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

config.json

@@ -0,0 +1,28 @@
+{
+  "_name_or_path": "outputs_qalign/out_stage_1",
+  "architectures": [
+    "Qwen2ForCausalLM"
+  ],
+  "attention_dropout": 0.0,
+  "eos_token_id": 151645,
+  "hidden_act": "silu",
+  "hidden_size": 1536,
+  "initializer_range": 0.02,
+  "intermediate_size": 8960,
+  "max_position_embeddings": 32768,
+  "max_window_layers": 21,
+  "model_type": "qwen2",
+  "num_attention_heads": 12,
+  "num_hidden_layers": 28,
+  "num_key_value_heads": 2,
+  "rms_norm_eps": 1e-06,
+  "rope_scaling": null,
+  "rope_theta": 1000000.0,
+  "sliding_window": 32768,
+  "tie_word_embeddings": true,
+  "torch_dtype": "bfloat16",
+  "transformers_version": "4.49.0",
+  "use_cache": false,
+  "use_sliding_window": false,
+  "vocab_size": 151936
+}

generation_config.json

@@ -0,0 +1,14 @@
+{
+  "bos_token_id": 151643,
+  "do_sample": true,
+  "eos_token_id": [
+    151645,
+    151643
+  ],
+  "pad_token_id": 151643,
+  "repetition_penalty": 1.1,
+  "temperature": 0.7,
+  "top_k": 20,
+  "top_p": 0.8,
+  "transformers_version": "4.49.0"
+}

latest

@@ -0,0 +1 @@
+global_step194

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1a01411c2fd299535b0f63d54a8bcf5de65cb46ed5faa14a201d4b974139808e
+size 3554214752

rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6cb795a5cea0baa625c50007a6c9da09c6bbb5c16b560424070384a479e7d8a6
+size 14512

rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5f19604377bd828eb366c68946ad997a4ff4d69beaeea93ee58915135768ec63
+size 14512

scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1bf927aeba7884e78dc27a27c5832aad0eefccdeed21230a97d2305c9ee44eff
+size 1064

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 sha256:9c5ae00e602b8860cbd784ba82a8aa14e8feecec692e7076590d014d7b7fdafa
+size 11421896

tokenizer_config.json

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

trainer_state.json

@@ -0,0 +1,1399 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.9980707395498393,
+  "eval_steps": 500,
+  "global_step": 194,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.005144694533762058,
+      "grad_norm": 3.161365032196045,
+      "learning_rate": 1.111111111111111e-06,
+      "loss": 0.7146,
+      "step": 1
+    },
+    {
+      "epoch": 0.010289389067524116,
+      "grad_norm": 3.145214080810547,
+      "learning_rate": 2.222222222222222e-06,
+      "loss": 0.6479,
+      "step": 2
+    },
+    {
+      "epoch": 0.015434083601286173,
+      "grad_norm": 3.2331063747406006,
+      "learning_rate": 3.3333333333333333e-06,
+      "loss": 0.5785,
+      "step": 3
+    },
+    {
+      "epoch": 0.020578778135048232,
+      "grad_norm": 3.2699339389801025,
+      "learning_rate": 4.444444444444444e-06,
+      "loss": 0.6538,
+      "step": 4
+    },
+    {
+      "epoch": 0.02572347266881029,
+      "grad_norm": 3.1358041763305664,
+      "learning_rate": 5.555555555555557e-06,
+      "loss": 0.651,
+      "step": 5
+    },
+    {
+      "epoch": 0.030868167202572346,
+      "grad_norm": 2.0558807849884033,
+      "learning_rate": 6.666666666666667e-06,
+      "loss": 0.5806,
+      "step": 6
+    },
+    {
+      "epoch": 0.0360128617363344,
+      "grad_norm": 1.1790688037872314,
+      "learning_rate": 7.77777777777778e-06,
+      "loss": 0.5116,
+      "step": 7
+    },
+    {
+      "epoch": 0.041157556270096464,
+      "grad_norm": 0.9816439747810364,
+      "learning_rate": 8.888888888888888e-06,
+      "loss": 0.5529,
+      "step": 8
+    },
+    {
+      "epoch": 0.04630225080385852,
+      "grad_norm": 0.8186548352241516,
+      "learning_rate": 1e-05,
+      "loss": 0.5629,
+      "step": 9
+    },
+    {
+      "epoch": 0.05144694533762058,
+      "grad_norm": 0.9718908667564392,
+      "learning_rate": 9.999279082014233e-06,
+      "loss": 0.5076,
+      "step": 10
+    },
+    {
+      "epoch": 0.05659163987138264,
+      "grad_norm": 0.8055087327957153,
+      "learning_rate": 9.997116535946028e-06,
+      "loss": 0.5523,
+      "step": 11
+    },
+    {
+      "epoch": 0.06173633440514469,
+      "grad_norm": 0.7194182872772217,
+      "learning_rate": 9.993512985402724e-06,
+      "loss": 0.5105,
+      "step": 12
+    },
+    {
+      "epoch": 0.06688102893890675,
+      "grad_norm": 0.9174315929412842,
+      "learning_rate": 9.988469469530086e-06,
+      "loss": 0.5066,
+      "step": 13
+    },
+    {
+      "epoch": 0.0720257234726688,
+      "grad_norm": 1.0070743560791016,
+      "learning_rate": 9.981987442712634e-06,
+      "loss": 0.5679,
+      "step": 14
+    },
+    {
+      "epoch": 0.07717041800643087,
+      "grad_norm": 0.7564347982406616,
+      "learning_rate": 9.974068774154252e-06,
+      "loss": 0.4988,
+      "step": 15
+    },
+    {
+      "epoch": 0.08231511254019293,
+      "grad_norm": 0.7040192484855652,
+      "learning_rate": 9.964715747339178e-06,
+      "loss": 0.5261,
+      "step": 16
+    },
+    {
+      "epoch": 0.08745980707395498,
+      "grad_norm": 0.6901726722717285,
+      "learning_rate": 9.95393105937351e-06,
+      "loss": 0.5519,
+      "step": 17
+    },
+    {
+      "epoch": 0.09260450160771704,
+      "grad_norm": 0.5697789788246155,
+      "learning_rate": 9.941717820207461e-06,
+      "loss": 0.4733,
+      "step": 18
+    },
+    {
+      "epoch": 0.0977491961414791,
+      "grad_norm": 0.5673728585243225,
+      "learning_rate": 9.928079551738542e-06,
+      "loss": 0.5225,
+      "step": 19
+    },
+    {
+      "epoch": 0.10289389067524116,
+      "grad_norm": 0.5412207841873169,
+      "learning_rate": 9.913020186795967e-06,
+      "loss": 0.5622,
+      "step": 20
+    },
+    {
+      "epoch": 0.10803858520900321,
+      "grad_norm": 0.5939576029777527,
+      "learning_rate": 9.89654406800655e-06,
+      "loss": 0.5056,
+      "step": 21
+    },
+    {
+      "epoch": 0.11318327974276528,
+      "grad_norm": 0.5736626386642456,
+      "learning_rate": 9.878655946542443e-06,
+      "loss": 0.5236,
+      "step": 22
+    },
+    {
+      "epoch": 0.11832797427652733,
+      "grad_norm": 0.5959703326225281,
+      "learning_rate": 9.85936098075104e-06,
+      "loss": 0.5045,
+      "step": 23
+    },
+    {
+      "epoch": 0.12347266881028938,
+      "grad_norm": 0.5663568377494812,
+      "learning_rate": 9.838664734667496e-06,
+      "loss": 0.5815,
+      "step": 24
+    },
+    {
+      "epoch": 0.12861736334405144,
+      "grad_norm": 0.5114029049873352,
+      "learning_rate": 9.81657317641022e-06,
+      "loss": 0.496,
+      "step": 25
+    },
+    {
+      "epoch": 0.1337620578778135,
+      "grad_norm": 0.5281174182891846,
+      "learning_rate": 9.79309267645989e-06,
+      "loss": 0.5243,
+      "step": 26
+    },
+    {
+      "epoch": 0.13890675241157557,
+      "grad_norm": 0.5573230981826782,
+      "learning_rate": 9.768230005822394e-06,
+      "loss": 0.5015,
+      "step": 27
+    },
+    {
+      "epoch": 0.1440514469453376,
+      "grad_norm": 0.5166211128234863,
+      "learning_rate": 9.741992334076309e-06,
+      "loss": 0.5105,
+      "step": 28
+    },
+    {
+      "epoch": 0.14919614147909968,
+      "grad_norm": 0.4607759118080139,
+      "learning_rate": 9.714387227305422e-06,
+      "loss": 0.4665,
+      "step": 29
+    },
+    {
+      "epoch": 0.15434083601286175,
+      "grad_norm": 0.504697322845459,
+      "learning_rate": 9.68542264591692e-06,
+      "loss": 0.5532,
+      "step": 30
+    },
+    {
+      "epoch": 0.1594855305466238,
+      "grad_norm": 0.4975253641605377,
+      "learning_rate": 9.65510694234587e-06,
+      "loss": 0.477,
+      "step": 31
+    },
+    {
+      "epoch": 0.16463022508038586,
+      "grad_norm": 0.5030412673950195,
+      "learning_rate": 9.623448858646658e-06,
+      "loss": 0.493,
+      "step": 32
+    },
+    {
+      "epoch": 0.16977491961414792,
+      "grad_norm": 0.489597350358963,
+      "learning_rate": 9.590457523972055e-06,
+      "loss": 0.4568,
+      "step": 33
+    },
+    {
+      "epoch": 0.17491961414790996,
+      "grad_norm": 0.5580692887306213,
+      "learning_rate": 9.55614245194068e-06,
+      "loss": 0.5396,
+      "step": 34
+    },
+    {
+      "epoch": 0.18006430868167203,
+      "grad_norm": 0.5050625801086426,
+      "learning_rate": 9.520513537893574e-06,
+      "loss": 0.5133,
+      "step": 35
+    },
+    {
+      "epoch": 0.18520900321543407,
+      "grad_norm": 0.5139254927635193,
+      "learning_rate": 9.48358105604072e-06,
+      "loss": 0.5307,
+      "step": 36
+    },
+    {
+      "epoch": 0.19035369774919614,
+      "grad_norm": 0.5394223928451538,
+      "learning_rate": 9.445355656498284e-06,
+      "loss": 0.5027,
+      "step": 37
+    },
+    {
+      "epoch": 0.1954983922829582,
+      "grad_norm": 0.4914088845252991,
+      "learning_rate": 9.40584836221749e-06,
+      "loss": 0.5215,
+      "step": 38
+    },
+    {
+      "epoch": 0.20064308681672025,
+      "grad_norm": 0.4907187521457672,
+      "learning_rate": 9.365070565805941e-06,
+      "loss": 0.5195,
+      "step": 39
+    },
+    {
+      "epoch": 0.2057877813504823,
+      "grad_norm": 0.49707943201065063,
+      "learning_rate": 9.323034026242378e-06,
+      "loss": 0.5493,
+      "step": 40
+    },
+    {
+      "epoch": 0.21093247588424438,
+      "grad_norm": 0.49830377101898193,
+      "learning_rate": 9.279750865485772e-06,
+      "loss": 0.4917,
+      "step": 41
+    },
+    {
+      "epoch": 0.21607717041800642,
+      "grad_norm": 0.5180136561393738,
+      "learning_rate": 9.235233564979756e-06,
+      "loss": 0.5193,
+      "step": 42
+    },
+    {
+      "epoch": 0.2212218649517685,
+      "grad_norm": 0.49484989047050476,
+      "learning_rate": 9.18949496205337e-06,
+      "loss": 0.5472,
+      "step": 43
+    },
+    {
+      "epoch": 0.22636655948553056,
+      "grad_norm": 0.4955751895904541,
+      "learning_rate": 9.142548246219212e-06,
+      "loss": 0.4818,
+      "step": 44
+    },
+    {
+      "epoch": 0.2315112540192926,
+      "grad_norm": 0.4689135253429413,
+      "learning_rate": 9.09440695537001e-06,
+      "loss": 0.4677,
+      "step": 45
+    },
+    {
+      "epoch": 0.23665594855305466,
+      "grad_norm": 0.5169051289558411,
+      "learning_rate": 9.045084971874738e-06,
+      "loss": 0.5033,
+      "step": 46
+    },
+    {
+      "epoch": 0.24180064308681673,
+      "grad_norm": 0.46415960788726807,
+      "learning_rate": 8.994596518575393e-06,
+      "loss": 0.4803,
+      "step": 47
+    },
+    {
+      "epoch": 0.24694533762057877,
+      "grad_norm": 0.4593764543533325,
+      "learning_rate": 8.942956154685596e-06,
+      "loss": 0.4792,
+      "step": 48
+    },
+    {
+      "epoch": 0.25209003215434084,
+      "grad_norm": 0.49039268493652344,
+      "learning_rate": 8.890178771592198e-06,
+      "loss": 0.5189,
+      "step": 49
+    },
+    {
+      "epoch": 0.2572347266881029,
+      "grad_norm": 0.48655083775520325,
+      "learning_rate": 8.836279588561084e-06,
+      "loss": 0.4865,
+      "step": 50
+    },
+    {
+      "epoch": 0.262379421221865,
+      "grad_norm": 0.4769374132156372,
+      "learning_rate": 8.781274148348438e-06,
+      "loss": 0.4924,
+      "step": 51
+    },
+    {
+      "epoch": 0.267524115755627,
+      "grad_norm": 0.47683149576187134,
+      "learning_rate": 8.725178312718727e-06,
+      "loss": 0.4901,
+      "step": 52
+    },
+    {
+      "epoch": 0.27266881028938905,
+      "grad_norm": 0.47773587703704834,
+      "learning_rate": 8.668008257870684e-06,
+      "loss": 0.4531,
+      "step": 53
+    },
+    {
+      "epoch": 0.27781350482315115,
+      "grad_norm": 0.47025877237319946,
+      "learning_rate": 8.609780469772623e-06,
+      "loss": 0.5129,
+      "step": 54
+    },
+    {
+      "epoch": 0.2829581993569132,
+      "grad_norm": 0.486856073141098,
+      "learning_rate": 8.550511739408428e-06,
+      "loss": 0.507,
+      "step": 55
+    },
+    {
+      "epoch": 0.2881028938906752,
+      "grad_norm": 0.4797123968601227,
+      "learning_rate": 8.490219157935589e-06,
+      "loss": 0.4873,
+      "step": 56
+    },
+    {
+      "epoch": 0.2932475884244373,
+      "grad_norm": 0.4642903208732605,
+      "learning_rate": 8.428920111756658e-06,
+      "loss": 0.4609,
+      "step": 57
+    },
+    {
+      "epoch": 0.29839228295819936,
+      "grad_norm": 0.48866933584213257,
+      "learning_rate": 8.366632277505598e-06,
+      "loss": 0.5264,
+      "step": 58
+    },
+    {
+      "epoch": 0.3035369774919614,
+      "grad_norm": 0.4838912785053253,
+      "learning_rate": 8.303373616950408e-06,
+      "loss": 0.4981,
+      "step": 59
+    },
+    {
+      "epoch": 0.3086816720257235,
+      "grad_norm": 0.47218450903892517,
+      "learning_rate": 8.239162371813552e-06,
+      "loss": 0.501,
+      "step": 60
+    },
+    {
+      "epoch": 0.31382636655948554,
+      "grad_norm": 0.48886698484420776,
+      "learning_rate": 8.17401705851163e-06,
+      "loss": 0.4924,
+      "step": 61
+    },
+    {
+      "epoch": 0.3189710610932476,
+      "grad_norm": 0.4737836718559265,
+      "learning_rate": 8.107956462815862e-06,
+      "loss": 0.5106,
+      "step": 62
+    },
+    {
+      "epoch": 0.32411575562700967,
+      "grad_norm": 0.44688716530799866,
+      "learning_rate": 8.040999634434883e-06,
+      "loss": 0.4655,
+      "step": 63
+    },
+    {
+      "epoch": 0.3292604501607717,
+      "grad_norm": 0.49828943610191345,
+      "learning_rate": 7.973165881521435e-06,
+      "loss": 0.5314,
+      "step": 64
+    },
+    {
+      "epoch": 0.33440514469453375,
+      "grad_norm": 0.4544566571712494,
+      "learning_rate": 7.90447476510452e-06,
+      "loss": 0.4624,
+      "step": 65
+    },
+    {
+      "epoch": 0.33954983922829585,
+      "grad_norm": 0.45815807580947876,
+      "learning_rate": 7.834946093448658e-06,
+      "loss": 0.4801,
+      "step": 66
+    },
+    {
+      "epoch": 0.3446945337620579,
+      "grad_norm": 0.4789896607398987,
+      "learning_rate": 7.764599916341817e-06,
+      "loss": 0.547,
+      "step": 67
+    },
+    {
+      "epoch": 0.3498392282958199,
+      "grad_norm": 0.47594505548477173,
+      "learning_rate": 7.69345651931372e-06,
+      "loss": 0.5217,
+      "step": 68
+    },
+    {
+      "epoch": 0.35498392282958197,
+      "grad_norm": 0.46930596232414246,
+      "learning_rate": 7.621536417786159e-06,
+      "loss": 0.4578,
+      "step": 69
+    },
+    {
+      "epoch": 0.36012861736334406,
+      "grad_norm": 0.45879995822906494,
+      "learning_rate": 7.548860351157028e-06,
+      "loss": 0.454,
+      "step": 70
+    },
+    {
+      "epoch": 0.3652733118971061,
+      "grad_norm": 0.48580196499824524,
+      "learning_rate": 7.475449276819753e-06,
+      "loss": 0.4777,
+      "step": 71
+    },
+    {
+      "epoch": 0.37041800643086814,
+      "grad_norm": 0.5085631012916565,
+      "learning_rate": 7.401324364119872e-06,
+      "loss": 0.5395,
+      "step": 72
+    },
+    {
+      "epoch": 0.37556270096463024,
+      "grad_norm": 0.45636940002441406,
+      "learning_rate": 7.326506988250488e-06,
+      "loss": 0.4908,
+      "step": 73
+    },
+    {
+      "epoch": 0.3807073954983923,
+      "grad_norm": 0.4770854711532593,
+      "learning_rate": 7.251018724088367e-06,
+      "loss": 0.5052,
+      "step": 74
+    },
+    {
+      "epoch": 0.3858520900321543,
+      "grad_norm": 0.49636557698249817,
+      "learning_rate": 7.174881339972448e-06,
+      "loss": 0.54,
+      "step": 75
+    },
+    {
+      "epoch": 0.3909967845659164,
+      "grad_norm": 0.4442436099052429,
+      "learning_rate": 7.09811679142657e-06,
+      "loss": 0.476,
+      "step": 76
+    },
+    {
+      "epoch": 0.39614147909967845,
+      "grad_norm": 0.46588826179504395,
+      "learning_rate": 7.020747214828221e-06,
+      "loss": 0.5081,
+      "step": 77
+    },
+    {
+      "epoch": 0.4012861736334405,
+      "grad_norm": 0.48112452030181885,
+      "learning_rate": 6.942794921025127e-06,
+      "loss": 0.48,
+      "step": 78
+    },
+    {
+      "epoch": 0.4064308681672026,
+      "grad_norm": 0.47660669684410095,
+      "learning_rate": 6.864282388901544e-06,
+      "loss": 0.4935,
+      "step": 79
+    },
+    {
+      "epoch": 0.4115755627009646,
+      "grad_norm": 0.4866008162498474,
+      "learning_rate": 6.785232258896078e-06,
+      "loss": 0.5333,
+      "step": 80
+    },
+    {
+      "epoch": 0.41672025723472667,
+      "grad_norm": 0.455619215965271,
+      "learning_rate": 6.705667326472926e-06,
+      "loss": 0.499,
+      "step": 81
+    },
+    {
+      "epoch": 0.42186495176848876,
+      "grad_norm": 0.46027281880378723,
+      "learning_rate": 6.625610535548418e-06,
+      "loss": 0.4728,
+      "step": 82
+    },
+    {
+      "epoch": 0.4270096463022508,
+      "grad_norm": 0.44711652398109436,
+      "learning_rate": 6.545084971874738e-06,
+      "loss": 0.5353,
+      "step": 83
+    },
+    {
+      "epoch": 0.43215434083601284,
+      "grad_norm": 0.48993319272994995,
+      "learning_rate": 6.464113856382752e-06,
+      "loss": 0.4929,
+      "step": 84
+    },
+    {
+      "epoch": 0.43729903536977494,
+      "grad_norm": 0.42511484026908875,
+      "learning_rate": 6.382720538485856e-06,
+      "loss": 0.4592,
+      "step": 85
+    },
+    {
+      "epoch": 0.442443729903537,
+      "grad_norm": 0.5443476438522339,
+      "learning_rate": 6.3009284893467655e-06,
+      "loss": 0.5057,
+      "step": 86
+    },
+    {
+      "epoch": 0.447588424437299,
+      "grad_norm": 0.49420294165611267,
+      "learning_rate": 6.218761295109209e-06,
+      "loss": 0.5119,
+      "step": 87
+    },
+    {
+      "epoch": 0.4527331189710611,
+      "grad_norm": 0.5062916278839111,
+      "learning_rate": 6.136242650096451e-06,
+      "loss": 0.5031,
+      "step": 88
+    },
+    {
+      "epoch": 0.45787781350482315,
+      "grad_norm": 0.49038904905319214,
+      "learning_rate": 6.053396349978632e-06,
+      "loss": 0.508,
+      "step": 89
+    },
+    {
+      "epoch": 0.4630225080385852,
+      "grad_norm": 0.4432820677757263,
+      "learning_rate": 5.970246284910877e-06,
+      "loss": 0.4359,
+      "step": 90
+    },
+    {
+      "epoch": 0.4681672025723473,
+      "grad_norm": 0.48448362946510315,
+      "learning_rate": 5.886816432644155e-06,
+      "loss": 0.4852,
+      "step": 91
+    },
+    {
+      "epoch": 0.4733118971061093,
+      "grad_norm": 0.4854327142238617,
+      "learning_rate": 5.803130851610887e-06,
+      "loss": 0.4811,
+      "step": 92
+    },
+    {
+      "epoch": 0.47845659163987136,
+      "grad_norm": 0.4899016320705414,
+      "learning_rate": 5.719213673987277e-06,
+      "loss": 0.5486,
+      "step": 93
+    },
+    {
+      "epoch": 0.48360128617363346,
+      "grad_norm": 0.49537014961242676,
+      "learning_rate": 5.635089098734394e-06,
+      "loss": 0.5107,
+      "step": 94
+    },
+    {
+      "epoch": 0.4887459807073955,
+      "grad_norm": 0.4889853298664093,
+      "learning_rate": 5.550781384619974e-06,
+      "loss": 0.4948,
+      "step": 95
+    },
+    {
+      "epoch": 0.49389067524115754,
+      "grad_norm": 0.46976524591445923,
+      "learning_rate": 5.466314843222993e-06,
+      "loss": 0.4913,
+      "step": 96
+    },
+    {
+      "epoch": 0.49903536977491963,
+      "grad_norm": 0.4508555233478546,
+      "learning_rate": 5.381713831923008e-06,
+      "loss": 0.4752,
+      "step": 97
+    },
+    {
+      "epoch": 0.5041800643086817,
+      "grad_norm": 0.48454639315605164,
+      "learning_rate": 5.297002746876284e-06,
+      "loss": 0.5265,
+      "step": 98
+    },
+    {
+      "epoch": 0.5093247588424438,
+      "grad_norm": 0.48890259861946106,
+      "learning_rate": 5.212206015980742e-06,
+      "loss": 0.4924,
+      "step": 99
+    },
+    {
+      "epoch": 0.5144694533762058,
+      "grad_norm": 0.47355538606643677,
+      "learning_rate": 5.127348091831755e-06,
+      "loss": 0.477,
+      "step": 100
+    },
+    {
+      "epoch": 0.5196141479099678,
+      "grad_norm": 0.48057636618614197,
+      "learning_rate": 5.042453444670829e-06,
+      "loss": 0.4687,
+      "step": 101
+    },
+    {
+      "epoch": 0.52475884244373,
+      "grad_norm": 0.4773067235946655,
+      "learning_rate": 4.957546555329174e-06,
+      "loss": 0.4999,
+      "step": 102
+    },
+    {
+      "epoch": 0.5299035369774919,
+      "grad_norm": 0.474159300327301,
+      "learning_rate": 4.872651908168244e-06,
+      "loss": 0.5025,
+      "step": 103
+    },
+    {
+      "epoch": 0.535048231511254,
+      "grad_norm": 0.5089556574821472,
+      "learning_rate": 4.78779398401926e-06,
+      "loss": 0.509,
+      "step": 104
+    },
+    {
+      "epoch": 0.5401929260450161,
+      "grad_norm": 0.4812917411327362,
+      "learning_rate": 4.702997253123716e-06,
+      "loss": 0.5066,
+      "step": 105
+    },
+    {
+      "epoch": 0.5453376205787781,
+      "grad_norm": 0.4695838689804077,
+      "learning_rate": 4.618286168076993e-06,
+      "loss": 0.5186,
+      "step": 106
+    },
+    {
+      "epoch": 0.5504823151125402,
+      "grad_norm": 0.46560072898864746,
+      "learning_rate": 4.533685156777008e-06,
+      "loss": 0.4899,
+      "step": 107
+    },
+    {
+      "epoch": 0.5556270096463023,
+      "grad_norm": 0.4533552825450897,
+      "learning_rate": 4.449218615380029e-06,
+      "loss": 0.486,
+      "step": 108
+    },
+    {
+      "epoch": 0.5607717041800643,
+      "grad_norm": 0.46256566047668457,
+      "learning_rate": 4.364910901265607e-06,
+      "loss": 0.5113,
+      "step": 109
+    },
+    {
+      "epoch": 0.5659163987138264,
+      "grad_norm": 0.483781635761261,
+      "learning_rate": 4.280786326012724e-06,
+      "loss": 0.4721,
+      "step": 110
+    },
+    {
+      "epoch": 0.5710610932475885,
+      "grad_norm": 0.45272862911224365,
+      "learning_rate": 4.196869148389114e-06,
+      "loss": 0.5051,
+      "step": 111
+    },
+    {
+      "epoch": 0.5762057877813505,
+      "grad_norm": 0.4708820879459381,
+      "learning_rate": 4.113183567355846e-06,
+      "loss": 0.487,
+      "step": 112
+    },
+    {
+      "epoch": 0.5813504823151125,
+      "grad_norm": 0.5137550830841064,
+      "learning_rate": 4.029753715089124e-06,
+      "loss": 0.5195,
+      "step": 113
+    },
+    {
+      "epoch": 0.5864951768488746,
+      "grad_norm": 0.47741350531578064,
+      "learning_rate": 3.94660365002137e-06,
+      "loss": 0.5042,
+      "step": 114
+    },
+    {
+      "epoch": 0.5916398713826366,
+      "grad_norm": 0.5155045986175537,
+      "learning_rate": 3.863757349903551e-06,
+      "loss": 0.471,
+      "step": 115
+    },
+    {
+      "epoch": 0.5967845659163987,
+      "grad_norm": 0.46846747398376465,
+      "learning_rate": 3.781238704890793e-06,
+      "loss": 0.5166,
+      "step": 116
+    },
+    {
+      "epoch": 0.6019292604501608,
+      "grad_norm": 0.4709222912788391,
+      "learning_rate": 3.6990715106532353e-06,
+      "loss": 0.5094,
+      "step": 117
+    },
+    {
+      "epoch": 0.6070739549839228,
+      "grad_norm": 0.5091134309768677,
+      "learning_rate": 3.6172794615141448e-06,
+      "loss": 0.4486,
+      "step": 118
+    },
+    {
+      "epoch": 0.6122186495176849,
+      "grad_norm": 0.4649624824523926,
+      "learning_rate": 3.5358861436172487e-06,
+      "loss": 0.5048,
+      "step": 119
+    },
+    {
+      "epoch": 0.617363344051447,
+      "grad_norm": 0.5000843405723572,
+      "learning_rate": 3.4549150281252635e-06,
+      "loss": 0.4784,
+      "step": 120
+    },
+    {
+      "epoch": 0.622508038585209,
+      "grad_norm": 0.4720861613750458,
+      "learning_rate": 3.374389464451583e-06,
+      "loss": 0.4794,
+      "step": 121
+    },
+    {
+      "epoch": 0.6276527331189711,
+      "grad_norm": 0.47109055519104004,
+      "learning_rate": 3.2943326735270766e-06,
+      "loss": 0.4828,
+      "step": 122
+    },
+    {
+      "epoch": 0.6327974276527332,
+      "grad_norm": 0.4736097455024719,
+      "learning_rate": 3.2147677411039236e-06,
+      "loss": 0.4627,
+      "step": 123
+    },
+    {
+      "epoch": 0.6379421221864952,
+      "grad_norm": 0.4775114059448242,
+      "learning_rate": 3.1357176110984578e-06,
+      "loss": 0.4861,
+      "step": 124
+    },
+    {
+      "epoch": 0.6430868167202572,
+      "grad_norm": 0.46678417921066284,
+      "learning_rate": 3.0572050789748732e-06,
+      "loss": 0.4946,
+      "step": 125
+    },
+    {
+      "epoch": 0.6482315112540193,
+      "grad_norm": 0.44837188720703125,
+      "learning_rate": 2.9792527851717805e-06,
+      "loss": 0.5074,
+      "step": 126
+    },
+    {
+      "epoch": 0.6533762057877813,
+      "grad_norm": 0.4877527952194214,
+      "learning_rate": 2.90188320857343e-06,
+      "loss": 0.5205,
+      "step": 127
+    },
+    {
+      "epoch": 0.6585209003215434,
+      "grad_norm": 0.4547199308872223,
+      "learning_rate": 2.8251186600275533e-06,
+      "loss": 0.4647,
+      "step": 128
+    },
+    {
+      "epoch": 0.6636655948553055,
+      "grad_norm": 0.45252618193626404,
+      "learning_rate": 2.748981275911633e-06,
+      "loss": 0.5093,
+      "step": 129
+    },
+    {
+      "epoch": 0.6688102893890675,
+      "grad_norm": 0.4427642822265625,
+      "learning_rate": 2.6734930117495133e-06,
+      "loss": 0.4422,
+      "step": 130
+    },
+    {
+      "epoch": 0.6739549839228296,
+      "grad_norm": 0.47340139746665955,
+      "learning_rate": 2.598675635880129e-06,
+      "loss": 0.475,
+      "step": 131
+    },
+    {
+      "epoch": 0.6790996784565917,
+      "grad_norm": 0.47648149728775024,
+      "learning_rate": 2.524550723180249e-06,
+      "loss": 0.4729,
+      "step": 132
+    },
+    {
+      "epoch": 0.6842443729903537,
+      "grad_norm": 0.4834476709365845,
+      "learning_rate": 2.4511396488429727e-06,
+      "loss": 0.4899,
+      "step": 133
+    },
+    {
+      "epoch": 0.6893890675241158,
+      "grad_norm": 0.45669957995414734,
+      "learning_rate": 2.3784635822138424e-06,
+      "loss": 0.5043,
+      "step": 134
+    },
+    {
+      "epoch": 0.6945337620578779,
+      "grad_norm": 0.4470764696598053,
+      "learning_rate": 2.306543480686281e-06,
+      "loss": 0.4704,
+      "step": 135
+    },
+    {
+      "epoch": 0.6996784565916399,
+      "grad_norm": 0.4960187077522278,
+      "learning_rate": 2.2354000836581834e-06,
+      "loss": 0.5208,
+      "step": 136
+    },
+    {
+      "epoch": 0.704823151125402,
+      "grad_norm": 0.4489913284778595,
+      "learning_rate": 2.1650539065513415e-06,
+      "loss": 0.4592,
+      "step": 137
+    },
+    {
+      "epoch": 0.7099678456591639,
+      "grad_norm": 0.46100521087646484,
+      "learning_rate": 2.0955252348954806e-06,
+      "loss": 0.4452,
+      "step": 138
+    },
+    {
+      "epoch": 0.715112540192926,
+      "grad_norm": 0.4687739908695221,
+      "learning_rate": 2.0268341184785674e-06,
+      "loss": 0.5146,
+      "step": 139
+    },
+    {
+      "epoch": 0.7202572347266881,
+      "grad_norm": 0.4546753168106079,
+      "learning_rate": 1.959000365565119e-06,
+      "loss": 0.4707,
+      "step": 140
+    },
+    {
+      "epoch": 0.7254019292604501,
+      "grad_norm": 0.4647018015384674,
+      "learning_rate": 1.8920435371841394e-06,
+      "loss": 0.4379,
+      "step": 141
+    },
+    {
+      "epoch": 0.7305466237942122,
+      "grad_norm": 0.473733127117157,
+      "learning_rate": 1.8259829414883728e-06,
+      "loss": 0.4845,
+      "step": 142
+    },
+    {
+      "epoch": 0.7356913183279743,
+      "grad_norm": 0.49459928274154663,
+      "learning_rate": 1.7608376281864502e-06,
+      "loss": 0.5376,
+      "step": 143
+    },
+    {
+      "epoch": 0.7408360128617363,
+      "grad_norm": 0.46276187896728516,
+      "learning_rate": 1.6966263830495939e-06,
+      "loss": 0.4691,
+      "step": 144
+    },
+    {
+      "epoch": 0.7459807073954984,
+      "grad_norm": 0.45211291313171387,
+      "learning_rate": 1.633367722494404e-06,
+      "loss": 0.4521,
+      "step": 145
+    },
+    {
+      "epoch": 0.7511254019292605,
+      "grad_norm": 0.4516507387161255,
+      "learning_rate": 1.5710798882433432e-06,
+      "loss": 0.4897,
+      "step": 146
+    },
+    {
+      "epoch": 0.7562700964630225,
+      "grad_norm": 0.46488216519355774,
+      "learning_rate": 1.5097808420644117e-06,
+      "loss": 0.4875,
+      "step": 147
+    },
+    {
+      "epoch": 0.7614147909967846,
+      "grad_norm": 0.47542884945869446,
+      "learning_rate": 1.4494882605915717e-06,
+      "loss": 0.4942,
+      "step": 148
+    },
+    {
+      "epoch": 0.7665594855305466,
+      "grad_norm": 0.4502287209033966,
+      "learning_rate": 1.390219530227378e-06,
+      "loss": 0.482,
+      "step": 149
+    },
+    {
+      "epoch": 0.7717041800643086,
+      "grad_norm": 0.45821210741996765,
+      "learning_rate": 1.3319917421293182e-06,
+      "loss": 0.5331,
+      "step": 150
+    },
+    {
+      "epoch": 0.7768488745980707,
+      "grad_norm": 0.4743272364139557,
+      "learning_rate": 1.2748216872812747e-06,
+      "loss": 0.5326,
+      "step": 151
+    },
+    {
+      "epoch": 0.7819935691318328,
+      "grad_norm": 0.45159199833869934,
+      "learning_rate": 1.2187258516515644e-06,
+      "loss": 0.4824,
+      "step": 152
+    },
+    {
+      "epoch": 0.7871382636655948,
+      "grad_norm": 0.4775025546550751,
+      "learning_rate": 1.1637204114389179e-06,
+      "loss": 0.5049,
+      "step": 153
+    },
+    {
+      "epoch": 0.7922829581993569,
+      "grad_norm": 0.4442935883998871,
+      "learning_rate": 1.1098212284078037e-06,
+      "loss": 0.4555,
+      "step": 154
+    },
+    {
+      "epoch": 0.797427652733119,
+      "grad_norm": 0.4545031487941742,
+      "learning_rate": 1.0570438453144044e-06,
+      "loss": 0.4885,
+      "step": 155
+    },
+    {
+      "epoch": 0.802572347266881,
+      "grad_norm": 0.4795566499233246,
+      "learning_rate": 1.0054034814246093e-06,
+      "loss": 0.5527,
+      "step": 156
+    },
+    {
+      "epoch": 0.8077170418006431,
+      "grad_norm": 0.4730350375175476,
+      "learning_rate": 9.549150281252633e-07,
+      "loss": 0.5218,
+      "step": 157
+    },
+    {
+      "epoch": 0.8128617363344052,
+      "grad_norm": 0.4760991632938385,
+      "learning_rate": 9.055930446299916e-07,
+      "loss": 0.5136,
+      "step": 158
+    },
+    {
+      "epoch": 0.8180064308681672,
+      "grad_norm": 0.4581911563873291,
+      "learning_rate": 8.574517537807897e-07,
+      "loss": 0.5176,
+      "step": 159
+    },
+    {
+      "epoch": 0.8231511254019293,
+      "grad_norm": 0.451224684715271,
+      "learning_rate": 8.105050379466334e-07,
+      "loss": 0.4777,
+      "step": 160
+    },
+    {
+      "epoch": 0.8282958199356913,
+      "grad_norm": 0.46320095658302307,
+      "learning_rate": 7.647664350202461e-07,
+      "loss": 0.4835,
+      "step": 161
+    },
+    {
+      "epoch": 0.8334405144694533,
+      "grad_norm": 0.4722878038883209,
+      "learning_rate": 7.202491345142288e-07,
+      "loss": 0.4445,
+      "step": 162
+    },
+    {
+      "epoch": 0.8385852090032154,
+      "grad_norm": 0.46132737398147583,
+      "learning_rate": 6.769659737576229e-07,
+      "loss": 0.4964,
+      "step": 163
+    },
+    {
+      "epoch": 0.8437299035369775,
+      "grad_norm": 0.478231817483902,
+      "learning_rate": 6.349294341940593e-07,
+      "loss": 0.5356,
+      "step": 164
+    },
+    {
+      "epoch": 0.8488745980707395,
+      "grad_norm": 0.44469162821769714,
+      "learning_rate": 5.941516377825102e-07,
+      "loss": 0.4855,
+      "step": 165
+    },
+    {
+      "epoch": 0.8540192926045016,
+      "grad_norm": 0.48774901032447815,
+      "learning_rate": 5.546443435017146e-07,
+      "loss": 0.4628,
+      "step": 166
+    },
+    {
+      "epoch": 0.8591639871382637,
+      "grad_norm": 0.44260117411613464,
+      "learning_rate": 5.16418943959282e-07,
+      "loss": 0.456,
+      "step": 167
+    },
+    {
+      "epoch": 0.8643086816720257,
+      "grad_norm": 0.48577848076820374,
+      "learning_rate": 4.794864621064266e-07,
+      "loss": 0.4974,
+      "step": 168
+    },
+    {
+      "epoch": 0.8694533762057878,
+      "grad_norm": 0.4491365849971771,
+      "learning_rate": 4.43857548059321e-07,
+      "loss": 0.4811,
+      "step": 169
+    },
+    {
+      "epoch": 0.8745980707395499,
+      "grad_norm": 0.4400619566440582,
+      "learning_rate": 4.095424760279454e-07,
+      "loss": 0.4606,
+      "step": 170
+    },
+    {
+      "epoch": 0.8797427652733119,
+      "grad_norm": 0.4469500184059143,
+      "learning_rate": 3.765511413533429e-07,
+      "loss": 0.512,
+      "step": 171
+    },
+    {
+      "epoch": 0.884887459807074,
+      "grad_norm": 0.45204490423202515,
+      "learning_rate": 3.4489305765413097e-07,
+      "loss": 0.4551,
+      "step": 172
+    },
+    {
+      "epoch": 0.890032154340836,
+      "grad_norm": 0.46731144189834595,
+      "learning_rate": 3.1457735408308153e-07,
+      "loss": 0.4868,
+      "step": 173
+    },
+    {
+      "epoch": 0.895176848874598,
+      "grad_norm": 0.4576306641101837,
+      "learning_rate": 2.85612772694579e-07,
+      "loss": 0.4504,
+      "step": 174
+    },
+    {
+      "epoch": 0.9003215434083601,
+      "grad_norm": 0.4661747217178345,
+      "learning_rate": 2.5800766592369077e-07,
+      "loss": 0.5027,
+      "step": 175
+    },
+    {
+      "epoch": 0.9054662379421222,
+      "grad_norm": 0.44758227467536926,
+      "learning_rate": 2.3176999417760637e-07,
+      "loss": 0.4859,
+      "step": 176
+    },
+    {
+      "epoch": 0.9106109324758842,
+      "grad_norm": 0.4612000584602356,
+      "learning_rate": 2.069073235401109e-07,
+      "loss": 0.5142,
+      "step": 177
+    },
+    {
+      "epoch": 0.9157556270096463,
+      "grad_norm": 0.4407828450202942,
+      "learning_rate": 1.834268235897807e-07,
+      "loss": 0.4805,
+      "step": 178
+    },
+    {
+      "epoch": 0.9209003215434084,
+      "grad_norm": 0.48513102531433105,
+      "learning_rate": 1.6133526533250566e-07,
+      "loss": 0.5101,
+      "step": 179
+    },
+    {
+      "epoch": 0.9260450160771704,
+      "grad_norm": 0.4615355134010315,
+      "learning_rate": 1.4063901924895985e-07,
+      "loss": 0.4986,
+      "step": 180
+    },
+    {
+      "epoch": 0.9311897106109325,
+      "grad_norm": 0.4797261655330658,
+      "learning_rate": 1.2134405345755773e-07,
+      "loss": 0.5108,
+      "step": 181
+    },
+    {
+      "epoch": 0.9363344051446946,
+      "grad_norm": 0.4593569338321686,
+      "learning_rate": 1.0345593199344972e-07,
+      "loss": 0.5166,
+      "step": 182
+    },
+    {
+      "epoch": 0.9414790996784566,
+      "grad_norm": 0.478330135345459,
+      "learning_rate": 8.697981320403337e-08,
+      "loss": 0.5229,
+      "step": 183
+    },
+    {
+      "epoch": 0.9466237942122186,
+      "grad_norm": 0.44957953691482544,
+      "learning_rate": 7.192044826145772e-08,
+      "loss": 0.469,
+      "step": 184
+    },
+    {
+      "epoch": 0.9517684887459807,
+      "grad_norm": 0.4661603569984436,
+      "learning_rate": 5.82821797925387e-08,
+      "loss": 0.4792,
+      "step": 185
+    },
+    {
+      "epoch": 0.9569131832797427,
+      "grad_norm": 0.4424455165863037,
+      "learning_rate": 4.60689406264897e-08,
+      "loss": 0.458,
+      "step": 186
+    },
+    {
+      "epoch": 0.9620578778135048,
+      "grad_norm": 0.4485902786254883,
+      "learning_rate": 3.528425266082325e-08,
+      "loss": 0.474,
+      "step": 187
+    },
+    {
+      "epoch": 0.9672025723472669,
+      "grad_norm": 0.43381765484809875,
+      "learning_rate": 2.593122584574892e-08,
+      "loss": 0.4625,
+      "step": 188
+    },
+    {
+      "epoch": 0.9723472668810289,
+      "grad_norm": 0.4678357243537903,
+      "learning_rate": 1.8012557287367394e-08,
+      "loss": 0.4841,
+      "step": 189
+    },
+    {
+      "epoch": 0.977491961414791,
+      "grad_norm": 0.47606784105300903,
+      "learning_rate": 1.1530530469914258e-08,
+      "loss": 0.5369,
+      "step": 190
+    },
+    {
+      "epoch": 0.9826366559485531,
+      "grad_norm": 0.44228696823120117,
+      "learning_rate": 6.487014597275631e-09,
+      "loss": 0.4824,
+      "step": 191
+    },
+    {
+      "epoch": 0.9877813504823151,
+      "grad_norm": 0.4588797390460968,
+      "learning_rate": 2.8834640539737723e-09,
+      "loss": 0.4457,
+      "step": 192
+    },
+    {
+      "epoch": 0.9929260450160772,
+      "grad_norm": 0.46713078022003174,
+      "learning_rate": 7.209179857675664e-10,
+      "loss": 0.5029,
+      "step": 193
+    },
+    {
+      "epoch": 0.9980707395498393,
+      "grad_norm": 0.4522612988948822,
+      "learning_rate": 0.0,
+      "loss": 0.4826,
+      "step": 194
+    },
+    {
+      "epoch": 0.9980707395498393,
+      "eval_loss": 0.46097689867019653,
+      "eval_runtime": 4.0485,
+      "eval_samples_per_second": 30.876,
+      "eval_steps_per_second": 3.952,
+      "step": 194
+    }
+  ],
+  "logging_steps": 1,
+  "max_steps": 194,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 1,
+  "save_steps": 500,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": true
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.384828301479772e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}

training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:269e347bc009e0cf055dc17e322a01f1dc99a6655c071f279d056932cc0063ad
+size 10872

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)