+    {
+      "epoch": 1.4077669902912622,
+      "grad_norm": 1.317031074434914,
+      "learning_rate": 4.892120147134378e-06,
+      "loss": 1.0709,
+      "step": 145
+    },
+    {
+      "epoch": 1.4174757281553398,
+      "grad_norm": 0.5996482737135205,
+      "learning_rate": 4.8906380000295615e-06,
+      "loss": 0.7866,
+      "step": 146
+    },
+    {
+      "epoch": 1.4271844660194175,
+      "grad_norm": 0.7791569768007636,
+      "learning_rate": 4.889145968385121e-06,
+      "loss": 0.8691,
+      "step": 147
+    },
+    {
+      "epoch": 1.4368932038834952,
+      "grad_norm": 0.6501971847860233,
+      "learning_rate": 4.887644058370139e-06,
+      "loss": 0.6976,
+      "step": 148
+    },
+    {
+      "epoch": 1.4466019417475728,
+      "grad_norm": 0.6511216890119076,
+      "learning_rate": 4.886132276194544e-06,
+      "loss": 0.7491,
+      "step": 149
+    },
+    {
+      "epoch": 1.4563106796116505,
+      "grad_norm": 0.612023999998433,
+      "learning_rate": 4.884610628109082e-06,
+      "loss": 0.717,
+      "step": 150
+    },
+    {
+      "epoch": 1.4660194174757282,
+      "grad_norm": 0.7773151050365769,
+      "learning_rate": 4.883079120405292e-06,
+      "loss": 1.004,
+      "step": 151
+    },
+    {
+      "epoch": 1.4757281553398058,
+      "grad_norm": 0.6158372837580877,
+      "learning_rate": 4.881537759415478e-06,
+      "loss": 0.6869,
+      "step": 152
+    },
+    {
+      "epoch": 1.4854368932038835,
+      "grad_norm": 0.8120612151123343,
+      "learning_rate": 4.879986551512684e-06,
+      "loss": 0.9272,
+      "step": 153
+    },
+    {
+      "epoch": 1.4951456310679612,
+      "grad_norm": 0.9467728776526185,
+      "learning_rate": 4.878425503110672e-06,
+      "loss": 0.7134,
+      "step": 154
+    },
+    {
+      "epoch": 1.5048543689320388,
+      "grad_norm": 0.7268524122061516,
+      "learning_rate": 4.876854620663887e-06,
+      "loss": 0.9148,
+      "step": 155
+    },
+    {
+      "epoch": 1.5145631067961165,
+      "grad_norm": 0.5995550004105531,
+      "learning_rate": 4.875273910667434e-06,
+      "loss": 0.74,
+      "step": 156
+    },
+    {
+      "epoch": 1.5242718446601942,
+      "grad_norm": 0.8211820446136313,
+      "learning_rate": 4.873683379657057e-06,
+      "loss": 0.7739,
+      "step": 157
+    },
+    {
+      "epoch": 1.5339805825242718,
+      "grad_norm": 0.6823628164842023,
+      "learning_rate": 4.8720830342091015e-06,
+      "loss": 1.1488,
+      "step": 158
+    },
+    {
+      "epoch": 1.5436893203883495,
+      "grad_norm": 0.7014265251368315,
+      "learning_rate": 4.870472880940496e-06,
+      "loss": 0.9573,
+      "step": 159
+    },
+    {
+      "epoch": 1.5533980582524272,
+      "grad_norm": 0.8352178057977285,
+      "learning_rate": 4.868852926508721e-06,
+      "loss": 0.9985,
+      "step": 160
+    },
+    {
+      "epoch": 1.5631067961165048,
+      "grad_norm": 0.756687144256477,
+      "learning_rate": 4.867223177611779e-06,
+      "loss": 0.938,
+      "step": 161
+    },
+    {
+      "epoch": 1.5728155339805825,
+      "grad_norm": 0.704162668693477,
+      "learning_rate": 4.865583640988173e-06,
+      "loss": 0.7891,
+      "step": 162
+    },
+    {
+      "epoch": 1.5825242718446602,
+      "grad_norm": 0.5937940435523842,
+      "learning_rate": 4.863934323416871e-06,
+      "loss": 0.7824,
+      "step": 163
+    },
+    {
+      "epoch": 1.5922330097087378,
+      "grad_norm": 0.5548908895552046,
+      "learning_rate": 4.862275231717288e-06,
+      "loss": 0.7329,
+      "step": 164
+    },
+    {
+      "epoch": 1.6019417475728155,
+      "grad_norm": 0.8668497443441112,
+      "learning_rate": 4.860606372749247e-06,
+      "loss": 0.7515,
+      "step": 165
+    },
+    {
+      "epoch": 1.6116504854368932,
+      "grad_norm": 0.8678487798318208,
+      "learning_rate": 4.858927753412958e-06,
+      "loss": 0.9059,
+      "step": 166
+    },
+    {
+      "epoch": 1.6213592233009708,
+      "grad_norm": 0.6236208019957972,
+      "learning_rate": 4.857239380648985e-06,
+      "loss": 0.7846,
+      "step": 167
+    },
+    {
+      "epoch": 1.6310679611650487,
+      "grad_norm": 0.6093138120471663,
+      "learning_rate": 4.855541261438223e-06,
+      "loss": 0.8363,
+      "step": 168
+    },
+    {
+      "epoch": 1.6407766990291264,
+      "grad_norm": 0.6297044385352409,
+      "learning_rate": 4.8538334028018605e-06,
+      "loss": 1.0275,
+      "step": 169
+    },
+    {
+      "epoch": 1.650485436893204,
+      "grad_norm": 0.7705428538880952,
+      "learning_rate": 4.8521158118013605e-06,
+      "loss": 0.8634,
+      "step": 170
+    },
+    {
+      "epoch": 1.6601941747572817,
+      "grad_norm": 0.5714628211852456,
+      "learning_rate": 4.850388495538423e-06,
+      "loss": 0.7612,
+      "step": 171
+    },
+    {
+      "epoch": 1.6699029126213594,
+      "grad_norm": 0.6093620351916359,
+      "learning_rate": 4.84865146115496e-06,
+      "loss": 0.6787,
+      "step": 172
+    },
+    {
+      "epoch": 1.679611650485437,
+      "grad_norm": 0.6119153926928719,
+      "learning_rate": 4.846904715833066e-06,
+      "loss": 0.6522,
+      "step": 173
+    },
+    {
+      "epoch": 1.6893203883495147,
+      "grad_norm": 1.4628803952904057,
+      "learning_rate": 4.8451482667949836e-06,
+      "loss": 0.9146,
+      "step": 174
+    },
+    {
+      "epoch": 1.6990291262135924,
+      "grad_norm": 0.6382465121072494,
+      "learning_rate": 4.843382121303082e-06,
+      "loss": 0.7933,
+      "step": 175
+    },
+    {
+      "epoch": 1.70873786407767,
+      "grad_norm": 0.6449697173083225,
+      "learning_rate": 4.841606286659819e-06,
+      "loss": 0.7666,
+      "step": 176
+    },
+    {
+      "epoch": 1.7184466019417477,
+      "grad_norm": 0.5895577438347263,
+      "learning_rate": 4.839820770207714e-06,
+      "loss": 0.6859,
+      "step": 177
+    },
+    {
+      "epoch": 1.7281553398058254,
+      "grad_norm": 0.6104244057063786,
+      "learning_rate": 4.8380255793293195e-06,
+      "loss": 0.691,
+      "step": 178
+    },
+    {
+      "epoch": 1.737864077669903,
+      "grad_norm": 0.6257950437898444,
+      "learning_rate": 4.8362207214471864e-06,
+      "loss": 0.667,
+      "step": 179
+    },
+    {
+      "epoch": 1.7475728155339807,
+      "grad_norm": 0.6404597691271853,
+      "learning_rate": 4.83440620402384e-06,
+      "loss": 0.7931,
+      "step": 180
+    },
+    {
+      "epoch": 1.7572815533980584,
+      "grad_norm": 0.554547749031446,
+      "learning_rate": 4.832582034561738e-06,
+      "loss": 0.7692,
+      "step": 181
+    },
+    {
+      "epoch": 1.766990291262136,
+      "grad_norm": 0.8379042242449998,
+      "learning_rate": 4.830748220603251e-06,
+      "loss": 1.0003,
+      "step": 182
+    },
+    {
+      "epoch": 1.7766990291262137,
+      "grad_norm": 1.371164550147673,
+      "learning_rate": 4.828904769730628e-06,
+      "loss": 1.1004,
+      "step": 183
+    },
+    {
+      "epoch": 1.7864077669902914,
+      "grad_norm": 0.663561268446458,
+      "learning_rate": 4.827051689565958e-06,
+      "loss": 0.7922,
+      "step": 184
+    },
+    {
+      "epoch": 1.796116504854369,
+      "grad_norm": 0.6107817143669849,
+      "learning_rate": 4.825188987771149e-06,
+      "loss": 0.9067,
+      "step": 185
+    },
+    {
+      "epoch": 1.8058252427184467,
+      "grad_norm": 0.5868092910997027,
+      "learning_rate": 4.82331667204789e-06,
+      "loss": 0.5892,
+      "step": 186
+    },
+    {
+      "epoch": 1.8155339805825244,
+      "grad_norm": 0.6461919671595425,
+      "learning_rate": 4.821434750137619e-06,
+      "loss": 0.725,
+      "step": 187
+    },
+    {
+      "epoch": 1.825242718446602,
+      "grad_norm": 0.6489841825930465,
+      "learning_rate": 4.819543229821494e-06,
+      "loss": 0.8711,
+      "step": 188
+    },
+    {
+      "epoch": 1.8349514563106797,
+      "grad_norm": 0.5584447600264582,
+      "learning_rate": 4.8176421189203605e-06,
+      "loss": 0.6982,
+      "step": 189
+    },
+    {
+      "epoch": 1.8446601941747574,
+      "grad_norm": 0.5659976253732854,
+      "learning_rate": 4.815731425294716e-06,
+      "loss": 0.7654,
+      "step": 190
+    },
+    {
+      "epoch": 1.854368932038835,
+      "grad_norm": 0.7825930846614281,
+      "learning_rate": 4.813811156844681e-06,
+      "loss": 0.9132,
+      "step": 191
+    },
+    {
+      "epoch": 1.8640776699029127,
+      "grad_norm": 0.662682846727773,
+      "learning_rate": 4.811881321509964e-06,
+      "loss": 0.849,
+      "step": 192
+    },
+    {
+      "epoch": 1.8737864077669903,
+      "grad_norm": 0.610458993645023,
+      "learning_rate": 4.809941927269829e-06,
+      "loss": 1.0337,
+      "step": 193
+    },
+    {
+      "epoch": 1.883495145631068,
+      "grad_norm": 0.6547992835486613,
+      "learning_rate": 4.807992982143064e-06,
+      "loss": 0.7589,
+      "step": 194
+    },
+    {
+      "epoch": 1.8932038834951457,
+      "grad_norm": 0.7542898472365844,
+      "learning_rate": 4.806034494187949e-06,
+      "loss": 1.0626,
+      "step": 195
+    },
+    {
+      "epoch": 1.9029126213592233,
+      "grad_norm": 0.6344494281728731,
+      "learning_rate": 4.804066471502216e-06,
+      "loss": 0.8039,
+      "step": 196
+    },
+    {
+      "epoch": 1.912621359223301,
+      "grad_norm": 0.6136658573521839,
+      "learning_rate": 4.802088922223024e-06,
+      "loss": 0.6425,
+      "step": 197
+    },
+    {
+      "epoch": 1.9223300970873787,
+      "grad_norm": 0.6483657220529678,
+      "learning_rate": 4.80010185452692e-06,
+      "loss": 0.8784,
+      "step": 198
+    },
+    {
+      "epoch": 1.9320388349514563,
+      "grad_norm": 0.5420123808990978,
+      "learning_rate": 4.798105276629806e-06,
+      "loss": 0.6578,
+      "step": 199
+    },
+    {
+      "epoch": 1.941747572815534,
+      "grad_norm": 0.6687950517102932,
+      "learning_rate": 4.796099196786908e-06,
+      "loss": 0.785,
+      "step": 200
+    },
+    {
+      "epoch": 1.9514563106796117,
+      "grad_norm": 0.6178617434820718,
+      "learning_rate": 4.794083623292737e-06,
+      "loss": 0.9773,
+      "step": 201
+    },
+    {
+      "epoch": 1.9611650485436893,
+      "grad_norm": 0.6711763346625357,
+      "learning_rate": 4.792058564481058e-06,
+      "loss": 0.7296,
+      "step": 202
+    },
+    {
+      "epoch": 1.970873786407767,
+      "grad_norm": 0.6084143889062773,
+      "learning_rate": 4.7900240287248554e-06,
+      "loss": 0.8921,
+      "step": 203
+    },
+    {
+      "epoch": 1.9805825242718447,
+      "grad_norm": 0.7549301985138527,
+      "learning_rate": 4.7879800244362975e-06,
+      "loss": 0.9962,
+      "step": 204
+    },
+    {
+      "epoch": 1.9902912621359223,
+      "grad_norm": 0.6828205600145024,
+      "learning_rate": 4.785926560066703e-06,
+      "loss": 0.9536,
+      "step": 205
+    },
+    {
+      "epoch": 2.0,
+      "grad_norm": 0.7048432912651719,
+      "learning_rate": 4.783863644106502e-06,
+      "loss": 0.91,
+      "step": 206
+    },
+    {
+      "epoch": 2.0097087378640777,
+      "grad_norm": 0.5925003641609397,
+      "learning_rate": 4.781791285085209e-06,
+      "loss": 0.7574,
+      "step": 207
+    },
+    {
+      "epoch": 2.0194174757281553,
+      "grad_norm": 0.6183118556276791,
+      "learning_rate": 4.779709491571378e-06,
+      "loss": 0.7517,
+      "step": 208
+    },
+    {
+      "epoch": 2.029126213592233,
+      "grad_norm": 0.5513269000902425,
+      "learning_rate": 4.777618272172573e-06,
+      "loss": 0.8483,
+      "step": 209
+    },
+    {
+      "epoch": 2.0388349514563107,
+      "grad_norm": 0.6449371069347959,
+      "learning_rate": 4.775517635535332e-06,
+      "loss": 0.9983,
+      "step": 210
+    },
+    {
+      "epoch": 2.0485436893203883,
+      "grad_norm": 0.6510146500554277,
+      "learning_rate": 4.77340759034513e-06,
+      "loss": 0.943,
+      "step": 211
+    },
+    {
+      "epoch": 2.058252427184466,
+      "grad_norm": 0.5516833176176236,
+      "learning_rate": 4.771288145326343e-06,
+      "loss": 0.8697,
+      "step": 212
+    },
+    {
+      "epoch": 2.0679611650485437,
+      "grad_norm": 0.5372833778405927,
+      "learning_rate": 4.769159309242213e-06,
+      "loss": 0.6181,
+      "step": 213
+    },
+    {
+      "epoch": 2.0776699029126213,
+      "grad_norm": 0.6989081916523086,
+      "learning_rate": 4.767021090894809e-06,
+      "loss": 0.9862,
+      "step": 214
+    },
+    {
+      "epoch": 2.087378640776699,
+      "grad_norm": 0.5906461946301448,
+      "learning_rate": 4.764873499124997e-06,
+      "loss": 0.9271,
+      "step": 215
+    },
+    {
+      "epoch": 2.0970873786407767,
+      "grad_norm": 0.546362671281846,
+      "learning_rate": 4.762716542812395e-06,
+      "loss": 0.7597,
+      "step": 216
+    },
+    {
+      "epoch": 2.1067961165048543,
+      "grad_norm": 0.5766822546336634,
+      "learning_rate": 4.7605502308753415e-06,
+      "loss": 0.7774,
+      "step": 217
+    },
+    {
+      "epoch": 2.116504854368932,
+      "grad_norm": 0.6388672043671925,
+      "learning_rate": 4.758374572270859e-06,
+      "loss": 0.9155,
+      "step": 218
+    },
+    {
+      "epoch": 2.1262135922330097,
+      "grad_norm": 0.5728853545011949,
+      "learning_rate": 4.756189575994614e-06,
+      "loss": 0.6017,
+      "step": 219
+    },
+    {
+      "epoch": 2.1359223300970873,
+      "grad_norm": 0.5366888703406183,
+      "learning_rate": 4.753995251080884e-06,
+      "loss": 0.7425,
+      "step": 220
+    },
+    {
+      "epoch": 2.145631067961165,
+      "grad_norm": 1.3209362433112186,
+      "learning_rate": 4.7517916066025126e-06,
+      "loss": 0.6868,
+      "step": 221
+    },
+    {
+      "epoch": 2.1553398058252426,
+      "grad_norm": 0.6484738703855876,
+      "learning_rate": 4.7495786516708806e-06,
+      "loss": 0.9286,
+      "step": 222
+    },
+    {
+      "epoch": 2.1650485436893203,
+      "grad_norm": 0.6262395253661787,
+      "learning_rate": 4.747356395435865e-06,
+      "loss": 0.6069,
+      "step": 223
+    },
+    {
+      "epoch": 2.174757281553398,
+      "grad_norm": 0.6311571664234611,
+      "learning_rate": 4.745124847085799e-06,
+      "loss": 0.7718,
+      "step": 224
+    },
+    {
+      "epoch": 2.1844660194174756,
+      "grad_norm": 0.613078099794739,
+      "learning_rate": 4.742884015847436e-06,
+      "loss": 0.805,
+      "step": 225
+    },
+    {
+      "epoch": 2.1941747572815533,
+      "grad_norm": 0.5854928160805744,
+      "learning_rate": 4.740633910985911e-06,
+      "loss": 0.8084,
+      "step": 226
+    },
+    {
+      "epoch": 2.203883495145631,
+      "grad_norm": 0.5976089299186353,
+      "learning_rate": 4.738374541804704e-06,
+      "loss": 0.6863,
+      "step": 227
+    },
+    {
+      "epoch": 2.2135922330097086,
+      "grad_norm": 0.6014124521376234,
+      "learning_rate": 4.7361059176456e-06,
+      "loss": 0.7472,
+      "step": 228
+    },
+    {
+      "epoch": 2.2233009708737863,
+      "grad_norm": 0.5685151477620302,
+      "learning_rate": 4.733828047888647e-06,
+      "loss": 0.8648,
+      "step": 229
+    },
+    {
+      "epoch": 2.233009708737864,
+      "grad_norm": 0.5695477126283384,
+      "learning_rate": 4.731540941952126e-06,
+      "loss": 0.5924,
+      "step": 230
+    },
+    {
+      "epoch": 2.2427184466019416,
+      "grad_norm": 0.5651868559249316,
+      "learning_rate": 4.7292446092925016e-06,
+      "loss": 0.6548,
+      "step": 231
+    },
+    {
+      "epoch": 2.2524271844660193,
+      "grad_norm": 0.5710356100550151,
+      "learning_rate": 4.726939059404392e-06,
+      "loss": 0.6689,
+      "step": 232
+    },
+    {
+      "epoch": 2.262135922330097,
+      "grad_norm": 0.5613313610051237,
+      "learning_rate": 4.724624301820524e-06,
+      "loss": 0.673,
+      "step": 233
+    },
+    {
+      "epoch": 2.2718446601941746,
+      "grad_norm": 0.7004250430825069,
+      "learning_rate": 4.722300346111695e-06,
+      "loss": 0.8458,
+      "step": 234
+    },
+    {
+      "epoch": 2.2815533980582523,
+      "grad_norm": 0.6216105948838587,
+      "learning_rate": 4.719967201886734e-06,
+      "loss": 0.7305,
+      "step": 235
+    },
+    {
+      "epoch": 2.29126213592233,
+      "grad_norm": 0.6326118950665,
+      "learning_rate": 4.717624878792461e-06,
+      "loss": 0.9773,
+      "step": 236
+    },
+    {
+      "epoch": 2.3009708737864076,
+      "grad_norm": 0.7022071993385745,
+      "learning_rate": 4.715273386513651e-06,
+      "loss": 0.9996,
+      "step": 237
+    },
+    {
+      "epoch": 2.3106796116504853,
+      "grad_norm": 0.5723457282915828,
+      "learning_rate": 4.712912734772988e-06,
+      "loss": 0.7054,
+      "step": 238
+    },
+    {
+      "epoch": 2.320388349514563,
+      "grad_norm": 0.6945599167433392,
+      "learning_rate": 4.710542933331025e-06,
+      "loss": 0.8439,
+      "step": 239
+    },
+    {
+      "epoch": 2.3300970873786406,
+      "grad_norm": 0.6298659607215348,
+      "learning_rate": 4.708163991986152e-06,
+      "loss": 0.9141,
+      "step": 240
+    },
+    {
+      "epoch": 2.3398058252427183,
+      "grad_norm": 0.5466796831126433,
+      "learning_rate": 4.705775920574546e-06,
+      "loss": 0.8445,
+      "step": 241
+    },
+    {
+      "epoch": 2.349514563106796,
+      "grad_norm": 0.8906548403954916,
+      "learning_rate": 4.703378728970134e-06,
+      "loss": 0.6775,
+      "step": 242
+    },
+    {
+      "epoch": 2.3592233009708736,
+      "grad_norm": 0.5453159831751623,
+      "learning_rate": 4.700972427084551e-06,
+      "loss": 0.8023,
+      "step": 243
+    },
+    {
+      "epoch": 2.3689320388349513,
+      "grad_norm": 0.6086212250840284,
+      "learning_rate": 4.698557024867105e-06,
+      "loss": 0.6731,
+      "step": 244
+    },
+    {
+      "epoch": 2.378640776699029,
+      "grad_norm": 0.5633438402824648,
+      "learning_rate": 4.696132532304727e-06,
+      "loss": 0.8508,
+      "step": 245
+    },
+    {
+      "epoch": 2.3883495145631066,
+      "grad_norm": 0.5713702706197664,
+      "learning_rate": 4.693698959421935e-06,
+      "loss": 0.7985,
+      "step": 246
+    },
+    {
+      "epoch": 2.3980582524271843,
+      "grad_norm": 0.7944485374871474,
+      "learning_rate": 4.691256316280789e-06,
+      "loss": 1.0714,
+      "step": 247
+    },
+    {
+      "epoch": 2.407766990291262,
+      "grad_norm": 0.7146714273823804,
+      "learning_rate": 4.688804612980855e-06,
+      "loss": 0.7486,
+      "step": 248
+    },
+    {
+      "epoch": 2.4174757281553396,
+      "grad_norm": 0.5620249512602807,
+      "learning_rate": 4.686343859659158e-06,
+      "loss": 0.9298,
+      "step": 249
+    },
+    {
+      "epoch": 2.4271844660194173,
+      "grad_norm": 0.573615588141239,
+      "learning_rate": 4.683874066490143e-06,
+      "loss": 0.7426,
+      "step": 250
+    },
+    {
+      "epoch": 2.436893203883495,
+      "grad_norm": 0.5949978059155007,
+      "learning_rate": 4.681395243685631e-06,
+      "loss": 0.7588,
+      "step": 251
+    },
+    {
+      "epoch": 2.4466019417475726,
+      "grad_norm": 0.5548882698312756,
+      "learning_rate": 4.67890740149478e-06,
+      "loss": 0.6808,
+      "step": 252
+    },
+    {
+      "epoch": 2.4563106796116507,
+      "grad_norm": 0.5998026064624129,
+      "learning_rate": 4.676410550204036e-06,
+      "loss": 0.9237,
+      "step": 253
+    },
+    {
+      "epoch": 2.466019417475728,
+      "grad_norm": 0.5343256980470983,
+      "learning_rate": 4.673904700137098e-06,
+      "loss": 0.607,
+      "step": 254
+    },
+    {
+      "epoch": 2.475728155339806,
+      "grad_norm": 0.5129919168354066,
+      "learning_rate": 4.671389861654873e-06,
+      "loss": 0.6515,
+      "step": 255
+    },
+    {
+      "epoch": 2.4854368932038833,
+      "grad_norm": 0.5256394449847143,
+      "learning_rate": 4.668866045155428e-06,
+      "loss": 0.7839,
+      "step": 256
+    },
+    {
+      "epoch": 2.4951456310679614,
+      "grad_norm": 0.6664881549664314,
+      "learning_rate": 4.666333261073956e-06,
+      "loss": 0.8259,
+      "step": 257
+    },
+    {
+      "epoch": 2.5048543689320386,
+      "grad_norm": 0.5878066563376306,
+      "learning_rate": 4.6637915198827265e-06,
+      "loss": 0.8741,
+      "step": 258
+    },
+    {
+      "epoch": 2.5145631067961167,
+      "grad_norm": 0.5762944457977316,
+      "learning_rate": 4.661240832091042e-06,
+      "loss": 0.719,
+      "step": 259
+    },
+    {
+      "epoch": 2.524271844660194,
+      "grad_norm": 0.635760206889188,
+      "learning_rate": 4.658681208245198e-06,
+      "loss": 0.8797,
+      "step": 260
+    },
+    {
+      "epoch": 2.533980582524272,
+      "grad_norm": 0.5296771268285247,
+      "learning_rate": 4.65611265892844e-06,
+      "loss": 0.7687,
+      "step": 261
+    },
+    {
+      "epoch": 2.5436893203883493,
+      "grad_norm": 0.5975151352963404,
+      "learning_rate": 4.653535194760912e-06,
+      "loss": 0.7371,
+      "step": 262
+    },
+    {
+      "epoch": 2.5533980582524274,
+      "grad_norm": 0.5669500366203135,
+      "learning_rate": 4.650948826399624e-06,
+      "loss": 0.5689,
+      "step": 263
+    },
+    {
+      "epoch": 2.5631067961165046,
+      "grad_norm": 0.5800778206254651,
+      "learning_rate": 4.648353564538397e-06,
+      "loss": 0.7246,
+      "step": 264
+    },
+    {
+      "epoch": 2.5728155339805827,
+      "grad_norm": 0.6321546100365262,
+      "learning_rate": 4.645749419907829e-06,
+      "loss": 0.7167,
+      "step": 265
+    },
+    {
+      "epoch": 2.58252427184466,
+      "grad_norm": 0.5319661367490109,
+      "learning_rate": 4.64313640327524e-06,
+      "loss": 0.5765,
+      "step": 266
+    },
+    {
+      "epoch": 2.592233009708738,
+      "grad_norm": 0.6013985476971264,
+      "learning_rate": 4.640514525444637e-06,
+      "loss": 0.8056,
+      "step": 267
+    },
+    {
+      "epoch": 2.6019417475728153,
+      "grad_norm": 0.6212321355444684,
+      "learning_rate": 4.637883797256663e-06,
+      "loss": 1.0431,
+      "step": 268
+    },
+    {
+      "epoch": 2.6116504854368934,
+      "grad_norm": 0.5534158811339417,
+      "learning_rate": 4.635244229588558e-06,
+      "loss": 0.7036,
+      "step": 269
+    },
+    {
+      "epoch": 2.6213592233009706,
+      "grad_norm": 0.6415745872801252,
+      "learning_rate": 4.632595833354105e-06,
+      "loss": 0.8674,
+      "step": 270
+    },
+    {
+      "epoch": 2.6310679611650487,
+      "grad_norm": 0.6285131287505951,
+      "learning_rate": 4.629938619503593e-06,
+      "loss": 0.7088,
+      "step": 271
+    },
+    {
+      "epoch": 2.6407766990291264,
+      "grad_norm": 0.5291268994493087,
+      "learning_rate": 4.627272599023772e-06,
+      "loss": 0.8775,
+      "step": 272
+    },
+    {
+      "epoch": 2.650485436893204,
+      "grad_norm": 0.5852978169428409,
+      "learning_rate": 4.6245977829378e-06,
+      "loss": 1.0301,
+      "step": 273
+    },
+    {
+      "epoch": 2.6601941747572817,
+      "grad_norm": 0.5619590148291557,
+      "learning_rate": 4.6219141823052035e-06,
+      "loss": 0.655,
+      "step": 274
+    },
+    {
+      "epoch": 2.6699029126213594,
+      "grad_norm": 0.586715896019821,
+      "learning_rate": 4.619221808221833e-06,
+      "loss": 0.9604,
+      "step": 275
+    },
+    {
+      "epoch": 2.679611650485437,
+      "grad_norm": 0.5855513689539318,
+      "learning_rate": 4.616520671819812e-06,
+      "loss": 0.6424,
+      "step": 276
+    },
+    {
+      "epoch": 2.6893203883495147,
+      "grad_norm": 0.647039460728171,
+      "learning_rate": 4.613810784267492e-06,
+      "loss": 0.8329,
+      "step": 277
+    },
+    {
+      "epoch": 2.6990291262135924,
+      "grad_norm": 0.5154279265816364,
+      "learning_rate": 4.61109215676941e-06,
+      "loss": 0.588,
+      "step": 278
+    },
+    {
+      "epoch": 2.70873786407767,
+      "grad_norm": 0.5832083686660431,
+      "learning_rate": 4.608364800566241e-06,
+      "loss": 0.7689,
+      "step": 279
+    },
+    {
+      "epoch": 2.7184466019417477,
+      "grad_norm": 0.5884877738799021,
+      "learning_rate": 4.605628726934747e-06,
+      "loss": 0.8834,
+      "step": 280
+    },
+    {
+      "epoch": 2.7281553398058254,
+      "grad_norm": 0.5126503459961328,
+      "learning_rate": 4.602883947187738e-06,
+      "loss": 0.7756,
+      "step": 281
+    },
+    {
+      "epoch": 2.737864077669903,
+      "grad_norm": 0.5509578989929416,
+      "learning_rate": 4.600130472674017e-06,
+      "loss": 0.63,
+      "step": 282
+    },
+    {
+      "epoch": 2.7475728155339807,
+      "grad_norm": 0.5937802430842546,
+      "learning_rate": 4.5973683147783405e-06,
+      "loss": 0.9692,
+      "step": 283
+    },
+    {
+      "epoch": 2.7572815533980584,
+      "grad_norm": 0.8437616641968025,
+      "learning_rate": 4.594597484921365e-06,
+      "loss": 0.7999,
+      "step": 284
+    },
+    {
+      "epoch": 2.766990291262136,
+      "grad_norm": 0.5815733690670376,
+      "learning_rate": 4.5918179945596055e-06,
+      "loss": 0.7197,
+      "step": 285
+    },
+    {
+      "epoch": 2.7766990291262137,
+      "grad_norm": 0.6766231348468958,
+      "learning_rate": 4.589029855185384e-06,
+      "loss": 0.8009,
+      "step": 286
+    },
+    {
+      "epoch": 2.7864077669902914,
+      "grad_norm": 0.6016607485166134,
+      "learning_rate": 4.586233078326785e-06,
+      "loss": 1.055,
+      "step": 287
+    },
+    {
+      "epoch": 2.796116504854369,
+      "grad_norm": 0.5385528216529567,
+      "learning_rate": 4.583427675547602e-06,
+      "loss": 0.8221,
+      "step": 288
+    },
+    {
+      "epoch": 2.8058252427184467,
+      "grad_norm": 0.5993566648276473,
+      "learning_rate": 4.580613658447301e-06,
+      "loss": 0.938,
+      "step": 289
+    },
+    {
+      "epoch": 2.8155339805825244,
+      "grad_norm": 0.5912484848249645,
+      "learning_rate": 4.577791038660959e-06,
+      "loss": 0.6534,
+      "step": 290
+    },
+    {
+      "epoch": 2.825242718446602,
+      "grad_norm": 0.5558723739620934,
+      "learning_rate": 4.574959827859226e-06,
+      "loss": 0.6328,
+      "step": 291
+    },
+    {
+      "epoch": 2.8349514563106797,
+      "grad_norm": 0.5215220174359476,
+      "learning_rate": 4.572120037748273e-06,
+      "loss": 0.6078,
+      "step": 292
+    },
+    {
+      "epoch": 2.8446601941747574,
+      "grad_norm": 0.5578352792075171,
+      "learning_rate": 4.5692716800697415e-06,
+      "loss": 0.7983,
+      "step": 293
+    },
+    {
+      "epoch": 2.854368932038835,
+      "grad_norm": 0.5631763967068151,
+      "learning_rate": 4.566414766600698e-06,
+      "loss": 0.7932,
+      "step": 294
+    },
+    {
+      "epoch": 2.8640776699029127,
+      "grad_norm": 0.5467170630989897,
+      "learning_rate": 4.563549309153589e-06,
+      "loss": 0.6513,
+      "step": 295
+    },
+    {
+      "epoch": 2.8737864077669903,
+      "grad_norm": 0.6087442326908157,
+      "learning_rate": 4.56067531957618e-06,
+      "loss": 0.7059,
+      "step": 296
+    },
+    {
+      "epoch": 2.883495145631068,
+      "grad_norm": 0.6345964624606057,
+      "learning_rate": 4.557792809751519e-06,
+      "loss": 0.8497,
+      "step": 297
+    },
+    {
+      "epoch": 2.8932038834951457,
+      "grad_norm": 0.6421474826199707,
+      "learning_rate": 4.554901791597883e-06,
+      "loss": 0.754,
+      "step": 298
+    },
+    {
+      "epoch": 2.9029126213592233,
+      "grad_norm": 0.5611662286254485,
+      "learning_rate": 4.552002277068725e-06,
+      "loss": 0.7587,
+      "step": 299
+    },
+    {
+      "epoch": 2.912621359223301,
+      "grad_norm": 0.6356794525355608,
+      "learning_rate": 4.549094278152631e-06,
+      "loss": 0.8163,
+      "step": 300
+    },
+    {
+      "epoch": 2.9223300970873787,
+      "grad_norm": 0.6336849057165008,
+      "learning_rate": 4.546177806873266e-06,
+      "loss": 0.682,
+      "step": 301
+    },
+    {
+      "epoch": 2.9320388349514563,
+      "grad_norm": 0.6059295762410952,
+      "learning_rate": 4.543252875289326e-06,
+      "loss": 0.6621,
+      "step": 302
+    },
+    {
+      "epoch": 2.941747572815534,
+      "grad_norm": 0.5887024077749718,
+      "learning_rate": 4.540319495494486e-06,
+      "loss": 0.7904,
+      "step": 303
+    },
+    {
+      "epoch": 2.9514563106796117,
+      "grad_norm": 0.6053171571638954,
+      "learning_rate": 4.537377679617353e-06,
+      "loss": 0.8159,
+      "step": 304
+    },
+    {
+      "epoch": 2.9611650485436893,
+      "grad_norm": 0.6162643989939205,
+      "learning_rate": 4.534427439821416e-06,
+      "loss": 0.7054,
+      "step": 305
+    },
+    {
+      "epoch": 2.970873786407767,
+      "grad_norm": 0.5653687306702669,
+      "learning_rate": 4.531468788304992e-06,
+      "loss": 0.6473,
+      "step": 306
+    },
+    {
+      "epoch": 2.9805825242718447,
+      "grad_norm": 0.6147243558106896,
+      "learning_rate": 4.5285017373011784e-06,
+      "loss": 0.6811,
+      "step": 307
+    },
+    {
+      "epoch": 2.9902912621359223,
+      "grad_norm": 0.5023058479913145,
+      "learning_rate": 4.5255262990778024e-06,
+      "loss": 0.5666,
+      "step": 308
+    },
+    {
+      "epoch": 3.0,
+      "grad_norm": 0.5339774541368152,
+      "learning_rate": 4.522542485937369e-06,
+      "loss": 0.7024,
+      "step": 309
+    },
+    {
+      "epoch": 3.0097087378640777,
+      "grad_norm": 0.5871214604141014,
+      "learning_rate": 4.519550310217013e-06,
+      "loss": 0.792,
+      "step": 310
+    },
+    {
+      "epoch": 3.0194174757281553,
+      "grad_norm": 0.550905273830553,
+      "learning_rate": 4.516549784288442e-06,
+      "loss": 0.6801,
+      "step": 311
+    },
+    {
+      "epoch": 3.029126213592233,
+      "grad_norm": 0.5295468316460928,
+      "learning_rate": 4.513540920557892e-06,
+      "loss": 0.6589,
+      "step": 312
+    },
+    {
+      "epoch": 3.0388349514563107,
+      "grad_norm": 0.5818378095787248,
+      "learning_rate": 4.510523731466072e-06,
+      "loss": 0.9006,
+      "step": 313
+    },
+    {
+      "epoch": 3.0485436893203883,
+      "grad_norm": 0.53839185417949,
+      "learning_rate": 4.507498229488116e-06,
+      "loss": 0.5935,
+      "step": 314
+    },
+    {
+      "epoch": 3.058252427184466,
+      "grad_norm": 0.5162101468182391,
+      "learning_rate": 4.504464427133527e-06,
+      "loss": 0.6615,
+      "step": 315
+    },
+    {
+      "epoch": 3.0679611650485437,
+      "grad_norm": 0.5627252995451077,
+      "learning_rate": 4.501422336946126e-06,
+      "loss": 0.6778,
+      "step": 316
+    },
+    {
+      "epoch": 3.0776699029126213,
+      "grad_norm": 0.5499924181430055,
+      "learning_rate": 4.498371971504005e-06,
+      "loss": 0.6824,
+      "step": 317
+    },
+    {
+      "epoch": 3.087378640776699,
+      "grad_norm": 0.5313000147551927,
+      "learning_rate": 4.49531334341947e-06,
+      "loss": 0.6953,
+      "step": 318
+    },
+    {
+      "epoch": 3.0970873786407767,
+      "grad_norm": 0.5053678892953669,
+      "learning_rate": 4.49224646533899e-06,
+      "loss": 0.624,
+      "step": 319
+    },
+    {
+      "epoch": 3.1067961165048543,
+      "grad_norm": 0.5607765091857444,
+      "learning_rate": 4.489171349943144e-06,
+      "loss": 0.6887,
+      "step": 320
+    },
+    {
+      "epoch": 3.116504854368932,
+      "grad_norm": 0.49971275746543925,
+      "learning_rate": 4.486088009946575e-06,
+      "loss": 0.599,
+      "step": 321
+    },
+    {
+      "epoch": 3.1262135922330097,
+      "grad_norm": 0.5463448565404533,
+      "learning_rate": 4.482996458097926e-06,
+      "loss": 0.7412,
+      "step": 322
+    },
+    {
+      "epoch": 3.1359223300970873,
+      "grad_norm": 0.5577175615372468,
+      "learning_rate": 4.479896707179796e-06,
+      "loss": 0.9504,
+      "step": 323
+    },
+    {
+      "epoch": 3.145631067961165,
+      "grad_norm": 0.739795464222875,
+      "learning_rate": 4.476788770008685e-06,
+      "loss": 0.9296,
+      "step": 324
+    },
+    {
+      "epoch": 3.1553398058252426,
+      "grad_norm": 0.5860377779118467,
+      "learning_rate": 4.473672659434941e-06,
+      "loss": 0.8433,
+      "step": 325
+    },
+    {
+      "epoch": 3.1650485436893203,
+      "grad_norm": 0.5234329633422101,
+      "learning_rate": 4.470548388342704e-06,
+      "loss": 0.7058,
+      "step": 326
+    },
+    {
+      "epoch": 3.174757281553398,
+      "grad_norm": 0.5307383652263797,
+      "learning_rate": 4.467415969649858e-06,
+      "loss": 0.7462,
+      "step": 327
+    },
+    {
+      "epoch": 3.1844660194174756,
+      "grad_norm": 0.5283773909342946,
+      "learning_rate": 4.464275416307973e-06,
+      "loss": 0.7497,
+      "step": 328
+    },
+    {
+      "epoch": 3.1941747572815533,
+      "grad_norm": 0.5706683381161813,
+      "learning_rate": 4.461126741302253e-06,
+      "loss": 0.7677,
+      "step": 329
+    },
+    {
+      "epoch": 3.203883495145631,
+      "grad_norm": 0.568157264873081,
+      "learning_rate": 4.457969957651485e-06,
+      "loss": 0.6759,
+      "step": 330
+    },
+    {
+      "epoch": 3.2135922330097086,
+      "grad_norm": 0.5665376751141662,
+      "learning_rate": 4.454805078407979e-06,
+      "loss": 0.9144,
+      "step": 331
+    },
+    {
+      "epoch": 3.2233009708737863,
+      "grad_norm": 0.5146029291558777,
+      "learning_rate": 4.451632116657521e-06,
+      "loss": 0.6786,
+      "step": 332
+    },
+    {
+      "epoch": 3.233009708737864,
+      "grad_norm": 0.4670736553738834,
+      "learning_rate": 4.448451085519314e-06,
+      "loss": 0.5646,
+      "step": 333
+    },
+    {
+      "epoch": 3.2427184466019416,
+      "grad_norm": 0.5983555089379993,
+      "learning_rate": 4.445261998145927e-06,
+      "loss": 0.7486,
+      "step": 334
+    },
+    {
+      "epoch": 3.2524271844660193,
+      "grad_norm": 0.5247722963016707,
+      "learning_rate": 4.442064867723236e-06,
+      "loss": 0.7228,
+      "step": 335
+    },
+    {
+      "epoch": 3.262135922330097,
+      "grad_norm": 0.5830371706199331,
+      "learning_rate": 4.438859707470376e-06,
+      "loss": 0.7545,
+      "step": 336
+    },
+    {
+      "epoch": 3.2718446601941746,
+      "grad_norm": 0.569317200535919,
+      "learning_rate": 4.435646530639679e-06,
+      "loss": 0.6446,
+      "step": 337
+    },
+    {
+      "epoch": 3.2815533980582523,
+      "grad_norm": 0.5229847762890999,
+      "learning_rate": 4.432425350516627e-06,
+      "loss": 0.7016,
+      "step": 338
+    },
+    {
+      "epoch": 3.29126213592233,
+      "grad_norm": 0.5556853198137727,
+      "learning_rate": 4.42919618041979e-06,
+      "loss": 0.581,
+      "step": 339
+    },
+    {
+      "epoch": 3.3009708737864076,
+      "grad_norm": 0.5504701249347626,
+      "learning_rate": 4.425959033700776e-06,
+      "loss": 0.4669,
+      "step": 340
+    },
+    {
+      "epoch": 3.3106796116504853,
+      "grad_norm": 0.5733436063811748,
+      "learning_rate": 4.422713923744174e-06,
+      "loss": 0.7609,
+      "step": 341
+    },
+    {
+      "epoch": 3.320388349514563,
+      "grad_norm": 0.5712264087934034,
+      "learning_rate": 4.419460863967496e-06,
+      "loss": 0.796,
+      "step": 342
+    },
+    {
+      "epoch": 3.3300970873786406,
+      "grad_norm": 0.5298396914849735,
+      "learning_rate": 4.416199867821126e-06,
+      "loss": 0.7347,
+      "step": 343
+    },
+    {
+      "epoch": 3.3398058252427183,
+      "grad_norm": 0.6167638285377818,
+      "learning_rate": 4.412930948788263e-06,
+      "loss": 0.7213,
+      "step": 344
+    },
+    {
+      "epoch": 3.349514563106796,
+      "grad_norm": 0.6769548528256155,
+      "learning_rate": 4.409654120384863e-06,
+      "loss": 0.9705,
+      "step": 345
+    },
+    {
+      "epoch": 3.3592233009708736,
+      "grad_norm": 0.5558234569754197,
+      "learning_rate": 4.406369396159585e-06,
+      "loss": 0.866,
+      "step": 346
+    },
+    {
+      "epoch": 3.3689320388349513,
+      "grad_norm": 0.5683210417700995,
+      "learning_rate": 4.403076789693735e-06,
+      "loss": 0.8529,
+      "step": 347
+    },
+    {
+      "epoch": 3.378640776699029,
+      "grad_norm": 0.6004152868285685,
+      "learning_rate": 4.399776314601212e-06,
+      "loss": 0.6924,
+      "step": 348
+    },
+    {
+      "epoch": 3.3883495145631066,
+      "grad_norm": 0.5915756102782764,
+      "learning_rate": 4.396467984528445e-06,
+      "loss": 0.809,
+      "step": 349
+    },
+    {
+      "epoch": 3.3980582524271843,
+      "grad_norm": 0.5726590426952238,
+      "learning_rate": 4.393151813154345e-06,
+      "loss": 0.6593,
+      "step": 350
+    },
+    {
+      "epoch": 3.407766990291262,
+      "grad_norm": 0.5071226489196179,
+      "learning_rate": 4.3898278141902396e-06,
+      "loss": 0.6473,
+      "step": 351
+    },
+    {
+      "epoch": 3.4174757281553396,
+      "grad_norm": 0.5733646491001667,
+      "learning_rate": 4.386496001379826e-06,
+      "loss": 0.8021,
+      "step": 352
+    },
+    {
+      "epoch": 3.4271844660194173,
+      "grad_norm": 0.5616644007222712,
+      "learning_rate": 4.383156388499106e-06,
+      "loss": 0.7276,
+      "step": 353
+    },
+    {
+      "epoch": 3.436893203883495,
+      "grad_norm": 0.5270931676906139,
+      "learning_rate": 4.3798089893563335e-06,
+      "loss": 0.6511,
+      "step": 354
+    },
+    {
+      "epoch": 3.4466019417475726,
+      "grad_norm": 0.5956176073578046,
+      "learning_rate": 4.3764538177919555e-06,
+      "loss": 0.5386,
+      "step": 355
+    },
+    {
+      "epoch": 3.4563106796116507,
+      "grad_norm": 0.6087361244099403,
+      "learning_rate": 4.3730908876785574e-06,
+      "loss": 0.7574,
+      "step": 356
+    },
+    {
+      "epoch": 3.466019417475728,
+      "grad_norm": 0.6124986417248809,
+      "learning_rate": 4.3697202129208e-06,
+      "loss": 0.797,
+      "step": 357
+    },
+    {
+      "epoch": 3.475728155339806,
+      "grad_norm": 0.5233985914467231,
+      "learning_rate": 4.36634180745537e-06,
+      "loss": 0.5764,
+      "step": 358
+    },
+    {
+      "epoch": 3.4854368932038833,
+      "grad_norm": 0.5590564532881283,
+      "learning_rate": 4.3629556852509145e-06,
+      "loss": 0.9866,
+      "step": 359
+    },
+    {
+      "epoch": 3.4951456310679614,
+      "grad_norm": 0.5100762691051512,
+      "learning_rate": 4.35956186030799e-06,
+      "loss": 0.6127,
+      "step": 360
+    },
+    {
+      "epoch": 3.5048543689320386,
+      "grad_norm": 0.6118806932618559,
+      "learning_rate": 4.356160346659001e-06,
+      "loss": 0.7548,
+      "step": 361
+    },
+    {
+      "epoch": 3.5145631067961167,
+      "grad_norm": 0.6631159487479685,
+      "learning_rate": 4.3527511583681384e-06,
+      "loss": 1.2605,
+      "step": 362
+    },
+    {
+      "epoch": 3.524271844660194,
+      "grad_norm": 0.6075337136532736,
+      "learning_rate": 4.34933430953133e-06,
+      "loss": 0.7971,
+      "step": 363
+    },
+    {
+      "epoch": 3.533980582524272,
+      "grad_norm": 0.6085963816816855,
+      "learning_rate": 4.345909814276177e-06,
+      "loss": 0.8115,
+      "step": 364
+    },
+    {
+      "epoch": 3.5436893203883493,
+      "grad_norm": 0.6489049958725044,
+      "learning_rate": 4.3424776867618935e-06,
+      "loss": 0.8511,
+      "step": 365
+    },
+    {
+      "epoch": 3.5533980582524274,
+      "grad_norm": 0.5533454922999531,
+      "learning_rate": 4.339037941179253e-06,
+      "loss": 0.7739,
+      "step": 366
+    },
+    {
+      "epoch": 3.5631067961165046,
+      "grad_norm": 0.5324277334095654,
+      "learning_rate": 4.335590591750526e-06,
+      "loss": 0.6825,
+      "step": 367
+    },
+    {
+      "epoch": 3.5728155339805827,
+      "grad_norm": 0.583139527089237,
+      "learning_rate": 4.332135652729423e-06,
+      "loss": 0.6584,
+      "step": 368
+    },
+    {
+      "epoch": 3.58252427184466,
+      "grad_norm": 0.5359360883573963,
+      "learning_rate": 4.328673138401036e-06,
+      "loss": 0.7771,
+      "step": 369
+    },
+    {
+      "epoch": 3.592233009708738,
+      "grad_norm": 0.5374666185272033,
+      "learning_rate": 4.325203063081776e-06,
+      "loss": 0.4958,
+      "step": 370
+    },
+    {
+      "epoch": 3.6019417475728153,
+      "grad_norm": 0.5182456153897488,
+      "learning_rate": 4.32172544111932e-06,
+      "loss": 0.6043,
+      "step": 371
+    },
+    {
+      "epoch": 3.6116504854368934,
+      "grad_norm": 0.5463365868711481,
+      "learning_rate": 4.318240286892544e-06,
+      "loss": 0.67,
+      "step": 372
+    },
+    {
+      "epoch": 3.6213592233009706,
+      "grad_norm": 0.5832479071806481,
+      "learning_rate": 4.314747614811471e-06,
+      "loss": 0.7424,
+      "step": 373
+    },
+    {
+      "epoch": 3.6310679611650487,
+      "grad_norm": 0.5054355197048168,
+      "learning_rate": 4.3112474393172055e-06,
+      "loss": 0.5487,
+      "step": 374
+    },
+    {
+      "epoch": 3.6407766990291264,
+      "grad_norm": 0.5382871622608469,
+      "learning_rate": 4.307739774881878e-06,
+      "loss": 0.6772,
+      "step": 375
+    },
+    {
+      "epoch": 3.650485436893204,
+      "grad_norm": 0.5458477824608565,
+      "learning_rate": 4.304224636008582e-06,
+      "loss": 0.5329,
+      "step": 376
+    },
+    {
+      "epoch": 3.6601941747572817,
+      "grad_norm": 0.5435683453057218,
+      "learning_rate": 4.300702037231318e-06,
+      "loss": 0.8113,
+      "step": 377
+    },
+    {
+      "epoch": 3.6699029126213594,
+      "grad_norm": 0.5383065920679537,
+      "learning_rate": 4.297171993114927e-06,
+      "loss": 0.5284,
+      "step": 378
+    },
+    {
+      "epoch": 3.679611650485437,
+      "grad_norm": 0.5880433935588353,
+      "learning_rate": 4.2936345182550365e-06,
+      "loss": 0.8539,
+      "step": 379
+    },
+    {
+      "epoch": 3.6893203883495147,
+      "grad_norm": 0.5952587100488826,
+      "learning_rate": 4.290089627277998e-06,
+      "loss": 0.5857,
+      "step": 380
+    },
+    {
+      "epoch": 3.6990291262135924,
+      "grad_norm": 0.5286616194694634,
+      "learning_rate": 4.286537334840825e-06,
+      "loss": 0.7462,
+      "step": 381
+    },
+    {
+      "epoch": 3.70873786407767,
+      "grad_norm": 0.6011010828422011,
+      "learning_rate": 4.2829776556311355e-06,
+      "loss": 0.6581,
+      "step": 382
+    },
+    {
+      "epoch": 3.7184466019417477,
+      "grad_norm": 0.59061291418775,
+      "learning_rate": 4.279410604367088e-06,
+      "loss": 0.8607,
+      "step": 383
+    },
+    {
+      "epoch": 3.7281553398058254,
+      "grad_norm": 0.6006128694830518,
+      "learning_rate": 4.275836195797323e-06,
+      "loss": 0.8253,
+      "step": 384
+    },
+    {
+      "epoch": 3.737864077669903,
+      "grad_norm": 0.559500801329847,
+      "learning_rate": 4.2722544447008995e-06,
+      "loss": 0.8451,
+      "step": 385
+    },
+    {
+      "epoch": 3.7475728155339807,
+      "grad_norm": 0.5834501093822622,
+      "learning_rate": 4.268665365887238e-06,
+      "loss": 0.8882,
+      "step": 386
+    },
+    {
+      "epoch": 3.7572815533980584,
+      "grad_norm": 0.5043323998416107,
+      "learning_rate": 4.265068974196056e-06,
+      "loss": 0.5369,
+      "step": 387
+    },
+    {
+      "epoch": 3.766990291262136,
+      "grad_norm": 0.5727251303865297,
+      "learning_rate": 4.261465284497307e-06,
+      "loss": 0.6106,
+      "step": 388
+    },
+    {
+      "epoch": 3.7766990291262137,
+      "grad_norm": 0.5372256278502759,
+      "learning_rate": 4.257854311691118e-06,
+      "loss": 0.7688,
+      "step": 389
+    },
+    {
+      "epoch": 3.7864077669902914,
+      "grad_norm": 0.53361182265545,
+      "learning_rate": 4.254236070707734e-06,
+      "loss": 0.6263,
+      "step": 390
+    },
+    {
+      "epoch": 3.796116504854369,
+      "grad_norm": 0.4761706762756821,
+      "learning_rate": 4.250610576507445e-06,
+      "loss": 0.5494,
+      "step": 391
+    },
+    {
+      "epoch": 3.8058252427184467,
+      "grad_norm": 0.5667551282882679,
+      "learning_rate": 4.246977844080537e-06,
+      "loss": 0.7841,
+      "step": 392
+    },
+    {
+      "epoch": 3.8155339805825244,
+      "grad_norm": 0.5581465960923723,
+      "learning_rate": 4.24333788844722e-06,
+      "loss": 0.727,
+      "step": 393
+    },
+    {
+      "epoch": 3.825242718446602,
+      "grad_norm": 0.641698316355312,
+      "learning_rate": 4.239690724657571e-06,
+      "loss": 0.7415,
+      "step": 394
+    },
+    {
+      "epoch": 3.8349514563106797,
+      "grad_norm": 0.5629852857617874,
+      "learning_rate": 4.236036367791471e-06,
+      "loss": 0.7685,
+      "step": 395
+    },
+    {
+      "epoch": 3.8446601941747574,
+      "grad_norm": 0.5910867032740321,
+      "learning_rate": 4.23237483295854e-06,
+      "loss": 0.797,
+      "step": 396
+    },
+    {
+      "epoch": 3.854368932038835,
+      "grad_norm": 0.5182753315022374,
+      "learning_rate": 4.228706135298081e-06,
+      "loss": 0.5786,
+      "step": 397
+    },
+    {
+      "epoch": 3.8640776699029127,
+      "grad_norm": 0.545028212220101,
+      "learning_rate": 4.225030289979006e-06,
+      "loss": 0.5823,
+      "step": 398
+    },
+    {
+      "epoch": 3.8737864077669903,
+      "grad_norm": 0.5228141064601322,
+      "learning_rate": 4.221347312199788e-06,
+      "loss": 0.5436,
+      "step": 399
+    },
+    {
+      "epoch": 3.883495145631068,
+      "grad_norm": 0.5397693619666596,
+      "learning_rate": 4.2176572171883865e-06,
+      "loss": 0.7937,
+      "step": 400
+    },
+    {
+      "epoch": 3.8932038834951457,
+      "grad_norm": 0.6886419465953945,
+      "learning_rate": 4.213960020202187e-06,
+      "loss": 0.8116,
+      "step": 401
+    },
+    {
+      "epoch": 3.9029126213592233,
+      "grad_norm": 0.5900622648932935,
+      "learning_rate": 4.2102557365279435e-06,
+      "loss": 0.917,
+      "step": 402
+    },
+    {
+      "epoch": 3.912621359223301,
+      "grad_norm": 0.544861080103798,
+      "learning_rate": 4.206544381481708e-06,
+      "loss": 0.8025,
+      "step": 403
+    },
+    {
+      "epoch": 3.9223300970873787,
+      "grad_norm": 0.518875885292913,
+      "learning_rate": 4.202825970408772e-06,
+      "loss": 0.5702,
+      "step": 404
+    },
+    {
+      "epoch": 3.9320388349514563,
+      "grad_norm": 0.656540638196741,
+      "learning_rate": 4.199100518683601e-06,
+      "loss": 0.6321,
+      "step": 405
+    },
+    {
+      "epoch": 3.941747572815534,
+      "grad_norm": 0.5191060858406762,
+      "learning_rate": 4.195368041709772e-06,
+      "loss": 0.7449,
+      "step": 406
+    },
+    {
+      "epoch": 3.9514563106796117,
+      "grad_norm": 0.4946680296221446,
+      "learning_rate": 4.191628554919907e-06,
+      "loss": 0.6497,
+      "step": 407
+    },
+    {
+      "epoch": 3.9611650485436893,
+      "grad_norm": 0.5824785363442938,
+      "learning_rate": 4.187882073775615e-06,
+      "loss": 0.7381,
+      "step": 408
+    },
+    {
+      "epoch": 3.970873786407767,
+      "grad_norm": 0.5193631938328553,
+      "learning_rate": 4.184128613767422e-06,
+      "loss": 0.6982,
+      "step": 409
+    },
+    {
+      "epoch": 3.9805825242718447,
+      "grad_norm": 0.5290930959018318,
+      "learning_rate": 4.18036819041471e-06,
+      "loss": 0.6859,
+      "step": 410
+    },
+    {
+      "epoch": 3.9902912621359223,
+      "grad_norm": 0.5272372038733293,
+      "learning_rate": 4.17660081926565e-06,
+      "loss": 0.5822,
+      "step": 411
+    },
+    {
+      "epoch": 4.0,
+      "grad_norm": 0.5630958714803344,
+      "learning_rate": 4.172826515897146e-06,
+      "loss": 0.6599,
+      "step": 412
+    },
+    {
+      "epoch": 4.009708737864078,
+      "grad_norm": 0.5200333011009248,
+      "learning_rate": 4.169045295914757e-06,
+      "loss": 0.7889,
+      "step": 413
+    },
+    {
+      "epoch": 4.019417475728155,
+      "grad_norm": 0.5331918145560857,
+      "learning_rate": 4.165257174952647e-06,
+      "loss": 0.5669,
+      "step": 414
+    },
+    {
+      "epoch": 4.029126213592233,
+      "grad_norm": 0.4838073648864831,
+      "learning_rate": 4.161462168673508e-06,
+      "loss": 0.6616,
+      "step": 415
+    },
+    {
+      "epoch": 4.038834951456311,
+      "grad_norm": 0.5285061889577369,
+      "learning_rate": 4.157660292768502e-06,
+      "loss": 0.7826,
+      "step": 416
+    },
+    {
+      "epoch": 4.048543689320389,
+      "grad_norm": 0.47876226674574307,
+      "learning_rate": 4.1538515629571985e-06,
+      "loss": 0.6324,
+      "step": 417
+    },
+    {
+      "epoch": 4.058252427184466,
+      "grad_norm": 0.4964106466439782,
+      "learning_rate": 4.1500359949875e-06,
+      "loss": 0.7882,
+      "step": 418
+    },
+    {
+      "epoch": 4.067961165048544,
+      "grad_norm": 0.5098927521794874,
+      "learning_rate": 4.1462136046355864e-06,
+      "loss": 0.679,
+      "step": 419
+    },
+    {
+      "epoch": 4.077669902912621,
+      "grad_norm": 0.5261711579674679,
+      "learning_rate": 4.142384407705846e-06,
+      "loss": 0.9621,
+      "step": 420
+    },
+    {
+      "epoch": 4.087378640776699,
+      "grad_norm": 0.4855524472712664,
+      "learning_rate": 4.138548420030808e-06,
+      "loss": 0.5577,
+      "step": 421
+    },
+    {
+      "epoch": 4.097087378640777,
+      "grad_norm": 0.46404425362535634,
+      "learning_rate": 4.13470565747108e-06,
+      "loss": 0.727,
+      "step": 422
+    },
+    {
+      "epoch": 4.106796116504855,
+      "grad_norm": 0.5013144099880142,
+      "learning_rate": 4.130856135915282e-06,
+      "loss": 0.7378,
+      "step": 423
+    },
+    {
+      "epoch": 4.116504854368932,
+      "grad_norm": 0.5316059543260906,
+      "learning_rate": 4.126999871279982e-06,
+      "loss": 0.6906,
+      "step": 424
+    },
+    {
+      "epoch": 4.12621359223301,
+      "grad_norm": 0.4944636869938843,
+      "learning_rate": 4.123136879509626e-06,
+      "loss": 0.6641,
+      "step": 425
+    },
+    {
+      "epoch": 4.135922330097087,
+      "grad_norm": 0.5095570582691661,
+      "learning_rate": 4.119267176576475e-06,
+      "loss": 0.7491,
+      "step": 426
+    },
+    {
+      "epoch": 4.145631067961165,
+      "grad_norm": 0.54005610895594,
+      "learning_rate": 4.11539077848054e-06,
+      "loss": 0.7558,
+      "step": 427
+    },
+    {
+      "epoch": 4.155339805825243,
+      "grad_norm": 0.5923182555591768,
+      "learning_rate": 4.111507701249513e-06,
+      "loss": 0.7929,
+      "step": 428
+    },
+    {
+      "epoch": 4.165048543689321,
+      "grad_norm": 0.6116168581706061,
+      "learning_rate": 4.107617960938702e-06,
+      "loss": 0.5406,
+      "step": 429
+    },
+    {
+      "epoch": 4.174757281553398,
+      "grad_norm": 0.5210538716839132,
+      "learning_rate": 4.103721573630965e-06,
+      "loss": 0.6724,
+      "step": 430
+    },
+    {
+      "epoch": 4.184466019417476,
+      "grad_norm": 0.47356150602269126,
+      "learning_rate": 4.099818555436645e-06,
+      "loss": 0.6737,
+      "step": 431
+    },
+    {
+      "epoch": 4.194174757281553,
+      "grad_norm": 0.6377655423855296,
+      "learning_rate": 4.095908922493499e-06,
+      "loss": 1.024,
+      "step": 432
+    },
+    {
+      "epoch": 4.203883495145631,
+      "grad_norm": 0.5471085748898301,
+      "learning_rate": 4.091992690966636e-06,
+      "loss": 0.5427,
+      "step": 433
+    },
+    {
+      "epoch": 4.213592233009709,
+      "grad_norm": 0.5818794345348097,
+      "learning_rate": 4.088069877048447e-06,
+      "loss": 0.6099,
+      "step": 434
+    },
+    {
+      "epoch": 4.223300970873787,
+      "grad_norm": 0.543595019351133,
+      "learning_rate": 4.084140496958539e-06,
+      "loss": 0.556,
+      "step": 435
+    },
+    {
+      "epoch": 4.233009708737864,
+      "grad_norm": 0.5101773462629047,
+      "learning_rate": 4.080204566943668e-06,
+      "loss": 0.5016,
+      "step": 436
+    },
+    {
+      "epoch": 4.242718446601942,
+      "grad_norm": 0.574864469322262,
+      "learning_rate": 4.076262103277673e-06,
+      "loss": 0.5518,
+      "step": 437
+    },
+    {
+      "epoch": 4.252427184466019,
+      "grad_norm": 0.529410686127538,
+      "learning_rate": 4.072313122261406e-06,
+      "loss": 0.9056,
+      "step": 438
+    },
+    {
+      "epoch": 4.262135922330097,
+      "grad_norm": 0.4724977490535202,
+      "learning_rate": 4.068357640222668e-06,
+      "loss": 0.5444,
+      "step": 439
+    },
+    {
+      "epoch": 4.271844660194175,
+      "grad_norm": 0.5371234765273634,
+      "learning_rate": 4.06439567351614e-06,
+      "loss": 0.5565,
+      "step": 440
+    },
+    {
+      "epoch": 4.281553398058253,
+      "grad_norm": 0.5554052963535526,
+      "learning_rate": 4.0604272385233105e-06,
+      "loss": 0.6978,
+      "step": 441
+    },
+    {
+      "epoch": 4.29126213592233,
+      "grad_norm": 0.5738857111157796,
+      "learning_rate": 4.056452351652418e-06,
+      "loss": 0.5298,
+      "step": 442
+    },
+    {
+      "epoch": 4.300970873786408,
+      "grad_norm": 0.5183938376729829,
+      "learning_rate": 4.052471029338375e-06,
+      "loss": 0.689,
+      "step": 443
+    },
+    {
+      "epoch": 4.310679611650485,
+      "grad_norm": 0.562235068394362,
+      "learning_rate": 4.048483288042703e-06,
+      "loss": 0.5329,
+      "step": 444
+    },
+    {
+      "epoch": 4.320388349514563,
+      "grad_norm": 0.5873359410798857,
+      "learning_rate": 4.0444891442534615e-06,
+      "loss": 0.6088,
+      "step": 445
+    },
+    {
+      "epoch": 4.330097087378641,
+      "grad_norm": 0.5232886853596697,
+      "learning_rate": 4.040488614485187e-06,
+      "loss": 0.6861,
+      "step": 446
+    },
+    {
+      "epoch": 4.339805825242719,
+      "grad_norm": 0.5707575802489561,
+      "learning_rate": 4.036481715278818e-06,
+      "loss": 0.7804,
+      "step": 447
+    },
+    {
+      "epoch": 4.349514563106796,
+      "grad_norm": 0.5743158436311879,
+      "learning_rate": 4.032468463201626e-06,
+      "loss": 0.7531,
+      "step": 448
+    },
+    {
+      "epoch": 4.359223300970874,
+      "grad_norm": 0.5272870524360327,
+      "learning_rate": 4.028448874847152e-06,
+      "loss": 0.4601,
+      "step": 449
+    },
+    {
+      "epoch": 4.368932038834951,
+      "grad_norm": 0.8245819176512555,
+      "learning_rate": 4.024422966835137e-06,
+      "loss": 0.7204,
+      "step": 450
+    },
+    {
+      "epoch": 4.378640776699029,
+      "grad_norm": 0.5818583543824783,
+      "learning_rate": 4.0203907558114475e-06,
+      "loss": 0.6319,
+      "step": 451
+    },
+    {
+      "epoch": 4.388349514563107,
+      "grad_norm": 0.5916617169196222,
+      "learning_rate": 4.016352258448016e-06,
+      "loss": 0.5578,
+      "step": 452
+    },
+    {
+      "epoch": 4.398058252427185,
+      "grad_norm": 0.5247948377375482,
+      "learning_rate": 4.0123074914427635e-06,
+      "loss": 0.5817,
+      "step": 453
+    },
+    {
+      "epoch": 4.407766990291262,
+      "grad_norm": 0.4841360885356806,
+      "learning_rate": 4.008256471519536e-06,
+      "loss": 0.5309,
+      "step": 454
+    },
+    {
+      "epoch": 4.41747572815534,
+      "grad_norm": 0.5386315901899501,
+      "learning_rate": 4.004199215428032e-06,
+      "loss": 0.6876,
+      "step": 455
+    },
+    {
+      "epoch": 4.427184466019417,
+      "grad_norm": 0.6303477188551428,
+      "learning_rate": 4.000135739943735e-06,
+      "loss": 0.7771,
+      "step": 456
+    },
+    {
+      "epoch": 4.436893203883495,
+      "grad_norm": 0.877305177258536,
+      "learning_rate": 3.996066061867844e-06,
+      "loss": 0.7183,
+      "step": 457
+    },
+    {
+      "epoch": 4.446601941747573,
+      "grad_norm": 0.5475353424276045,
+      "learning_rate": 3.991990198027203e-06,
+      "loss": 0.6071,
+      "step": 458
+    },
+    {
+      "epoch": 4.456310679611651,
+      "grad_norm": 0.5369581006252573,
+      "learning_rate": 3.987908165274233e-06,
+      "loss": 0.6663,
+      "step": 459
+    },
+    {
+      "epoch": 4.466019417475728,
+      "grad_norm": 0.5481338983710499,
+      "learning_rate": 3.9838199804868635e-06,
+      "loss": 0.7901,
+      "step": 460
+    },
+    {
+      "epoch": 4.475728155339806,
+      "grad_norm": 0.5673339500133648,
+      "learning_rate": 3.979725660568456e-06,
+      "loss": 0.7221,
+      "step": 461
+    },
+    {
+      "epoch": 4.485436893203883,
+      "grad_norm": 0.5276433542097172,
+      "learning_rate": 3.975625222447742e-06,
+      "loss": 0.5454,
+      "step": 462
+    },
+    {
+      "epoch": 4.495145631067961,
+      "grad_norm": 0.5773222334966291,
+      "learning_rate": 3.97151868307875e-06,
+      "loss": 0.8548,
+      "step": 463
+    },
+    {
+      "epoch": 4.504854368932039,
+      "grad_norm": 0.5288063685496589,
+      "learning_rate": 3.9674060594407345e-06,
+      "loss": 0.6481,
+      "step": 464
+    },
+    {
+      "epoch": 4.514563106796117,
+      "grad_norm": 0.5234908733008361,
+      "learning_rate": 3.963287368538105e-06,
+      "loss": 0.4874,
+      "step": 465
+    },
+    {
+      "epoch": 4.524271844660194,
+      "grad_norm": 0.5548781615871586,
+      "learning_rate": 3.959162627400361e-06,
+      "loss": 0.7774,
+      "step": 466
+    },
+    {
+      "epoch": 4.533980582524272,
+      "grad_norm": 0.6088390104886395,
+      "learning_rate": 3.9550318530820145e-06,
+      "loss": 0.7768,
+      "step": 467
+    },
+    {
+      "epoch": 4.543689320388349,
+      "grad_norm": 0.5362660929991437,
+      "learning_rate": 3.9508950626625244e-06,
+      "loss": 0.5664,
+      "step": 468
+    },
+    {
+      "epoch": 4.553398058252427,
+      "grad_norm": 0.5403037827442222,
+      "learning_rate": 3.946752273246224e-06,
+      "loss": 0.7226,
+      "step": 469
+    },
+    {
+      "epoch": 4.563106796116505,
+      "grad_norm": 0.5160659142912444,
+      "learning_rate": 3.942603501962249e-06,
+      "loss": 0.609,
+      "step": 470
+    },
+    {
+      "epoch": 4.572815533980583,
+      "grad_norm": 0.6197532352354348,
+      "learning_rate": 3.9384487659644716e-06,
+      "loss": 0.5509,
+      "step": 471
+    },
+    {
+      "epoch": 4.58252427184466,
+      "grad_norm": 0.6600383068172292,
+      "learning_rate": 3.934288082431423e-06,
+      "loss": 0.7257,
+      "step": 472
+    },
+    {
+      "epoch": 4.592233009708738,
+      "grad_norm": 0.5641531137481023,
+      "learning_rate": 3.930121468566227e-06,
+      "loss": 0.8232,
+      "step": 473
+    },
+    {
+      "epoch": 4.601941747572815,
+      "grad_norm": 0.5450663648523331,
+      "learning_rate": 3.925948941596528e-06,
+      "loss": 0.6493,
+      "step": 474
+    },
+    {
+      "epoch": 4.611650485436893,
+      "grad_norm": 0.5134107913642527,
+      "learning_rate": 3.92177051877442e-06,
+      "loss": 0.554,
+      "step": 475
+    },
+    {
+      "epoch": 4.621359223300971,
+      "grad_norm": 0.5988443116784232,
+      "learning_rate": 3.917586217376369e-06,
+      "loss": 0.7745,
+      "step": 476
+    },
+    {
+      "epoch": 4.631067961165049,
+      "grad_norm": 0.5138233596193919,
+      "learning_rate": 3.913396054703155e-06,
+      "loss": 0.7129,
+      "step": 477
+    },
+    {
+      "epoch": 4.640776699029126,
+      "grad_norm": 0.5323215028941234,
+      "learning_rate": 3.909200048079786e-06,
+      "loss": 0.7206,
+      "step": 478
+    },
+    {
+      "epoch": 4.650485436893204,
+      "grad_norm": 0.5875971843837734,
+      "learning_rate": 3.9049982148554384e-06,
+      "loss": 0.8862,
+      "step": 479
+    },
+    {
+      "epoch": 4.660194174757281,
+      "grad_norm": 0.5645023952141737,
+      "learning_rate": 3.900790572403376e-06,
+      "loss": 0.6745,
+      "step": 480
+    },
+    {
+      "epoch": 4.669902912621359,
+      "grad_norm": 0.5154008565576635,
+      "learning_rate": 3.896577138120881e-06,
+      "loss": 0.6947,
+      "step": 481
+    },
+    {
+      "epoch": 4.679611650485437,
+      "grad_norm": 0.5394552750418433,
+      "learning_rate": 3.892357929429187e-06,
+      "loss": 0.6791,
+      "step": 482
+    },
+    {
+      "epoch": 4.689320388349515,
+      "grad_norm": 0.5827930464161047,
+      "learning_rate": 3.8881329637734e-06,
+      "loss": 0.5154,
+      "step": 483
+    },
+    {
+      "epoch": 4.699029126213592,
+      "grad_norm": 0.5361181581166996,
+      "learning_rate": 3.883902258622431e-06,
+      "loss": 0.6468,
+      "step": 484
+    },
+    {
+      "epoch": 4.70873786407767,
+      "grad_norm": 0.5711060546895186,
+      "learning_rate": 3.8796658314689205e-06,
+      "loss": 0.7167,
+      "step": 485
+    },
+    {
+      "epoch": 4.718446601941747,
+      "grad_norm": 0.5434222419984399,
+      "learning_rate": 3.875423699829168e-06,
+      "loss": 0.5867,
+      "step": 486
+    },
+    {
+      "epoch": 4.728155339805825,
+      "grad_norm": 0.5096613481897871,
+      "learning_rate": 3.871175881243061e-06,
+      "loss": 0.6356,
+      "step": 487
+    },
+    {
+      "epoch": 4.737864077669903,
+      "grad_norm": 0.542382695866742,
+      "learning_rate": 3.866922393273999e-06,
+      "loss": 0.7997,
+      "step": 488
+    },
+    {
+      "epoch": 4.747572815533981,
+      "grad_norm": 0.5295202287930876,
+      "learning_rate": 3.862663253508822e-06,
+      "loss": 0.6506,
+      "step": 489
+    },
+    {
+      "epoch": 4.757281553398058,
+      "grad_norm": 0.5116429751055901,
+      "learning_rate": 3.858398479557739e-06,
+      "loss": 0.6037,
+      "step": 490
+    },
+    {
+      "epoch": 4.766990291262136,
+      "grad_norm": 0.5283075153391654,
+      "learning_rate": 3.8541280890542565e-06,
+      "loss": 0.6142,
+      "step": 491
+    },
+    {
+      "epoch": 4.776699029126213,
+      "grad_norm": 0.5741236822380017,
+      "learning_rate": 3.849852099655102e-06,
+      "loss": 0.7518,
+      "step": 492
+    },
+    {
+      "epoch": 4.786407766990291,
+      "grad_norm": 0.5062208506690622,
+      "learning_rate": 3.845570529040151e-06,
+      "loss": 0.6338,
+      "step": 493
+    },
+    {
+      "epoch": 4.796116504854369,
+      "grad_norm": 0.5321985087384002,
+      "learning_rate": 3.841283394912361e-06,
+      "loss": 0.6375,
+      "step": 494
+    },
+    {
+      "epoch": 4.805825242718447,
+      "grad_norm": 0.5310703753315527,
+      "learning_rate": 3.836990714997686e-06,
+      "loss": 0.7015,
+      "step": 495
+    },
+    {
+      "epoch": 4.815533980582524,
+      "grad_norm": 0.5620272702074033,
+      "learning_rate": 3.832692507045015e-06,
+      "loss": 0.731,
+      "step": 496
+    },
+    {
+      "epoch": 4.825242718446602,
+      "grad_norm": 0.5060945256107214,
+      "learning_rate": 3.828388788826091e-06,
+      "loss": 0.7914,
+      "step": 497
+    },
+    {
+      "epoch": 4.834951456310679,
+      "grad_norm": 0.5538607078294117,
+      "learning_rate": 3.824079578135442e-06,
+      "loss": 0.6944,
+      "step": 498
+    },
+    {
+      "epoch": 4.844660194174757,
+      "grad_norm": 0.5452852198799888,
+      "learning_rate": 3.819764892790307e-06,
+      "loss": 0.7015,
+      "step": 499
+    },
+    {
+      "epoch": 4.854368932038835,
+      "grad_norm": 0.557748542801134,
+      "learning_rate": 3.815444750630555e-06,
+      "loss": 0.6616,
+      "step": 500
+    },
+    {
+      "epoch": 4.864077669902913,
+      "grad_norm": 0.555226374195002,
+      "learning_rate": 3.811119169518624e-06,
+      "loss": 0.8484,
+      "step": 501
+    },
+    {
+      "epoch": 4.87378640776699,
+      "grad_norm": 0.583053708240317,
+      "learning_rate": 3.8067881673394363e-06,
+      "loss": 1.071,
+      "step": 502
+    },
+    {
+      "epoch": 4.883495145631068,
+      "grad_norm": 0.6078873319472481,
+      "learning_rate": 3.802451762000331e-06,
+      "loss": 0.6256,
+      "step": 503
+    },
+    {
+      "epoch": 4.893203883495145,
+      "grad_norm": 0.539423072387751,
+      "learning_rate": 3.7981099714309856e-06,
+      "loss": 0.5403,
+      "step": 504
+    },
+    {
+      "epoch": 4.902912621359223,
+      "grad_norm": 0.6603698148330428,
+      "learning_rate": 3.7937628135833453e-06,
+      "loss": 0.6906,
+      "step": 505
+    },
+    {
+      "epoch": 4.9126213592233015,
+      "grad_norm": 0.5562121401027199,
+      "learning_rate": 3.7894103064315463e-06,
+      "loss": 0.7039,
+      "step": 506
+    },
+    {
+      "epoch": 4.922330097087379,
+      "grad_norm": 0.5336235803797111,
+      "learning_rate": 3.7850524679718424e-06,
+      "loss": 0.59,
+      "step": 507
+    },
+    {
+      "epoch": 4.932038834951456,
+      "grad_norm": 0.5331917381976123,
+      "learning_rate": 3.7806893162225328e-06,
+      "loss": 0.54,
+      "step": 508
+    },
+    {
+      "epoch": 4.941747572815534,
+      "grad_norm": 0.5058791106293693,
+      "learning_rate": 3.7763208692238818e-06,
+      "loss": 0.6653,
+      "step": 509
+    },
+    {
+      "epoch": 4.951456310679612,
+      "grad_norm": 0.5299074738912896,
+      "learning_rate": 3.7719471450380518e-06,
+      "loss": 0.6194,
+      "step": 510
+    },
+    {
+      "epoch": 4.961165048543689,
+      "grad_norm": 0.5327819159805345,
+      "learning_rate": 3.7675681617490212e-06,
+      "loss": 0.805,
+      "step": 511
+    },
+    {
+      "epoch": 4.970873786407767,
+      "grad_norm": 0.5298660538103491,
+      "learning_rate": 3.7631839374625167e-06,
+      "loss": 0.7929,
+      "step": 512
+    },
+    {
+      "epoch": 4.980582524271845,
+      "grad_norm": 0.4764247185946088,
+      "learning_rate": 3.758794490305932e-06,
+      "loss": 0.611,
+      "step": 513
+    },
+    {
+      "epoch": 4.990291262135923,
+      "grad_norm": 0.5451924892487114,
+      "learning_rate": 3.7543998384282565e-06,
+      "loss": 0.6893,
+      "step": 514
+    },
+    {
+      "epoch": 5.0,
+      "grad_norm": 0.5454441184842063,
+      "learning_rate": 3.7500000000000005e-06,
+      "loss": 0.7546,
+      "step": 515
+    }
+  ],
+  "logging_steps": 1,
+  "max_steps": 1545,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 15,
+  "save_steps": 500,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 24890826055680.0,
+  "train_batch_size": 1,
+  "trial_name": null,
+  "trial_params": null
+}

limo/full/checkpoint-515/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:61d018a068ded862a14e259873b308e3a1b0e8da7491deb0bc57ea90e32e6192
+size 7416

limo/full/checkpoint-515/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)