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
+tokenizer.json filter=lfs diff=lfs merge=lfs -text

added_tokens.json

@@ -0,0 +1,28 @@
+{
+  "</think>": 151668,
+  "</tool_call>": 151658,
+  "</tool_response>": 151666,
+  "<think>": 151667,
+  "<tool_call>": 151657,
+  "<tool_response>": 151665,
+  "<|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
+}

config.json

@@ -0,0 +1,38 @@
+{
+  "architectures": [
+    "Qwen3MoeForCausalLM"
+  ],
+  "attention_bias": false,
+  "attention_dropout": 0.0,
+  "bos_token_id": 151643,
+  "decoder_sparse_step": 1,
+  "eos_token_id": 151643,
+  "head_dim": 128,
+  "hidden_act": "silu",
+  "hidden_size": 2048,
+  "initializer_range": 0.02,
+  "intermediate_size": 6144,
+  "max_position_embeddings": 32768,
+  "max_window_layers": 48,
+  "mlp_only_layers": [],
+  "model_type": "qwen3_moe",
+  "moe_intermediate_size": 768,
+  "norm_topk_prob": true,
+  "num_attention_heads": 32,
+  "num_experts": 128,
+  "num_experts_per_tok": 8,
+  "num_hidden_layers": 48,
+  "num_key_value_heads": 4,
+  "output_router_logits": false,
+  "rms_norm_eps": 1e-06,
+  "rope_scaling": null,
+  "rope_theta": 1000000.0,
+  "router_aux_loss_coef": 0.001,
+  "sliding_window": null,
+  "tie_word_embeddings": false,
+  "torch_dtype": "bfloat16",
+  "transformers_version": "4.51.3",
+  "use_cache": false,
+  "use_sliding_window": false,
+  "vocab_size": 151936
+}

generation_config.json

@@ -0,0 +1,6 @@
+{
+  "bos_token_id": 151643,
+  "eos_token_id": 151643,
+  "max_new_tokens": 2048,
+  "transformers_version": "4.51.3"
+}

latest

@@ -0,0 +1 @@
+global_step261

model-00001-of-00013.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7998ae356bd4a5ae28a627e1f3d7b8664aabe626e51ad201d1b9b66ff5b4a537
+size 4997184968

model-00002-of-00013.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9486371259db2b91fcff80f383adf1132d3b7fc6e5e06a974274367af88576d2
+size 4997741608

model-00003-of-00013.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:aa6c7fc6e307422644f4aba73841afb7d7b761d5bc2363aec51b098dac09b2b6
+size 4997742208

model-00004-of-00013.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bfdcb83a9e5e283a5bc7cb0bf26c1f86463817ee9c10cd6df6b47cd5a4cffcc5
+size 4997743184

model-00005-of-00013.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bd3196281566b3cf61c97f7a414d3ac8aecfe9105d108159fc890dfa76b2b706
+size 4997743184

model-00006-of-00013.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:87dd71f82ceb21d019c1670369ee2b23e032ca1642c9e85db8061444eedbf21a
+size 4997743184

model-00007-of-00013.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:086fc7485939cc5ba03cc0f5c0fcb2d1f9f2f66b8c973e758a7f0ceb6a23b7ef
+size 4997743184

model-00008-of-00013.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6bcf8c405427dd3f1ae3dc662ad1c8ea72125edba03ec4c58e9386e5b9b12375
+size 4997743184

model-00009-of-00013.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a5251fbe2ddbaa62b210cc0ff8850c38d314c00756d67e11424a951c5c42554e
+size 4997743184

model-00010-of-00013.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c2c9c47d4cdb826d5e87eabdeeca3e034848cae0c290e450690a4d6a6f80e2d2
+size 4997743184

model-00011-of-00013.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:35c1abd1ac9630409fad9171ab1497c9518fd768953064106370ce6264d569c4
+size 4997743184

model-00012-of-00013.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bb3cf3f4d49d9fdd50726b71cef58ec32160dad3139c4ddc5213d4248a1d2311
+size 4997743184

model-00013-of-00013.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c7a2fb4a5d3f53a979d0472e19195f74a7ea15c5129134a9f20e535ccd500f14
+size 1094220288

rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:478b41e9f26d338fd8f896e08cad1adab7c423b61f1b45754113bc78d256a3f9
+size 16389

rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ce29a8767a7d907dd24987aa2c3e654d4317f3042fbc13b5b72cadb46d43311a
+size 16389

rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:61a48db011646b4e9a867bf12f4a233cad5dfbfe309686f8996c250196d3783a
+size 16389

rng_state_3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b9562ee822472a4f01dcd6349ab3d1ef42a48915fe3b92e843a0c37db53c8421
+size 16389

rng_state_4.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e7d2767d83c3bf27f12db022b0632e2c4f8c164274ba75e380cf18f9d5f21819
+size 16389

rng_state_5.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:76816358d4e5db8149d60d55234db658d67a13c0c1ce05d7404cf7125a676a5c
+size 16389

rng_state_6.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1562e7520c977d178183d641f70abcf3f57da2489938756cfbebf9b6e6c1a9fd
+size 16389

rng_state_7.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a6b6cabaed045c5398cd1b732f7ec48bd363f3b43cd24e0e70e641a42bd00c28
+size 16389

scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4b4f96b441d64817f399608944699940a0e23d28331baa9957ef3e9a97aecb60
+size 1465

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
+  }
+}

tokenizer.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:aeb13307a71acd8fe81861d94ad54ab689df773318809eed3cbe794b4492dae4
+size 11422654

tokenizer_config.json

@@ -0,0 +1,241 @@
+{
+  "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
+    },
+    "151665": {
+      "content": "<tool_response>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151666": {
+      "content": "</tool_response>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151667": {
+      "content": "<think>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151668": {
+      "content": "</think>",
+      "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\\n' }}\n    {%- endif %}\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    {%- endif %}\n{%- endif %}\n{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}\n{%- for message in messages[::-1] %}\n    {%- set index = (messages|length - 1) - loop.index0 %}\n    {%- if ns.multi_step_tool and message.role == \"user\" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}\n        {%- set ns.multi_step_tool = false %}\n        {%- set ns.last_query_index = index %}\n    {%- endif %}\n{%- endfor %}\n{%- for message in messages %}\n    {%- if (message.role == \"user\") or (message.role == \"system\" and not loop.first) %}\n        {{- '<|im_start|>' + message.role + '\\n' + message.content + '<|im_end|>' + '\\n' }}\n    {%- elif message.role == \"assistant\" %}\n        {%- set content = message.content %}\n        {%- set reasoning_content = '' %}\n        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}\n            {%- set reasoning_content = message.reasoning_content %}\n        {%- else %}\n            {%- if '</think>' in message.content %}\n                {%- set content = message.content.split('</think>')[-1].lstrip('\\n') %}\n                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\\n').split('<think>')[-1].lstrip('\\n') %}\n            {%- endif %}\n        {%- endif %}\n        {%- if loop.index0 > ns.last_query_index %}\n            {%- if loop.last or (not loop.last and reasoning_content) %}\n                {{- '<|im_start|>' + message.role + '\\n<think>\\n' + reasoning_content.strip('\\n') + '\\n</think>\\n\\n' + content.lstrip('\\n') }}\n            {%- else %}\n                {{- '<|im_start|>' + message.role + '\\n' + content }}\n            {%- endif %}\n        {%- else %}\n            {{- '<|im_start|>' + message.role + '\\n' + content }}\n        {%- endif %}\n        {%- if message.tool_calls %}\n            {%- for tool_call in message.tool_calls %}\n                {%- if (loop.first and content) or (not loop.first) %}\n                    {{- '\\n' }}\n                {%- endif %}\n                {%- if tool_call.function %}\n                    {%- set tool_call = tool_call.function %}\n                {%- endif %}\n                {{- '<tool_call>\\n{\"name\": \"' }}\n                {{- tool_call.name }}\n                {{- '\", \"arguments\": ' }}\n                {%- if tool_call.arguments is string %}\n                    {{- tool_call.arguments }}\n                {%- else %}\n                    {{- tool_call.arguments | tojson }}\n                {%- endif %}\n                {{- '}\\n</tool_call>' }}\n            {%- endfor %}\n        {%- endif %}\n        {{- '<|im_end|>\\n' }}\n    {%- elif message.role == \"tool\" %}\n        {%- if loop.first 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    {%- if enable_thinking is defined and enable_thinking is false %}\n        {{- '<think>\\n\\n</think>\\n\\n' }}\n    {%- endif %}\n{%- endif %}",
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|im_end|>",
+  "errors": "replace",
+  "extra_special_tokens": {},
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "padding_side": "right",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}

trainer_state.json

@@ -0,0 +1,1861 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.16704,
+  "eval_steps": 500,
+  "global_step": 261,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.00064,
+      "grad_norm": 2.3657337561838188,
+      "learning_rate": 0.0,
+      "loss": 0.5764,
+      "step": 1
+    },
+    {
+      "epoch": 0.00128,
+      "grad_norm": 2.577238188562681,
+      "learning_rate": 6.329113924050633e-07,
+      "loss": 0.6154,
+      "step": 2
+    },
+    {
+      "epoch": 0.00192,
+      "grad_norm": 2.486400760684678,
+      "learning_rate": 1.2658227848101265e-06,
+      "loss": 0.5765,
+      "step": 3
+    },
+    {
+      "epoch": 0.00256,
+      "grad_norm": 2.427317708214583,
+      "learning_rate": 1.8987341772151901e-06,
+      "loss": 0.5898,
+      "step": 4
+    },
+    {
+      "epoch": 0.0032,
+      "grad_norm": 2.3689778807783104,
+      "learning_rate": 2.531645569620253e-06,
+      "loss": 0.5846,
+      "step": 5
+    },
+    {
+      "epoch": 0.00384,
+      "grad_norm": 2.533864042007203,
+      "learning_rate": 3.1645569620253167e-06,
+      "loss": 0.6002,
+      "step": 6
+    },
+    {
+      "epoch": 0.00448,
+      "grad_norm": 1.879595075845785,
+      "learning_rate": 3.7974683544303802e-06,
+      "loss": 0.5581,
+      "step": 7
+    },
+    {
+      "epoch": 0.00512,
+      "grad_norm": 2.056563531274739,
+      "learning_rate": 4.430379746835443e-06,
+      "loss": 0.5859,
+      "step": 8
+    },
+    {
+      "epoch": 0.00576,
+      "grad_norm": 1.4703684433180533,
+      "learning_rate": 5.063291139240506e-06,
+      "loss": 0.5479,
+      "step": 9
+    },
+    {
+      "epoch": 0.0064,
+      "grad_norm": 1.2614967629103893,
+      "learning_rate": 5.69620253164557e-06,
+      "loss": 0.5303,
+      "step": 10
+    },
+    {
+      "epoch": 0.00704,
+      "grad_norm": 1.1460599905324345,
+      "learning_rate": 6.329113924050633e-06,
+      "loss": 0.5568,
+      "step": 11
+    },
+    {
+      "epoch": 0.00768,
+      "grad_norm": 1.6204644872566638,
+      "learning_rate": 6.9620253164556965e-06,
+      "loss": 0.5111,
+      "step": 12
+    },
+    {
+      "epoch": 0.00832,
+      "grad_norm": 1.6361395875225335,
+      "learning_rate": 7.5949367088607605e-06,
+      "loss": 0.4678,
+      "step": 13
+    },
+    {
+      "epoch": 0.00896,
+      "grad_norm": 1.6402122162855226,
+      "learning_rate": 8.227848101265822e-06,
+      "loss": 0.4811,
+      "step": 14
+    },
+    {
+      "epoch": 0.0096,
+      "grad_norm": 1.2490639825476322,
+      "learning_rate": 8.860759493670886e-06,
+      "loss": 0.442,
+      "step": 15
+    },
+    {
+      "epoch": 0.01024,
+      "grad_norm": 1.0588915477376701,
+      "learning_rate": 9.49367088607595e-06,
+      "loss": 0.4263,
+      "step": 16
+    },
+    {
+      "epoch": 0.01088,
+      "grad_norm": 1.40604786962106,
+      "learning_rate": 1.0126582278481012e-05,
+      "loss": 0.4549,
+      "step": 17
+    },
+    {
+      "epoch": 0.01152,
+      "grad_norm": 1.242365760774633,
+      "learning_rate": 1.0759493670886076e-05,
+      "loss": 0.4531,
+      "step": 18
+    },
+    {
+      "epoch": 0.01216,
+      "grad_norm": 0.9396101260126887,
+      "learning_rate": 1.139240506329114e-05,
+      "loss": 0.415,
+      "step": 19
+    },
+    {
+      "epoch": 0.0128,
+      "grad_norm": 0.7578271337493023,
+      "learning_rate": 1.2025316455696203e-05,
+      "loss": 0.4268,
+      "step": 20
+    },
+    {
+      "epoch": 0.01344,
+      "grad_norm": 0.7671817679988738,
+      "learning_rate": 1.2658227848101267e-05,
+      "loss": 0.4314,
+      "step": 21
+    },
+    {
+      "epoch": 0.01408,
+      "grad_norm": 0.6494114557426621,
+      "learning_rate": 1.3291139240506329e-05,
+      "loss": 0.3666,
+      "step": 22
+    },
+    {
+      "epoch": 0.01472,
+      "grad_norm": 0.6806456497077759,
+      "learning_rate": 1.3924050632911393e-05,
+      "loss": 0.3823,
+      "step": 23
+    },
+    {
+      "epoch": 0.01536,
+      "grad_norm": 0.6595430154008608,
+      "learning_rate": 1.4556962025316457e-05,
+      "loss": 0.3529,
+      "step": 24
+    },
+    {
+      "epoch": 0.016,
+      "grad_norm": 0.7689071362735744,
+      "learning_rate": 1.5189873417721521e-05,
+      "loss": 0.3932,
+      "step": 25
+    },
+    {
+      "epoch": 0.01664,
+      "grad_norm": 0.6854685982052705,
+      "learning_rate": 1.5822784810126583e-05,
+      "loss": 0.3738,
+      "step": 26
+    },
+    {
+      "epoch": 0.01728,
+      "grad_norm": 0.558852717411362,
+      "learning_rate": 1.6455696202531644e-05,
+      "loss": 0.3708,
+      "step": 27
+    },
+    {
+      "epoch": 0.01792,
+      "grad_norm": 0.49876042518765007,
+      "learning_rate": 1.7088607594936708e-05,
+      "loss": 0.3696,
+      "step": 28
+    },
+    {
+      "epoch": 0.01856,
+      "grad_norm": 0.5271567613623678,
+      "learning_rate": 1.7721518987341772e-05,
+      "loss": 0.3558,
+      "step": 29
+    },
+    {
+      "epoch": 0.0192,
+      "grad_norm": 0.5978920045933569,
+      "learning_rate": 1.8354430379746836e-05,
+      "loss": 0.3827,
+      "step": 30
+    },
+    {
+      "epoch": 0.01984,
+      "grad_norm": 0.4975313988318536,
+      "learning_rate": 1.89873417721519e-05,
+      "loss": 0.3772,
+      "step": 31
+    },
+    {
+      "epoch": 0.02048,
+      "grad_norm": 0.5766330443378189,
+      "learning_rate": 1.962025316455696e-05,
+      "loss": 0.3832,
+      "step": 32
+    },
+    {
+      "epoch": 0.02112,
+      "grad_norm": 0.5567055560538855,
+      "learning_rate": 2.0253164556962025e-05,
+      "loss": 0.3477,
+      "step": 33
+    },
+    {
+      "epoch": 0.02176,
+      "grad_norm": 0.46871971492473,
+      "learning_rate": 2.088607594936709e-05,
+      "loss": 0.3604,
+      "step": 34
+    },
+    {
+      "epoch": 0.0224,
+      "grad_norm": 0.43316810660778116,
+      "learning_rate": 2.1518987341772153e-05,
+      "loss": 0.3592,
+      "step": 35
+    },
+    {
+      "epoch": 0.02304,
+      "grad_norm": 0.4312186195231089,
+      "learning_rate": 2.2151898734177217e-05,
+      "loss": 0.3711,
+      "step": 36
+    },
+    {
+      "epoch": 0.02368,
+      "grad_norm": 0.46951838290882675,
+      "learning_rate": 2.278481012658228e-05,
+      "loss": 0.3618,
+      "step": 37
+    },
+    {
+      "epoch": 0.02432,
+      "grad_norm": 0.468516818608629,
+      "learning_rate": 2.341772151898734e-05,
+      "loss": 0.3578,
+      "step": 38
+    },
+    {
+      "epoch": 0.02496,
+      "grad_norm": 0.45909911923620816,
+      "learning_rate": 2.4050632911392405e-05,
+      "loss": 0.3521,
+      "step": 39
+    },
+    {
+      "epoch": 0.0256,
+      "grad_norm": 0.36223572387755915,
+      "learning_rate": 2.468354430379747e-05,
+      "loss": 0.3439,
+      "step": 40
+    },
+    {
+      "epoch": 0.02624,
+      "grad_norm": 0.41920026035277985,
+      "learning_rate": 2.5316455696202533e-05,
+      "loss": 0.3052,
+      "step": 41
+    },
+    {
+      "epoch": 0.02688,
+      "grad_norm": 0.46410873807902825,
+      "learning_rate": 2.5949367088607597e-05,
+      "loss": 0.3324,
+      "step": 42
+    },
+    {
+      "epoch": 0.02752,
+      "grad_norm": 0.44804958203514283,
+      "learning_rate": 2.6582278481012658e-05,
+      "loss": 0.322,
+      "step": 43
+    },
+    {
+      "epoch": 0.02816,
+      "grad_norm": 0.4408934955100543,
+      "learning_rate": 2.7215189873417722e-05,
+      "loss": 0.3399,
+      "step": 44
+    },
+    {
+      "epoch": 0.0288,
+      "grad_norm": 0.4536362310706478,
+      "learning_rate": 2.7848101265822786e-05,
+      "loss": 0.3508,
+      "step": 45
+    },
+    {
+      "epoch": 0.02944,
+      "grad_norm": 0.4735438789842479,
+      "learning_rate": 2.848101265822785e-05,
+      "loss": 0.3431,
+      "step": 46
+    },
+    {
+      "epoch": 0.03008,
+      "grad_norm": 0.3960360490220664,
+      "learning_rate": 2.9113924050632914e-05,
+      "loss": 0.3358,
+      "step": 47
+    },
+    {
+      "epoch": 0.03072,
+      "grad_norm": 0.4449245975582831,
+      "learning_rate": 2.9746835443037974e-05,
+      "loss": 0.3472,
+      "step": 48
+    },
+    {
+      "epoch": 0.03136,
+      "grad_norm": 0.41598379019590886,
+      "learning_rate": 3.0379746835443042e-05,
+      "loss": 0.3295,
+      "step": 49
+    },
+    {
+      "epoch": 0.032,
+      "grad_norm": 0.39770652562313213,
+      "learning_rate": 3.10126582278481e-05,
+      "loss": 0.339,
+      "step": 50
+    },
+    {
+      "epoch": 0.03264,
+      "grad_norm": 0.4036450937079526,
+      "learning_rate": 3.1645569620253167e-05,
+      "loss": 0.326,
+      "step": 51
+    },
+    {
+      "epoch": 0.03328,
+      "grad_norm": 0.43082778666471144,
+      "learning_rate": 3.227848101265823e-05,
+      "loss": 0.3258,
+      "step": 52
+    },
+    {
+      "epoch": 0.03392,
+      "grad_norm": 0.402275292978643,
+      "learning_rate": 3.291139240506329e-05,
+      "loss": 0.2905,
+      "step": 53
+    },
+    {
+      "epoch": 0.03456,
+      "grad_norm": 0.43612618786182106,
+      "learning_rate": 3.354430379746836e-05,
+      "loss": 0.334,
+      "step": 54
+    },
+    {
+      "epoch": 0.0352,
+      "grad_norm": 0.45984837714015897,
+      "learning_rate": 3.4177215189873416e-05,
+      "loss": 0.3283,
+      "step": 55
+    },
+    {
+      "epoch": 0.03584,
+      "grad_norm": 0.4484461315187549,
+      "learning_rate": 3.4810126582278487e-05,
+      "loss": 0.3277,
+      "step": 56
+    },
+    {
+      "epoch": 0.03648,
+      "grad_norm": 0.4052074034352622,
+      "learning_rate": 3.5443037974683544e-05,
+      "loss": 0.3271,
+      "step": 57
+    },
+    {
+      "epoch": 0.03712,
+      "grad_norm": 0.46620186767863914,
+      "learning_rate": 3.607594936708861e-05,
+      "loss": 0.3464,
+      "step": 58
+    },
+    {
+      "epoch": 0.03776,
+      "grad_norm": 0.42593292833096424,
+      "learning_rate": 3.670886075949367e-05,
+      "loss": 0.3331,
+      "step": 59
+    },
+    {
+      "epoch": 0.0384,
+      "grad_norm": 0.4815687673225032,
+      "learning_rate": 3.7341772151898736e-05,
+      "loss": 0.3269,
+      "step": 60
+    },
+    {
+      "epoch": 0.03904,
+      "grad_norm": 0.455342196474994,
+      "learning_rate": 3.79746835443038e-05,
+      "loss": 0.3343,
+      "step": 61
+    },
+    {
+      "epoch": 0.03968,
+      "grad_norm": 0.4316078425366754,
+      "learning_rate": 3.8607594936708864e-05,
+      "loss": 0.3145,
+      "step": 62
+    },
+    {
+      "epoch": 0.04032,
+      "grad_norm": 0.4604468007808017,
+      "learning_rate": 3.924050632911392e-05,
+      "loss": 0.3206,
+      "step": 63
+    },
+    {
+      "epoch": 0.04096,
+      "grad_norm": 0.516259514156323,
+      "learning_rate": 3.987341772151899e-05,
+      "loss": 0.3538,
+      "step": 64
+    },
+    {
+      "epoch": 0.0416,
+      "grad_norm": 0.3911206186305256,
+      "learning_rate": 4.050632911392405e-05,
+      "loss": 0.3116,
+      "step": 65
+    },
+    {
+      "epoch": 0.04224,
+      "grad_norm": 0.5059680673369054,
+      "learning_rate": 4.113924050632912e-05,
+      "loss": 0.3461,
+      "step": 66
+    },
+    {
+      "epoch": 0.04288,
+      "grad_norm": 0.4477077725372915,
+      "learning_rate": 4.177215189873418e-05,
+      "loss": 0.3142,
+      "step": 67
+    },
+    {
+      "epoch": 0.04352,
+      "grad_norm": 0.4568181894979607,
+      "learning_rate": 4.240506329113924e-05,
+      "loss": 0.3258,
+      "step": 68
+    },
+    {
+      "epoch": 0.04416,
+      "grad_norm": 0.436426112682463,
+      "learning_rate": 4.3037974683544305e-05,
+      "loss": 0.2942,
+      "step": 69
+    },
+    {
+      "epoch": 0.0448,
+      "grad_norm": 0.4130845172287253,
+      "learning_rate": 4.367088607594937e-05,
+      "loss": 0.3293,
+      "step": 70
+    },
+    {
+      "epoch": 0.04544,
+      "grad_norm": 0.4907470169969436,
+      "learning_rate": 4.430379746835443e-05,
+      "loss": 0.3152,
+      "step": 71
+    },
+    {
+      "epoch": 0.04608,
+      "grad_norm": 0.44557023947548996,
+      "learning_rate": 4.49367088607595e-05,
+      "loss": 0.323,
+      "step": 72
+    },
+    {
+      "epoch": 0.04672,
+      "grad_norm": 0.43347032964452825,
+      "learning_rate": 4.556962025316456e-05,
+      "loss": 0.3132,
+      "step": 73
+    },
+    {
+      "epoch": 0.04736,
+      "grad_norm": 0.41315778348357357,
+      "learning_rate": 4.6202531645569625e-05,
+      "loss": 0.3076,
+      "step": 74
+    },
+    {
+      "epoch": 0.048,
+      "grad_norm": 0.38689299616228223,
+      "learning_rate": 4.683544303797468e-05,
+      "loss": 0.2971,
+      "step": 75
+    },
+    {
+      "epoch": 0.04864,
+      "grad_norm": 0.6649393785343345,
+      "learning_rate": 4.7468354430379746e-05,
+      "loss": 0.3196,
+      "step": 76
+    },
+    {
+      "epoch": 0.04928,
+      "grad_norm": 0.4741860571119842,
+      "learning_rate": 4.810126582278481e-05,
+      "loss": 0.3261,
+      "step": 77
+    },
+    {
+      "epoch": 0.04992,
+      "grad_norm": 0.39850926668653464,
+      "learning_rate": 4.8734177215189874e-05,
+      "loss": 0.2909,
+      "step": 78
+    },
+    {
+      "epoch": 0.05056,
+      "grad_norm": 0.45993342960946926,
+      "learning_rate": 4.936708860759494e-05,
+      "loss": 0.3232,
+      "step": 79
+    },
+    {
+      "epoch": 0.0512,
+      "grad_norm": 0.41758744126936076,
+      "learning_rate": 5e-05,
+      "loss": 0.3138,
+      "step": 80
+    },
+    {
+      "epoch": 0.05184,
+      "grad_norm": 0.4190137054884969,
+      "learning_rate": 4.999994390459423e-05,
+      "loss": 0.324,
+      "step": 81
+    },
+    {
+      "epoch": 0.05248,
+      "grad_norm": 0.38007532443802017,
+      "learning_rate": 4.999977561862863e-05,
+      "loss": 0.3167,
+      "step": 82
+    },
+    {
+      "epoch": 0.05312,
+      "grad_norm": 0.3894813419979866,
+      "learning_rate": 4.999949514285842e-05,
+      "loss": 0.3169,
+      "step": 83
+    },
+    {
+      "epoch": 0.05376,
+      "grad_norm": 0.4646671073219221,
+      "learning_rate": 4.999910247854227e-05,
+      "loss": 0.3295,
+      "step": 84
+    },
+    {
+      "epoch": 0.0544,
+      "grad_norm": 0.46440352293136283,
+      "learning_rate": 4.99985976274423e-05,
+      "loss": 0.3259,
+      "step": 85
+    },
+    {
+      "epoch": 0.05504,
+      "grad_norm": 0.4754742191954697,
+      "learning_rate": 4.9997980591824126e-05,
+      "loss": 0.3073,
+      "step": 86
+    },
+    {
+      "epoch": 0.05568,
+      "grad_norm": 0.4102331568111904,
+      "learning_rate": 4.999725137445674e-05,
+      "loss": 0.3179,
+      "step": 87
+    },
+    {
+      "epoch": 0.05632,
+      "grad_norm": 0.49865407563428993,
+      "learning_rate": 4.999640997861264e-05,
+      "loss": 0.3198,
+      "step": 88
+    },
+    {
+      "epoch": 0.05696,
+      "grad_norm": 0.450995150648752,
+      "learning_rate": 4.999545640806766e-05,
+      "loss": 0.3215,
+      "step": 89
+    },
+    {
+      "epoch": 0.0576,
+      "grad_norm": 0.42715082345561,
+      "learning_rate": 4.999439066710111e-05,
+      "loss": 0.3406,
+      "step": 90
+    },
+    {
+      "epoch": 0.05824,
+      "grad_norm": 0.47007665924261693,
+      "learning_rate": 4.999321276049563e-05,
+      "loss": 0.3169,
+      "step": 91
+    },
+    {
+      "epoch": 0.05888,
+      "grad_norm": 0.37692477648671757,
+      "learning_rate": 4.9991922693537226e-05,
+      "loss": 0.3052,
+      "step": 92
+    },
+    {
+      "epoch": 0.05952,
+      "grad_norm": 0.5254380682221013,
+      "learning_rate": 4.9990520472015246e-05,
+      "loss": 0.2928,
+      "step": 93
+    },
+    {
+      "epoch": 0.06016,
+      "grad_norm": 0.4524991571079602,
+      "learning_rate": 4.998900610222236e-05,
+      "loss": 0.3035,
+      "step": 94
+    },
+    {
+      "epoch": 0.0608,
+      "grad_norm": 0.39419071585883453,
+      "learning_rate": 4.998737959095449e-05,
+      "loss": 0.2764,
+      "step": 95
+    },
+    {
+      "epoch": 0.06144,
+      "grad_norm": 0.44021580096048935,
+      "learning_rate": 4.998564094551081e-05,
+      "loss": 0.3137,
+      "step": 96
+    },
+    {
+      "epoch": 0.06208,
+      "grad_norm": 0.4397078772822271,
+      "learning_rate": 4.998379017369375e-05,
+      "loss": 0.3092,
+      "step": 97
+    },
+    {
+      "epoch": 0.06272,
+      "grad_norm": 0.4612518337175716,
+      "learning_rate": 4.9981827283808855e-05,
+      "loss": 0.3065,
+      "step": 98
+    },
+    {
+      "epoch": 0.06336,
+      "grad_norm": 0.4491229684653282,
+      "learning_rate": 4.997975228466489e-05,
+      "loss": 0.3327,
+      "step": 99
+    },
+    {
+      "epoch": 0.064,
+      "grad_norm": 0.40909998580765855,
+      "learning_rate": 4.997756518557368e-05,
+      "loss": 0.2787,
+      "step": 100
+    },
+    {
+      "epoch": 0.06464,
+      "grad_norm": 0.3785712105730604,
+      "learning_rate": 4.997526599635011e-05,
+      "loss": 0.2837,
+      "step": 101
+    },
+    {
+      "epoch": 0.06528,
+      "grad_norm": 0.5009245567404541,
+      "learning_rate": 4.997285472731211e-05,
+      "loss": 0.3325,
+      "step": 102
+    },
+    {
+      "epoch": 0.06592,
+      "grad_norm": 0.41940515716785753,
+      "learning_rate": 4.997033138928056e-05,
+      "loss": 0.3126,
+      "step": 103
+    },
+    {
+      "epoch": 0.06656,
+      "grad_norm": 0.4356254674524969,
+      "learning_rate": 4.996769599357927e-05,
+      "loss": 0.3079,
+      "step": 104
+    },
+    {
+      "epoch": 0.0672,
+      "grad_norm": 0.4485794513852225,
+      "learning_rate": 4.996494855203493e-05,
+      "loss": 0.3227,
+      "step": 105
+    },
+    {
+      "epoch": 0.06784,
+      "grad_norm": 0.5339271359603054,
+      "learning_rate": 4.996208907697706e-05,
+      "loss": 0.3197,
+      "step": 106
+    },
+    {
+      "epoch": 0.06848,
+      "grad_norm": 0.47913200680485823,
+      "learning_rate": 4.995911758123791e-05,
+      "loss": 0.2829,
+      "step": 107
+    },
+    {
+      "epoch": 0.06912,
+      "grad_norm": 0.48619705540043057,
+      "learning_rate": 4.995603407815249e-05,
+      "loss": 0.295,
+      "step": 108
+    },
+    {
+      "epoch": 0.06976,
+      "grad_norm": 0.44791381540090275,
+      "learning_rate": 4.9952838581558406e-05,
+      "loss": 0.3009,
+      "step": 109
+    },
+    {
+      "epoch": 0.0704,
+      "grad_norm": 0.4217427510368746,
+      "learning_rate": 4.994953110579588e-05,
+      "loss": 0.3124,
+      "step": 110
+    },
+    {
+      "epoch": 0.07104,
+      "grad_norm": 0.44605995933104997,
+      "learning_rate": 4.9946111665707645e-05,
+      "loss": 0.3232,
+      "step": 111
+    },
+    {
+      "epoch": 0.07168,
+      "grad_norm": 0.3580173680896881,
+      "learning_rate": 4.99425802766389e-05,
+      "loss": 0.288,
+      "step": 112
+    },
+    {
+      "epoch": 0.07232,
+      "grad_norm": 0.4281198427243427,
+      "learning_rate": 4.993893695443721e-05,
+      "loss": 0.3008,
+      "step": 113
+    },
+    {
+      "epoch": 0.07296,
+      "grad_norm": 0.3748012845273175,
+      "learning_rate": 4.993518171545247e-05,
+      "loss": 0.2898,
+      "step": 114
+    },
+    {
+      "epoch": 0.0736,
+      "grad_norm": 0.41133247941906176,
+      "learning_rate": 4.993131457653682e-05,
+      "loss": 0.3007,
+      "step": 115
+    },
+    {
+      "epoch": 0.07424,
+      "grad_norm": 0.4049279526177947,
+      "learning_rate": 4.992733555504454e-05,
+      "loss": 0.3156,
+      "step": 116
+    },
+    {
+      "epoch": 0.07488,
+      "grad_norm": 0.3907854290589919,
+      "learning_rate": 4.9923244668832025e-05,
+      "loss": 0.305,
+      "step": 117
+    },
+    {
+      "epoch": 0.07552,
+      "grad_norm": 3.3672297158831523,
+      "learning_rate": 4.991904193625768e-05,
+      "loss": 0.3099,
+      "step": 118
+    },
+    {
+      "epoch": 0.07616,
+      "grad_norm": 0.45896342700641674,
+      "learning_rate": 4.991472737618181e-05,
+      "loss": 0.2766,
+      "step": 119
+    },
+    {
+      "epoch": 0.0768,
+      "grad_norm": 0.41249483633287526,
+      "learning_rate": 4.991030100796659e-05,
+      "loss": 0.3026,
+      "step": 120
+    },
+    {
+      "epoch": 0.07744,
+      "grad_norm": 0.39435147330028136,
+      "learning_rate": 4.9905762851475915e-05,
+      "loss": 0.3151,
+      "step": 121
+    },
+    {
+      "epoch": 0.07808,
+      "grad_norm": 0.38174217860012233,
+      "learning_rate": 4.990111292707537e-05,
+      "loss": 0.3132,
+      "step": 122
+    },
+    {
+      "epoch": 0.07872,
+      "grad_norm": 0.4106762400890441,
+      "learning_rate": 4.989635125563211e-05,
+      "loss": 0.2883,
+      "step": 123
+    },
+    {
+      "epoch": 0.07936,
+      "grad_norm": 0.3777494709551966,
+      "learning_rate": 4.989147785851477e-05,
+      "loss": 0.3075,
+      "step": 124
+    },
+    {
+      "epoch": 0.08,
+      "grad_norm": 0.3812281839176885,
+      "learning_rate": 4.988649275759335e-05,
+      "loss": 0.3173,
+      "step": 125
+    },
+    {
+      "epoch": 0.08064,
+      "grad_norm": 0.3800442807387135,
+      "learning_rate": 4.988139597523916e-05,
+      "loss": 0.2897,
+      "step": 126
+    },
+    {
+      "epoch": 0.08128,
+      "grad_norm": 0.4101363073522896,
+      "learning_rate": 4.98761875343247e-05,
+      "loss": 0.2822,
+      "step": 127
+    },
+    {
+      "epoch": 0.08192,
+      "grad_norm": 0.3515414786001439,
+      "learning_rate": 4.987086745822352e-05,
+      "loss": 0.2831,
+      "step": 128
+    },
+    {
+      "epoch": 0.08256,
+      "grad_norm": 0.38143533758711873,
+      "learning_rate": 4.986543577081017e-05,
+      "loss": 0.3193,
+      "step": 129
+    },
+    {
+      "epoch": 0.0832,
+      "grad_norm": 0.44066201212578954,
+      "learning_rate": 4.985989249646007e-05,
+      "loss": 0.3006,
+      "step": 130
+    },
+    {
+      "epoch": 0.08384,
+      "grad_norm": 0.3699946189372801,
+      "learning_rate": 4.985423766004939e-05,
+      "loss": 0.309,
+      "step": 131
+    },
+    {
+      "epoch": 0.08448,
+      "grad_norm": 0.4732003041900607,
+      "learning_rate": 4.9848471286954965e-05,
+      "loss": 0.3253,
+      "step": 132
+    },
+    {
+      "epoch": 0.08512,
+      "grad_norm": 0.35724915771132804,
+      "learning_rate": 4.984259340305416e-05,
+      "loss": 0.2876,
+      "step": 133
+    },
+    {
+      "epoch": 0.08576,
+      "grad_norm": 0.39925687072469707,
+      "learning_rate": 4.9836604034724753e-05,
+      "loss": 0.2906,
+      "step": 134
+    },
+    {
+      "epoch": 0.0864,
+      "grad_norm": 0.43582991089095635,
+      "learning_rate": 4.9830503208844835e-05,
+      "loss": 0.2928,
+      "step": 135
+    },
+    {
+      "epoch": 0.08704,
+      "grad_norm": 0.3638802742085539,
+      "learning_rate": 4.982429095279266e-05,
+      "loss": 0.2924,
+      "step": 136
+    },
+    {
+      "epoch": 0.08768,
+      "grad_norm": 0.44414884826796713,
+      "learning_rate": 4.981796729444655e-05,
+      "loss": 0.3225,
+      "step": 137
+    },
+    {
+      "epoch": 0.08832,
+      "grad_norm": 0.4179539121719692,
+      "learning_rate": 4.981153226218477e-05,
+      "loss": 0.3067,
+      "step": 138
+    },
+    {
+      "epoch": 0.08896,
+      "grad_norm": 0.3518602463362222,
+      "learning_rate": 4.9804985884885366e-05,
+      "loss": 0.2854,
+      "step": 139
+    },
+    {
+      "epoch": 0.0896,
+      "grad_norm": 0.4013030661247676,
+      "learning_rate": 4.979832819192608e-05,
+      "loss": 0.3113,
+      "step": 140
+    },
+    {
+      "epoch": 0.09024,
+      "grad_norm": 0.3700912839771966,
+      "learning_rate": 4.97915592131842e-05,
+      "loss": 0.2873,
+      "step": 141
+    },
+    {
+      "epoch": 0.09088,
+      "grad_norm": 0.41429834446700914,
+      "learning_rate": 4.97846789790364e-05,
+      "loss": 0.3064,
+      "step": 142
+    },
+    {
+      "epoch": 0.09152,
+      "grad_norm": 0.3743350290044281,
+      "learning_rate": 4.977768752035865e-05,
+      "loss": 0.311,
+      "step": 143
+    },
+    {
+      "epoch": 0.09216,
+      "grad_norm": 0.4436631708270473,
+      "learning_rate": 4.9770584868526046e-05,
+      "loss": 0.3105,
+      "step": 144
+    },
+    {
+      "epoch": 0.0928,
+      "grad_norm": 0.4341903366251428,
+      "learning_rate": 4.9763371055412685e-05,
+      "loss": 0.307,
+      "step": 145
+    },
+    {
+      "epoch": 0.09344,
+      "grad_norm": 0.4147022361169306,
+      "learning_rate": 4.9756046113391496e-05,
+      "loss": 0.3288,
+      "step": 146
+    },
+    {
+      "epoch": 0.09408,
+      "grad_norm": 0.3884970059580215,
+      "learning_rate": 4.974861007533413e-05,
+      "loss": 0.2847,
+      "step": 147
+    },
+    {
+      "epoch": 0.09472,
+      "grad_norm": 0.4258271028958422,
+      "learning_rate": 4.97410629746108e-05,
+      "loss": 0.2908,
+      "step": 148
+    },
+    {
+      "epoch": 0.09536,
+      "grad_norm": 0.41527352386736543,
+      "learning_rate": 4.9733404845090116e-05,
+      "loss": 0.2828,
+      "step": 149
+    },
+    {
+      "epoch": 0.096,
+      "grad_norm": 0.41352693469232554,
+      "learning_rate": 4.972563572113894e-05,
+      "loss": 0.2895,
+      "step": 150
+    },
+    {
+      "epoch": 0.09664,
+      "grad_norm": 0.4391732737160089,
+      "learning_rate": 4.971775563762226e-05,
+      "loss": 0.2908,
+      "step": 151
+    },
+    {
+      "epoch": 0.09728,
+      "grad_norm": 0.42637768991672836,
+      "learning_rate": 4.970976462990299e-05,
+      "loss": 0.3103,
+      "step": 152
+    },
+    {
+      "epoch": 0.09792,
+      "grad_norm": 0.38123935717217666,
+      "learning_rate": 4.970166273384183e-05,
+      "loss": 0.2859,
+      "step": 153
+    },
+    {
+      "epoch": 0.09856,
+      "grad_norm": 0.3432072306652636,
+      "learning_rate": 4.969344998579712e-05,
+      "loss": 0.2653,
+      "step": 154
+    },
+    {
+      "epoch": 0.0992,
+      "grad_norm": 0.38378318854854493,
+      "learning_rate": 4.9685126422624644e-05,
+      "loss": 0.3135,
+      "step": 155
+    },
+    {
+      "epoch": 0.09984,
+      "grad_norm": 0.3871767364493639,
+      "learning_rate": 4.96766920816775e-05,
+      "loss": 0.3129,
+      "step": 156
+    },
+    {
+      "epoch": 0.10048,
+      "grad_norm": 0.361665333999729,
+      "learning_rate": 4.966814700080592e-05,
+      "loss": 0.3175,
+      "step": 157
+    },
+    {
+      "epoch": 0.10112,
+      "grad_norm": 0.394466639141269,
+      "learning_rate": 4.965949121835707e-05,
+      "loss": 0.333,
+      "step": 158
+    },
+    {
+      "epoch": 0.10176,
+      "grad_norm": 0.396292343881353,
+      "learning_rate": 4.965072477317493e-05,
+      "loss": 0.3084,
+      "step": 159
+    },
+    {
+      "epoch": 0.1024,
+      "grad_norm": 0.3468981265806776,
+      "learning_rate": 4.964184770460007e-05,
+      "loss": 0.2927,
+      "step": 160
+    },
+    {
+      "epoch": 0.10304,
+      "grad_norm": 0.39865168519337385,
+      "learning_rate": 4.963286005246953e-05,
+      "loss": 0.2815,
+      "step": 161
+    },
+    {
+      "epoch": 0.10368,
+      "grad_norm": 0.37451640530098157,
+      "learning_rate": 4.9623761857116594e-05,
+      "loss": 0.3188,
+      "step": 162
+    },
+    {
+      "epoch": 0.10432,
+      "grad_norm": 0.37483826123841657,
+      "learning_rate": 4.961455315937059e-05,
+      "loss": 0.2813,
+      "step": 163
+    },
+    {
+      "epoch": 0.10496,
+      "grad_norm": 0.33130533329924106,
+      "learning_rate": 4.960523400055679e-05,
+      "loss": 0.2924,
+      "step": 164
+    },
+    {
+      "epoch": 0.1056,
+      "grad_norm": 0.3364997696428208,
+      "learning_rate": 4.959580442249615e-05,
+      "loss": 0.3001,
+      "step": 165
+    },
+    {
+      "epoch": 0.10624,
+      "grad_norm": 0.3183485250122132,
+      "learning_rate": 4.958626446750515e-05,
+      "loss": 0.3195,
+      "step": 166
+    },
+    {
+      "epoch": 0.10688,
+      "grad_norm": 0.3417933668875517,
+      "learning_rate": 4.9576614178395595e-05,
+      "loss": 0.2991,
+      "step": 167
+    },
+    {
+      "epoch": 0.10752,
+      "grad_norm": 0.3487109852934047,
+      "learning_rate": 4.956685359847445e-05,
+      "loss": 0.2959,
+      "step": 168
+    },
+    {
+      "epoch": 0.10816,
+      "grad_norm": 0.31315058374358634,
+      "learning_rate": 4.95569827715436e-05,
+      "loss": 0.2799,
+      "step": 169
+    },
+    {
+      "epoch": 0.1088,
+      "grad_norm": 0.35913320358975465,
+      "learning_rate": 4.9547001741899693e-05,
+      "loss": 0.297,
+      "step": 170
+    },
+    {
+      "epoch": 0.10944,
+      "grad_norm": 0.3331478821334656,
+      "learning_rate": 4.953691055433391e-05,
+      "loss": 0.2901,
+      "step": 171
+    },
+    {
+      "epoch": 0.11008,
+      "grad_norm": 0.39224813006559756,
+      "learning_rate": 4.952670925413182e-05,
+      "loss": 0.3026,
+      "step": 172
+    },
+    {
+      "epoch": 0.11072,
+      "grad_norm": 0.3480414792318722,
+      "learning_rate": 4.951639788707307e-05,
+      "loss": 0.2845,
+      "step": 173
+    },
+    {
+      "epoch": 0.11136,
+      "grad_norm": 0.34970614984057585,
+      "learning_rate": 4.950597649943132e-05,
+      "loss": 0.2775,
+      "step": 174
+    },
+    {
+      "epoch": 0.112,
+      "grad_norm": 0.347207358588287,
+      "learning_rate": 4.9495445137973905e-05,
+      "loss": 0.2923,
+      "step": 175
+    },
+    {
+      "epoch": 0.11264,
+      "grad_norm": 0.34884090954360897,
+      "learning_rate": 4.948480384996172e-05,
+      "loss": 0.2862,
+      "step": 176
+    },
+    {
+      "epoch": 0.11328,
+      "grad_norm": 0.32979237586574495,
+      "learning_rate": 4.947405268314894e-05,
+      "loss": 0.2983,
+      "step": 177
+    },
+    {
+      "epoch": 0.11392,
+      "grad_norm": 0.43991275496703847,
+      "learning_rate": 4.9463191685782865e-05,
+      "loss": 0.3068,
+      "step": 178
+    },
+    {
+      "epoch": 0.11456,
+      "grad_norm": 0.3039302314073942,
+      "learning_rate": 4.945222090660364e-05,
+      "loss": 0.2841,
+      "step": 179
+    },
+    {
+      "epoch": 0.1152,
+      "grad_norm": 0.38589337384034356,
+      "learning_rate": 4.944114039484411e-05,
+      "loss": 0.3091,
+      "step": 180
+    },
+    {
+      "epoch": 0.11584,
+      "grad_norm": 0.34105198100381573,
+      "learning_rate": 4.942995020022954e-05,
+      "loss": 0.3228,
+      "step": 181
+    },
+    {
+      "epoch": 0.11648,
+      "grad_norm": 0.34780611263986955,
+      "learning_rate": 4.9418650372977384e-05,
+      "loss": 0.2874,
+      "step": 182
+    },
+    {
+      "epoch": 0.11712,
+      "grad_norm": 0.32541445621708004,
+      "learning_rate": 4.940724096379714e-05,
+      "loss": 0.282,
+      "step": 183
+    },
+    {
+      "epoch": 0.11776,
+      "grad_norm": 0.3082894269244735,
+      "learning_rate": 4.9395722023890035e-05,
+      "loss": 0.2799,
+      "step": 184
+    },
+    {
+      "epoch": 0.1184,
+      "grad_norm": 0.32812745331187065,
+      "learning_rate": 4.938409360494883e-05,
+      "loss": 0.2827,
+      "step": 185
+    },
+    {
+      "epoch": 0.11904,
+      "grad_norm": 0.34755898651365463,
+      "learning_rate": 4.937235575915761e-05,
+      "loss": 0.2917,
+      "step": 186
+    },
+    {
+      "epoch": 0.11968,
+      "grad_norm": 0.3413504764617461,
+      "learning_rate": 4.93605085391915e-05,
+      "loss": 0.2918,
+      "step": 187
+    },
+    {
+      "epoch": 0.12032,
+      "grad_norm": 0.3705985820501455,
+      "learning_rate": 4.934855199821647e-05,
+      "loss": 0.2826,
+      "step": 188
+    },
+    {
+      "epoch": 0.12096,
+      "grad_norm": 0.3414000814514884,
+      "learning_rate": 4.93364861898891e-05,
+      "loss": 0.276,
+      "step": 189
+    },
+    {
+      "epoch": 0.1216,
+      "grad_norm": 0.3589734824395955,
+      "learning_rate": 4.932431116835628e-05,
+      "loss": 0.2959,
+      "step": 190
+    },
+    {
+      "epoch": 0.12224,
+      "grad_norm": 0.345670488984043,
+      "learning_rate": 4.931202698825503e-05,
+      "loss": 0.281,
+      "step": 191
+    },
+    {
+      "epoch": 0.12288,
+      "grad_norm": 0.3235692312545686,
+      "learning_rate": 4.929963370471226e-05,
+      "loss": 0.2886,
+      "step": 192
+    },
+    {
+      "epoch": 0.12352,
+      "grad_norm": 0.3034427958798936,
+      "learning_rate": 4.928713137334445e-05,
+      "loss": 0.2812,
+      "step": 193
+    },
+    {
+      "epoch": 0.12416,
+      "grad_norm": 0.34712496586477004,
+      "learning_rate": 4.927452005025748e-05,
+      "loss": 0.3018,
+      "step": 194
+    },
+    {
+      "epoch": 0.1248,
+      "grad_norm": 0.3222467927855389,
+      "learning_rate": 4.926179979204633e-05,
+      "loss": 0.3023,
+      "step": 195
+    },
+    {
+      "epoch": 0.12544,
+      "grad_norm": 0.3243348951725544,
+      "learning_rate": 4.9248970655794836e-05,
+      "loss": 0.2917,
+      "step": 196
+    },
+    {
+      "epoch": 0.12608,
+      "grad_norm": 0.3096795363251828,
+      "learning_rate": 4.923603269907545e-05,
+      "loss": 0.2737,
+      "step": 197
+    },
+    {
+      "epoch": 0.12672,
+      "grad_norm": 0.31152060487499617,
+      "learning_rate": 4.922298597994897e-05,
+      "loss": 0.2828,
+      "step": 198
+    },
+    {
+      "epoch": 0.12736,
+      "grad_norm": 0.3405465489598549,
+      "learning_rate": 4.920983055696428e-05,
+      "loss": 0.2927,
+      "step": 199
+    },
+    {
+      "epoch": 0.128,
+      "grad_norm": 0.3423946912166355,
+      "learning_rate": 4.919656648915807e-05,
+      "loss": 0.2835,
+      "step": 200
+    },
+    {
+      "epoch": 0.12864,
+      "grad_norm": 0.3464727040800733,
+      "learning_rate": 4.918319383605462e-05,
+      "loss": 0.2575,
+      "step": 201
+    },
+    {
+      "epoch": 0.12928,
+      "grad_norm": 0.3620779326329196,
+      "learning_rate": 4.916971265766547e-05,
+      "loss": 0.2944,
+      "step": 202
+    },
+    {
+      "epoch": 0.12992,
+      "grad_norm": 0.3699616777622434,
+      "learning_rate": 4.915612301448919e-05,
+      "loss": 0.2942,
+      "step": 203
+    },
+    {
+      "epoch": 0.13056,
+      "grad_norm": 0.38798124559835534,
+      "learning_rate": 4.914242496751111e-05,
+      "loss": 0.2882,
+      "step": 204
+    },
+    {
+      "epoch": 0.1312,
+      "grad_norm": 0.4306348775923099,
+      "learning_rate": 4.912861857820303e-05,
+      "loss": 0.3026,
+      "step": 205
+    },
+    {
+      "epoch": 0.13184,
+      "grad_norm": 0.37275998073469474,
+      "learning_rate": 4.911470390852295e-05,
+      "loss": 0.3008,
+      "step": 206
+    },
+    {
+      "epoch": 0.13248,
+      "grad_norm": 0.33076859723493407,
+      "learning_rate": 4.91006810209148e-05,
+      "loss": 0.2881,
+      "step": 207
+    },
+    {
+      "epoch": 0.13312,
+      "grad_norm": 0.38285708110841127,
+      "learning_rate": 4.9086549978308125e-05,
+      "loss": 0.299,
+      "step": 208
+    },
+    {
+      "epoch": 0.13376,
+      "grad_norm": 0.3130925296009032,
+      "learning_rate": 4.9072310844117875e-05,
+      "loss": 0.2714,
+      "step": 209
+    },
+    {
+      "epoch": 0.1344,
+      "grad_norm": 0.3556404363171034,
+      "learning_rate": 4.9057963682244045e-05,
+      "loss": 0.291,
+      "step": 210
+    },
+    {
+      "epoch": 0.13504,
+      "grad_norm": 0.33759319834147206,
+      "learning_rate": 4.904350855707142e-05,
+      "loss": 0.2868,
+      "step": 211
+    },
+    {
+      "epoch": 0.13568,
+      "grad_norm": 0.3171722280640699,
+      "learning_rate": 4.9028945533469276e-05,
+      "loss": 0.2646,
+      "step": 212
+    },
+    {
+      "epoch": 0.13632,
+      "grad_norm": 0.39240914087101075,
+      "learning_rate": 4.9014274676791134e-05,
+      "loss": 0.302,
+      "step": 213
+    },
+    {
+      "epoch": 0.13696,
+      "grad_norm": 0.34184628116877047,
+      "learning_rate": 4.899949605287439e-05,
+      "loss": 0.2792,
+      "step": 214
+    },
+    {
+      "epoch": 0.1376,
+      "grad_norm": 0.3700664187254164,
+      "learning_rate": 4.898460972804009e-05,
+      "loss": 0.3088,
+      "step": 215
+    },
+    {
+      "epoch": 0.13824,
+      "grad_norm": 0.36421921549881453,
+      "learning_rate": 4.896961576909258e-05,
+      "loss": 0.2808,
+      "step": 216
+    },
+    {
+      "epoch": 0.13888,
+      "grad_norm": 0.3565240796039332,
+      "learning_rate": 4.895451424331924e-05,
+      "loss": 0.2714,
+      "step": 217
+    },
+    {
+      "epoch": 0.13952,
+      "grad_norm": 0.3396297883795516,
+      "learning_rate": 4.893930521849016e-05,
+      "loss": 0.267,
+      "step": 218
+    },
+    {
+      "epoch": 0.14016,
+      "grad_norm": 0.39278607376772506,
+      "learning_rate": 4.892398876285785e-05,
+      "loss": 0.2896,
+      "step": 219
+    },
+    {
+      "epoch": 0.1408,
+      "grad_norm": 0.38437399542076783,
+      "learning_rate": 4.890856494515696e-05,
+      "loss": 0.2875,
+      "step": 220
+    },
+    {
+      "epoch": 0.14144,
+      "grad_norm": 0.37920192384172824,
+      "learning_rate": 4.889303383460389e-05,
+      "loss": 0.2806,
+      "step": 221
+    },
+    {
+      "epoch": 0.14208,
+      "grad_norm": 0.35504283495161315,
+      "learning_rate": 4.887739550089656e-05,
+      "loss": 0.2781,
+      "step": 222
+    },
+    {
+      "epoch": 0.14272,
+      "grad_norm": 0.3866107761723795,
+      "learning_rate": 4.886165001421408e-05,
+      "loss": 0.3091,
+      "step": 223
+    },
+    {
+      "epoch": 0.14336,
+      "grad_norm": 0.30439690273605985,
+      "learning_rate": 4.884579744521638e-05,
+      "loss": 0.2842,
+      "step": 224
+    },
+    {
+      "epoch": 0.144,
+      "grad_norm": 0.3551417352689114,
+      "learning_rate": 4.8829837865043984e-05,
+      "loss": 0.2909,
+      "step": 225
+    },
+    {
+      "epoch": 0.14464,
+      "grad_norm": 0.309111612445454,
+      "learning_rate": 4.881377134531762e-05,
+      "loss": 0.268,
+      "step": 226
+    },
+    {
+      "epoch": 0.14528,
+      "grad_norm": 0.3282993602648903,
+      "learning_rate": 4.8797597958137914e-05,
+      "loss": 0.28,
+      "step": 227
+    },
+    {
+      "epoch": 0.14592,
+      "grad_norm": 0.3581890513495606,
+      "learning_rate": 4.878131777608509e-05,
+      "loss": 0.2842,
+      "step": 228
+    },
+    {
+      "epoch": 0.14656,
+      "grad_norm": 0.32315603560305584,
+      "learning_rate": 4.876493087221862e-05,
+      "loss": 0.2961,
+      "step": 229
+    },
+    {
+      "epoch": 0.1472,
+      "grad_norm": 0.33067996529729515,
+      "learning_rate": 4.87484373200769e-05,
+      "loss": 0.2891,
+      "step": 230
+    },
+    {
+      "epoch": 0.14784,
+      "grad_norm": 0.3178546016599421,
+      "learning_rate": 4.8731837193676944e-05,
+      "loss": 0.2781,
+      "step": 231
+    },
+    {
+      "epoch": 0.14848,
+      "grad_norm": 0.3414289060571217,
+      "learning_rate": 4.871513056751401e-05,
+      "loss": 0.2983,
+      "step": 232
+    },
+    {
+      "epoch": 0.14912,
+      "grad_norm": 0.3420355616767074,
+      "learning_rate": 4.8698317516561295e-05,
+      "loss": 0.2987,
+      "step": 233
+    },
+    {
+      "epoch": 0.14976,
+      "grad_norm": 0.347117902288678,
+      "learning_rate": 4.8681398116269596e-05,
+      "loss": 0.2857,
+      "step": 234
+    },
+    {
+      "epoch": 0.1504,
+      "grad_norm": 0.31648517023684664,
+      "learning_rate": 4.8664372442566954e-05,
+      "loss": 0.2921,
+      "step": 235
+    },
+    {
+      "epoch": 0.15104,
+      "grad_norm": 0.3308519821164363,
+      "learning_rate": 4.864724057185834e-05,
+      "loss": 0.2672,
+      "step": 236
+    },
+    {
+      "epoch": 0.15168,
+      "grad_norm": 0.31208238851833453,
+      "learning_rate": 4.86300025810253e-05,
+      "loss": 0.2798,
+      "step": 237
+    },
+    {
+      "epoch": 0.15232,
+      "grad_norm": 0.2945629591271816,
+      "learning_rate": 4.861265854742559e-05,
+      "loss": 0.2732,
+      "step": 238
+    },
+    {
+      "epoch": 0.15296,
+      "grad_norm": 0.3383764371143183,
+      "learning_rate": 4.859520854889287e-05,
+      "loss": 0.2923,
+      "step": 239
+    },
+    {
+      "epoch": 0.1536,
+      "grad_norm": 0.2935231514068793,
+      "learning_rate": 4.857765266373631e-05,
+      "loss": 0.2945,
+      "step": 240
+    },
+    {
+      "epoch": 0.15424,
+      "grad_norm": 0.3265200869108488,
+      "learning_rate": 4.855999097074027e-05,
+      "loss": 0.2925,
+      "step": 241
+    },
+    {
+      "epoch": 0.15488,
+      "grad_norm": 0.301565323998506,
+      "learning_rate": 4.8542223549163946e-05,
+      "loss": 0.271,
+      "step": 242
+    },
+    {
+      "epoch": 0.15552,
+      "grad_norm": 0.3121729061732836,
+      "learning_rate": 4.8524350478740985e-05,
+      "loss": 0.266,
+      "step": 243
+    },
+    {
+      "epoch": 0.15616,
+      "grad_norm": 0.3612945591439241,
+      "learning_rate": 4.850637183967917e-05,
+      "loss": 0.2825,
+      "step": 244
+    },
+    {
+      "epoch": 0.1568,
+      "grad_norm": 0.32526112140366076,
+      "learning_rate": 4.848828771266002e-05,
+      "loss": 0.2875,
+      "step": 245
+    },
+    {
+      "epoch": 0.15744,
+      "grad_norm": 0.3724034080949958,
+      "learning_rate": 4.847009817883844e-05,
+      "loss": 0.2829,
+      "step": 246
+    },
+    {
+      "epoch": 0.15808,
+      "grad_norm": 0.31233652215083263,
+      "learning_rate": 4.845180331984239e-05,
+      "loss": 0.2713,
+      "step": 247
+    },
+    {
+      "epoch": 0.15872,
+      "grad_norm": 0.3480570732169317,
+      "learning_rate": 4.843340321777246e-05,
+      "loss": 0.2641,
+      "step": 248
+    },
+    {
+      "epoch": 0.15936,
+      "grad_norm": 0.320447396199795,
+      "learning_rate": 4.8414897955201556e-05,
+      "loss": 0.261,
+      "step": 249
+    },
+    {
+      "epoch": 0.16,
+      "grad_norm": 0.30011124568595027,
+      "learning_rate": 4.8396287615174484e-05,
+      "loss": 0.2566,
+      "step": 250
+    },
+    {
+      "epoch": 0.16064,
+      "grad_norm": 0.3416591207698499,
+      "learning_rate": 4.837757228120762e-05,
+      "loss": 0.2886,
+      "step": 251
+    },
+    {
+      "epoch": 0.16128,
+      "grad_norm": 0.3347844484474534,
+      "learning_rate": 4.835875203728849e-05,
+      "loss": 0.2546,
+      "step": 252
+    },
+    {
+      "epoch": 0.16192,
+      "grad_norm": 0.3805038941147735,
+      "learning_rate": 4.8339826967875443e-05,
+      "loss": 0.2867,
+      "step": 253
+    },
+    {
+      "epoch": 0.16256,
+      "grad_norm": 0.36899369410322225,
+      "learning_rate": 4.832079715789724e-05,
+      "loss": 0.2969,
+      "step": 254
+    },
+    {
+      "epoch": 0.1632,
+      "grad_norm": 0.3671046083988311,
+      "learning_rate": 4.830166269275267e-05,
+      "loss": 0.2903,
+      "step": 255
+    },
+    {
+      "epoch": 0.16384,
+      "grad_norm": 0.3268166037896424,
+      "learning_rate": 4.828242365831016e-05,
+      "loss": 0.2744,
+      "step": 256
+    },
+    {
+      "epoch": 0.16448,
+      "grad_norm": 0.38930768380996184,
+      "learning_rate": 4.8263080140907455e-05,
+      "loss": 0.2898,
+      "step": 257
+    },
+    {
+      "epoch": 0.16512,
+      "grad_norm": 0.3527164726419632,
+      "learning_rate": 4.824363222735114e-05,
+      "loss": 0.2751,
+      "step": 258
+    },
+    {
+      "epoch": 0.16576,
+      "grad_norm": 0.3831694490194351,
+      "learning_rate": 4.82240800049163e-05,
+      "loss": 0.2797,
+      "step": 259
+    },
+    {
+      "epoch": 0.1664,
+      "grad_norm": 0.39066664017067054,
+      "learning_rate": 4.820442356134612e-05,
+      "loss": 0.2716,
+      "step": 260
+    },
+    {
+      "epoch": 0.16704,
+      "grad_norm": 0.36528936400377576,
+      "learning_rate": 4.8184662984851505e-05,
+      "loss": 0.3001,
+      "step": 261
+    }
+  ],
+  "logging_steps": 1,
+  "max_steps": 1562,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 1,
+  "save_steps": 261,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 119007652085760.0,
+  "train_batch_size": 2,
+  "trial_name": null,
+  "trial_params": null
+}

training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5946d40dd8d3826d5c86fe25022910cc155f3c909b1feb80eb2aa211e8f5d20b
+size 8017

zero_to_fp32.py

@@ -0,0 +1,760 @@
+#!/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 . output_dir/
+#   or
+#   python zero_to_fp32.py . output_dir/ --safe_serialization
+
+import argparse
+import torch
+import glob
+import math
+import os
+import re
+import gc
+import json
+import numpy as np
+from tqdm import tqdm
+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, weights_only=False)
+
+        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 tqdm(files, desc='Loading checkpoint shards'):
+        state_dict = torch.load(f, map_location=device, mmap=True, weights_only=False)
+        # 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}")
+
+    fp32_flat_groups = [state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key] 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")
+
+
+class GatheredTensor:
+    """
+    A pseudo tensor that collects partitioned weights.
+    It is more memory efficient when there are multiple groups.
+    """
+
+    def __init__(self, flat_groups, flat_groups_offset, offset, partitioned_numel, shape):
+        self.flat_groups = flat_groups
+        self.flat_groups_offset = flat_groups_offset
+        self.offset = offset
+        self.partitioned_numel = partitioned_numel
+        self.shape = shape
+        self.dtype = self.flat_groups[0][0].dtype
+
+    def contiguous(self):
+        """
+        Merge partitioned weights from flat_groups into a single tensor.
+        """
+        end_idx = self.offset + self.partitioned_numel
+        world_size = len(self.flat_groups)
+        pad_flat_param_chunks = []
+
+        for rank_i in range(world_size):
+            # for each rank, we need to collect weights from related group/groups
+            flat_groups_at_rank_i = self.flat_groups[rank_i]
+            start_group_id = None
+            end_group_id = None
+            for group_id in range(len(self.flat_groups_offset)):
+                if self.flat_groups_offset[group_id] <= self.offset < self.flat_groups_offset[group_id + 1]:
+                    start_group_id = group_id
+                if self.flat_groups_offset[group_id] < end_idx <= self.flat_groups_offset[group_id + 1]:
+                    end_group_id = group_id
+                    break
+            # collect weights from related group/groups
+            for group_id in range(start_group_id, end_group_id + 1):
+                flat_tensor = flat_groups_at_rank_i[group_id]
+                start_offset = self.offset - self.flat_groups_offset[group_id]
+                end_offset = min(end_idx, self.flat_groups_offset[group_id + 1]) - self.flat_groups_offset[group_id]
+                pad_flat_param_chunks.append(flat_tensor[start_offset:end_offset])
+
+        # collect weights from all ranks
+        pad_flat_param = torch.cat(pad_flat_param_chunks, dim=0)
+        param = pad_flat_param[:self.shape.numel()].view(self.shape).contiguous()
+        return param
+
+
+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 = sum([flat_group.numel() for flat_group in fp32_flat_groups[0]]) * 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
+    flat_groups_offset = [0] + list(np.cumsum([flat_tensor.numel() for flat_tensor in fp32_flat_groups[0]]))
+    for name, shape in tqdm(param_shapes.items(), desc='Gathering sharded weights'):
+        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}"
+            )
+
+        # memory efficient tensor
+        tensor = GatheredTensor(fp32_flat_groups, flat_groups_offset, offset, partitioned_numel, shape)
+        state_dict[name] = tensor
+        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 to_torch_tensor(state_dict, return_empty_tensor=False):
+    """
+    Convert state_dict of GatheredTensor to torch tensor
+    """
+    torch_state_dict = {}
+    converted_tensors = {}
+    for name, tensor in state_dict.items():
+        tensor_id = id(tensor)
+        if tensor_id in converted_tensors:  # shared tensors
+            shared_tensor = torch_state_dict[converted_tensors[tensor_id]]
+            torch_state_dict[name] = shared_tensor
+        else:
+            converted_tensors[tensor_id] = name
+            if return_empty_tensor:
+                torch_state_dict[name] = torch.empty(tensor.shape, dtype=tensor.dtype)
+            else:
+                torch_state_dict[name] = tensor.contiguous()
+    return torch_state_dict
+
+
+def get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir,
+                                             tag=None,
+                                             exclude_frozen_parameters=False,
+                                             lazy_mode=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
+        - ``lazy_mode``: get state_dict in lazy mode. It returns a dict of pesduo tensor instead of torch tensor, which is more memory efficient.
+          Convert the pesduo tensor to torch tensor by ``.contiguous()``
+
+    Returns:
+        - pytorch ``state_dict``
+
+    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.
+
+    Note: the above usage may not work if your application doesn't have sufficient free CPU memory.
+    You may need to use the offline approach using the ``zero_to_fp32.py`` script that is saved with
+    the checkpoint. Or you can load state_dict in lazy mode ::
+
+        from deepspeed.utils.zero_to_fp32 import get_fp32_state_dict_from_zero_checkpoint
+        state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, lazy_mode=True) # not on cpu
+        for name, lazy_tensor in state_dict.item():
+            tensor = lazy_tensor.contiguous()  # to cpu
+            print(name, tensor)
+            # del tensor to release memory if it no longer in use
+    """
+    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")
+
+    state_dict = _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters)
+    if lazy_mode:
+        return state_dict
+    else:
+        return to_torch_tensor(state_dict)
+
+
+def convert_zero_checkpoint_to_fp32_state_dict(checkpoint_dir,
+                                               output_dir,
+                                               max_shard_size="5GB",
+                                               safe_serialization=False,
+                                               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_dir``: directory to the pytorch fp32 state_dict output files
+        - ``max_shard_size``: the maximum size for a checkpoint before being sharded, default value is 5GB
+        - ``safe_serialization``:  whether to save the model using `safetensors` or the traditional PyTorch way (that uses `pickle`).
+        - ``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
+    """
+
+    # Dependency pre-check
+    if safe_serialization:
+        try:
+            from safetensors.torch import save_file
+        except ImportError:
+            print('If you want to use `safe_serialization`, please `pip install safetensors`')
+            raise
+    if max_shard_size is not None:
+        try:
+            from huggingface_hub import split_torch_state_dict_into_shards
+        except ImportError:
+            print('If you want to use `max_shard_size`, please `pip install huggingface_hub`')
+            raise
+
+    # Convert zero checkpoint to state_dict
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir,
+                                                          tag,
+                                                          exclude_frozen_parameters,
+                                                          lazy_mode=True)
+
+    # Shard the model if it is too big.
+    weights_name = "model.safetensors" if safe_serialization else "pytorch_model.bin"
+    if max_shard_size is not None:
+        filename_pattern = weights_name.replace(".bin", "{suffix}.bin").replace(".safetensors", "{suffix}.safetensors")
+        # an memory-efficient approach for sharding
+        empty_state_dict = to_torch_tensor(state_dict, return_empty_tensor=True)
+        state_dict_split = split_torch_state_dict_into_shards(empty_state_dict,
+                                                              filename_pattern=filename_pattern,
+                                                              max_shard_size=max_shard_size)
+    else:
+        from collections import namedtuple
+        StateDictSplit = namedtuple("StateDictSplit", ["is_sharded", "filename_to_tensors"])
+        state_dict_split = StateDictSplit(is_sharded=False,
+                                          filename_to_tensors={weights_name: list(state_dict.keys())})
+
+    # Save the model by shard
+    os.makedirs(output_dir, exist_ok=True)
+    filename_to_tensors = state_dict_split.filename_to_tensors.items()
+    for shard_file, tensors in tqdm(filename_to_tensors, desc="Saving checkpoint shards"):
+        shard_state_dict = {tensor_name: state_dict[tensor_name] for tensor_name in tensors}
+        shard_state_dict = to_torch_tensor(shard_state_dict)
+        output_path = os.path.join(output_dir, shard_file)
+        if safe_serialization:
+            save_file(shard_state_dict, output_path, metadata={"format": "pt"})
+        else:
+            torch.save(shard_state_dict, output_path)
+        # release the memory of current shard
+        for tensor_name in list(shard_state_dict.keys()):
+            del state_dict[tensor_name]
+            del shard_state_dict[tensor_name]
+        del shard_state_dict
+        gc.collect()
+
+    # Save index if sharded
+    if state_dict_split.is_sharded:
+        index = {
+            "metadata": state_dict_split.metadata,
+            "weight_map": state_dict_split.tensor_to_filename,
+        }
+        save_index_file = "model.safetensors.index.json" if safe_serialization else "pytorch_model.bin.index.json"
+        save_index_file = os.path.join(output_dir, save_index_file)
+        with open(save_index_file, "w", encoding="utf-8") as f:
+            content = json.dumps(index, indent=2, sort_keys=True) + "\n"
+            f.write(content)
+
+
+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_dir",
+                        type=str,
+                        help="directory to the pytorch fp32 state_dict output files"
+                        "(e.g. path/checkpoint-12-output/)")
+    parser.add_argument(
+        "--max_shard_size",
+        type=str,
+        default="5GB",
+        help="The maximum size for a checkpoint before being sharded. Checkpoints shard will then be each of size"
+        "lower than this size. If expressed as a string, needs to be digits followed by a unit (like `5MB`"
+        "We default it to 5GB in order for models to be able to run easily on free-tier google colab instances"
+        "without CPU OOM issues.")
+    parser.add_argument(
+        "--safe_serialization",
+        default=False,
+        action='store_true',
+        help="Whether to save the model using `safetensors` or the traditional PyTorch way (that uses `pickle`).")
+    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_dir,
+                                               max_shard_size=args.max_shard_size,
+                                               safe_serialization=args.safe_serialization,
+                                               tag=args.tag,
+                                               exclude_frozen_parameters=args.exclude_frozen_parameters)