Upload folder using huggingface_hub

.gitattributes

@@ -35,3 +35,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
 base_model_215M filter=lfs diff=lfs merge=lfs -text
 pretrain_Qwen2.5-1.5B/output/checkpoint-390/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+sft_train_Qwen2.5-1.5B/output/checkpoint-900/tokenizer.json filter=lfs diff=lfs merge=lfs -text

sft_train_Qwen2.5-1.5B/output/checkpoint-900/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
+}

sft_train_Qwen2.5-1.5B/output/checkpoint-900/chat_template.jinja

@@ -0,0 +1,54 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0]['role'] == 'system' %}
+        {{- messages[0]['content'] }}
+    {%- else %}
+        {{- 'You are a helpful assistant.' }}
+    {%- endif %}
+    {{- "\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>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\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" }}
+{%- else %}
+    {%- if messages[0]['role'] == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0]['content'] + '<|im_end|>\n' }}
+    {%- else %}
+        {{- '<|im_start|>system\nYou are a helpful assistant.<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) or (message.role == "assistant" and not message.tool_calls) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {{- '<|im_start|>' + message.role }}
+        {%- if message.content %}
+            {{- '\n' + message.content }}
+        {%- endif %}
+        {%- for tool_call in message.tool_calls %}
+            {%- if tool_call.function is defined %}
+                {%- set tool_call = tool_call.function %}
+            {%- endif %}
+            {{- '\n<tool_call>\n{"name": "' }}
+            {{- tool_call.name }}
+            {{- '", "arguments": ' }}
+            {{- tool_call.arguments | tojson }}
+            {{- '}\n</tool_call>' }}
+        {%- endfor %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if (loop.index0 == 0) or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+{%- endif %}

sft_train_Qwen2.5-1.5B/output/checkpoint-900/config.json

@@ -0,0 +1,59 @@
+{
+  "architectures": [
+    "Qwen2ForCausalLM"
+  ],
+  "attention_dropout": 0.0,
+  "dtype": "bfloat16",
+  "eos_token_id": 151643,
+  "hidden_act": "silu",
+  "hidden_size": 1536,
+  "initializer_range": 0.02,
+  "intermediate_size": 8960,
+  "layer_types": [
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention"
+  ],
+  "max_position_embeddings": 131072,
+  "max_window_layers": 28,
+  "model_type": "qwen2",
+  "num_attention_heads": 12,
+  "num_hidden_layers": 28,
+  "num_key_value_heads": 2,
+  "pad_token_id": 151643,
+  "rms_norm_eps": 1e-06,
+  "rope_scaling": null,
+  "rope_theta": 1000000.0,
+  "sliding_window": null,
+  "tie_word_embeddings": true,
+  "transformers_version": "4.57.3",
+  "use_cache": true,
+  "use_mrope": false,
+  "use_sliding_window": false,
+  "vocab_size": 151936
+}

sft_train_Qwen2.5-1.5B/output/checkpoint-900/generation_config.json

@@ -0,0 +1,8 @@
+{
+  "eos_token_id": [
+    151643
+  ],
+  "max_new_tokens": 2048,
+  "pad_token_id": 151643,
+  "transformers_version": "4.57.3"
+}

sft_train_Qwen2.5-1.5B/output/checkpoint-900/global_step900/bf16_zero_pp_rank_0_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:39cc62a1f5e68565f69c0f3a3a58e66a8992c8524890bcb21ef61b78e678bd44
+size 18524602449

sft_train_Qwen2.5-1.5B/output/checkpoint-900/global_step900/mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5a37e0c34c4d2fb043e87bc05c1a36038496a45c29ecc11fc26b5091accefb7a
+size 3087521177

sft_train_Qwen2.5-1.5B/output/checkpoint-900/latest

@@ -0,0 +1 @@
+global_step900

sft_train_Qwen2.5-1.5B/output/checkpoint-900/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b2936862d29959d2b9a34d1c4a9f4ed7e03ac8884afbcc97bd81cbf0eba3eb40
+size 3554214752

sft_train_Qwen2.5-1.5B/output/checkpoint-900/rng_state.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2527c1c927a9fcfad4470e680222a0a0635da820509c6fdfb6fb31471ed46aed
+size 14645

sft_train_Qwen2.5-1.5B/output/checkpoint-900/special_tokens_map.json

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

sft_train_Qwen2.5-1.5B/output/checkpoint-900/tokenizer.json

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

sft_train_Qwen2.5-1.5B/output/checkpoint-900/tokenizer_config.json

@@ -0,0 +1,207 @@
+{
+  "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,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": {},
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}

sft_train_Qwen2.5-1.5B/output/checkpoint-900/trainer_state.json

@@ -0,0 +1,1294 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.8768000000000002,
+  "eval_steps": 500,
+  "global_step": 900,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.016,
+      "grad_norm": 0.28587257862091064,
+      "learning_rate": 3.0376396902828087e-05,
+      "loss": 3.2411,
+      "step": 5
+    },
+    {
+      "epoch": 0.032,
+      "grad_norm": 0.11348269879817963,
+      "learning_rate": 4.3458798967609374e-05,
+      "loss": 0.3728,
+      "step": 10
+    },
+    {
+      "epoch": 0.048,
+      "grad_norm": 0.1157122552394867,
+      "learning_rate": 5.111151359486345e-05,
+      "loss": 0.2897,
+      "step": 15
+    },
+    {
+      "epoch": 0.064,
+      "grad_norm": 0.09483788162469864,
+      "learning_rate": 5.6541201032390644e-05,
+      "loss": 0.2627,
+      "step": 20
+    },
+    {
+      "epoch": 0.08,
+      "grad_norm": 0.031023913994431496,
+      "learning_rate": 6.0752793805656173e-05,
+      "loss": 0.2285,
+      "step": 25
+    },
+    {
+      "epoch": 0.096,
+      "grad_norm": 0.03515796363353729,
+      "learning_rate": 6.419391565964472e-05,
+      "loss": 0.244,
+      "step": 30
+    },
+    {
+      "epoch": 0.112,
+      "grad_norm": 0.03319098800420761,
+      "learning_rate": 6.710334273172837e-05,
+      "loss": 0.2256,
+      "step": 35
+    },
+    {
+      "epoch": 0.128,
+      "grad_norm": 0.026092184707522392,
+      "learning_rate": 6.962360309717191e-05,
+      "loss": 0.2129,
+      "step": 40
+    },
+    {
+      "epoch": 0.144,
+      "grad_norm": 0.028749588876962662,
+      "learning_rate": 7.184663028689878e-05,
+      "loss": 0.1969,
+      "step": 45
+    },
+    {
+      "epoch": 0.16,
+      "grad_norm": 0.018630679696798325,
+      "learning_rate": 7.383519587043745e-05,
+      "loss": 0.1978,
+      "step": 50
+    },
+    {
+      "epoch": 0.176,
+      "grad_norm": 0.01749802753329277,
+      "learning_rate": 7.56340722534583e-05,
+      "loss": 0.2027,
+      "step": 55
+    },
+    {
+      "epoch": 0.192,
+      "grad_norm": 0.015604503452777863,
+      "learning_rate": 7.727631772442599e-05,
+      "loss": 0.2093,
+      "step": 60
+    },
+    {
+      "epoch": 0.208,
+      "grad_norm": 0.017327846959233284,
+      "learning_rate": 7.878703711203575e-05,
+      "loss": 0.179,
+      "step": 65
+    },
+    {
+      "epoch": 0.224,
+      "grad_norm": 0.019031211733818054,
+      "learning_rate": 8.018574479650967e-05,
+      "loss": 0.2262,
+      "step": 70
+    },
+    {
+      "epoch": 0.24,
+      "grad_norm": 0.018662473186850548,
+      "learning_rate": 8.148791049769153e-05,
+      "loss": 0.2065,
+      "step": 75
+    },
+    {
+      "epoch": 0.256,
+      "grad_norm": 0.015872152522206306,
+      "learning_rate": 8.27060051619532e-05,
+      "loss": 0.192,
+      "step": 80
+    },
+    {
+      "epoch": 0.272,
+      "grad_norm": 0.017076680436730385,
+      "learning_rate": 8.385022921691829e-05,
+      "loss": 0.1903,
+      "step": 85
+    },
+    {
+      "epoch": 0.288,
+      "grad_norm": 0.015185699798166752,
+      "learning_rate": 8.492903235168008e-05,
+      "loss": 0.2045,
+      "step": 90
+    },
+    {
+      "epoch": 0.304,
+      "grad_norm": 0.014806922525167465,
+      "learning_rate": 8.594949257574365e-05,
+      "loss": 0.1769,
+      "step": 95
+    },
+    {
+      "epoch": 0.32,
+      "grad_norm": 0.015597835183143616,
+      "learning_rate": 8.691759793521875e-05,
+      "loss": 0.1891,
+      "step": 100
+    },
+    {
+      "epoch": 0.336,
+      "grad_norm": 0.014892748557031155,
+      "learning_rate": 8.783845942376372e-05,
+      "loss": 0.1943,
+      "step": 105
+    },
+    {
+      "epoch": 0.352,
+      "grad_norm": 0.015355441719293594,
+      "learning_rate": 8.87164743182396e-05,
+      "loss": 0.1965,
+      "step": 110
+    },
+    {
+      "epoch": 0.368,
+      "grad_norm": 0.015447203069925308,
+      "learning_rate": 8.955545317691439e-05,
+      "loss": 0.1789,
+      "step": 115
+    },
+    {
+      "epoch": 0.384,
+      "grad_norm": 0.01500595360994339,
+      "learning_rate": 9.035871978920727e-05,
+      "loss": 0.2024,
+      "step": 120
+    },
+    {
+      "epoch": 0.4,
+      "grad_norm": 0.016250288113951683,
+      "learning_rate": 9.112919070848428e-05,
+      "loss": 0.1895,
+      "step": 125
+    },
+    {
+      "epoch": 0.416,
+      "grad_norm": 0.015516160987317562,
+      "learning_rate": 9.186943917681704e-05,
+      "loss": 0.1804,
+      "step": 130
+    },
+    {
+      "epoch": 0.432,
+      "grad_norm": 0.016289617866277695,
+      "learning_rate": 9.258174697893415e-05,
+      "loss": 0.1733,
+      "step": 135
+    },
+    {
+      "epoch": 0.448,
+      "grad_norm": 0.015045730397105217,
+      "learning_rate": 9.326814686129093e-05,
+      "loss": 0.1751,
+      "step": 140
+    },
+    {
+      "epoch": 0.464,
+      "grad_norm": 0.014139682054519653,
+      "learning_rate": 9.393045750415324e-05,
+      "loss": 0.1831,
+      "step": 145
+    },
+    {
+      "epoch": 0.48,
+      "grad_norm": 0.018029669299721718,
+      "learning_rate": 9.457031256247281e-05,
+      "loss": 0.1994,
+      "step": 150
+    },
+    {
+      "epoch": 0.496,
+      "grad_norm": 0.013372590765357018,
+      "learning_rate": 9.518918494316514e-05,
+      "loss": 0.1919,
+      "step": 155
+    },
+    {
+      "epoch": 0.512,
+      "grad_norm": 0.014621068723499775,
+      "learning_rate": 9.578840722673448e-05,
+      "loss": 0.1801,
+      "step": 160
+    },
+    {
+      "epoch": 0.528,
+      "grad_norm": 0.016799382865428925,
+      "learning_rate": 9.636918894549365e-05,
+      "loss": 0.2055,
+      "step": 165
+    },
+    {
+      "epoch": 0.544,
+      "grad_norm": 0.013853991404175758,
+      "learning_rate": 9.693263128169958e-05,
+      "loss": 0.1871,
+      "step": 170
+    },
+    {
+      "epoch": 0.56,
+      "grad_norm": 0.015933020040392876,
+      "learning_rate": 9.747973963455648e-05,
+      "loss": 0.2115,
+      "step": 175
+    },
+    {
+      "epoch": 0.576,
+      "grad_norm": 0.015511302277445793,
+      "learning_rate": 9.801143441646136e-05,
+      "loss": 0.1983,
+      "step": 180
+    },
+    {
+      "epoch": 0.592,
+      "grad_norm": 0.014855251647531986,
+      "learning_rate": 9.852856036971404e-05,
+      "loss": 0.1795,
+      "step": 185
+    },
+    {
+      "epoch": 0.608,
+      "grad_norm": 0.014359841123223305,
+      "learning_rate": 9.903189464052493e-05,
+      "loss": 0.2008,
+      "step": 190
+    },
+    {
+      "epoch": 0.624,
+      "grad_norm": 0.01440009381622076,
+      "learning_rate": 9.952215380407112e-05,
+      "loss": 0.2027,
+      "step": 195
+    },
+    {
+      "epoch": 0.64,
+      "grad_norm": 0.014104939065873623,
+      "learning_rate": 0.0001,
+      "loss": 0.2136,
+      "step": 200
+    },
+    {
+      "epoch": 0.656,
+      "grad_norm": 0.013765454292297363,
+      "learning_rate": 0.0001,
+      "loss": 0.2055,
+      "step": 205
+    },
+    {
+      "epoch": 0.672,
+      "grad_norm": 0.014615650288760662,
+      "learning_rate": 0.0001,
+      "loss": 0.1979,
+      "step": 210
+    },
+    {
+      "epoch": 0.688,
+      "grad_norm": 0.015578454360365868,
+      "learning_rate": 0.0001,
+      "loss": 0.2063,
+      "step": 215
+    },
+    {
+      "epoch": 0.704,
+      "grad_norm": 0.01497430820018053,
+      "learning_rate": 0.0001,
+      "loss": 0.1825,
+      "step": 220
+    },
+    {
+      "epoch": 0.72,
+      "grad_norm": 0.013819866813719273,
+      "learning_rate": 0.0001,
+      "loss": 0.2087,
+      "step": 225
+    },
+    {
+      "epoch": 0.736,
+      "grad_norm": 0.015137173235416412,
+      "learning_rate": 0.0001,
+      "loss": 0.1959,
+      "step": 230
+    },
+    {
+      "epoch": 0.752,
+      "grad_norm": 0.01289480458945036,
+      "learning_rate": 0.0001,
+      "loss": 0.1929,
+      "step": 235
+    },
+    {
+      "epoch": 0.768,
+      "grad_norm": 0.013631526380777359,
+      "learning_rate": 0.0001,
+      "loss": 0.1785,
+      "step": 240
+    },
+    {
+      "epoch": 0.784,
+      "grad_norm": 0.014563720673322678,
+      "learning_rate": 0.0001,
+      "loss": 0.1963,
+      "step": 245
+    },
+    {
+      "epoch": 0.8,
+      "grad_norm": 0.012929687276482582,
+      "learning_rate": 0.0001,
+      "loss": 0.2059,
+      "step": 250
+    },
+    {
+      "epoch": 0.816,
+      "grad_norm": 0.016089975833892822,
+      "learning_rate": 0.0001,
+      "loss": 0.1839,
+      "step": 255
+    },
+    {
+      "epoch": 0.832,
+      "grad_norm": 0.01351515669375658,
+      "learning_rate": 0.0001,
+      "loss": 0.1992,
+      "step": 260
+    },
+    {
+      "epoch": 0.848,
+      "grad_norm": 0.015158702619373798,
+      "learning_rate": 0.0001,
+      "loss": 0.1998,
+      "step": 265
+    },
+    {
+      "epoch": 0.864,
+      "grad_norm": 0.012720300815999508,
+      "learning_rate": 0.0001,
+      "loss": 0.2056,
+      "step": 270
+    },
+    {
+      "epoch": 0.88,
+      "grad_norm": 0.19446808099746704,
+      "learning_rate": 0.0001,
+      "loss": 0.2368,
+      "step": 275
+    },
+    {
+      "epoch": 0.896,
+      "grad_norm": 0.03085729293525219,
+      "learning_rate": 0.0001,
+      "loss": 0.2156,
+      "step": 280
+    },
+    {
+      "epoch": 0.912,
+      "grad_norm": 0.019313033670186996,
+      "learning_rate": 0.0001,
+      "loss": 0.2155,
+      "step": 285
+    },
+    {
+      "epoch": 0.928,
+      "grad_norm": 0.04357197508215904,
+      "learning_rate": 0.0001,
+      "loss": 0.2169,
+      "step": 290
+    },
+    {
+      "epoch": 0.944,
+      "grad_norm": 0.016985150054097176,
+      "learning_rate": 0.0001,
+      "loss": 0.1925,
+      "step": 295
+    },
+    {
+      "epoch": 0.96,
+      "grad_norm": 0.016494382172822952,
+      "learning_rate": 0.0001,
+      "loss": 0.2156,
+      "step": 300
+    },
+    {
+      "epoch": 0.976,
+      "grad_norm": 0.018017994239926338,
+      "learning_rate": 0.0001,
+      "loss": 0.213,
+      "step": 305
+    },
+    {
+      "epoch": 0.992,
+      "grad_norm": 0.04853694513440132,
+      "learning_rate": 0.0001,
+      "loss": 0.2181,
+      "step": 310
+    },
+    {
+      "epoch": 1.0064,
+      "grad_norm": 0.014205807819962502,
+      "learning_rate": 0.0001,
+      "loss": 0.2023,
+      "step": 315
+    },
+    {
+      "epoch": 1.0224,
+      "grad_norm": 0.019374940544366837,
+      "learning_rate": 0.0001,
+      "loss": 0.215,
+      "step": 320
+    },
+    {
+      "epoch": 1.0384,
+      "grad_norm": 0.024870608001947403,
+      "learning_rate": 0.0001,
+      "loss": 0.1994,
+      "step": 325
+    },
+    {
+      "epoch": 1.0544,
+      "grad_norm": 0.015315101481974125,
+      "learning_rate": 0.0001,
+      "loss": 0.1903,
+      "step": 330
+    },
+    {
+      "epoch": 1.0704,
+      "grad_norm": 0.01554437167942524,
+      "learning_rate": 0.0001,
+      "loss": 0.169,
+      "step": 335
+    },
+    {
+      "epoch": 1.0864,
+      "grad_norm": 0.017209002748131752,
+      "learning_rate": 0.0001,
+      "loss": 0.1823,
+      "step": 340
+    },
+    {
+      "epoch": 1.1024,
+      "grad_norm": 0.013879024423658848,
+      "learning_rate": 0.0001,
+      "loss": 0.1704,
+      "step": 345
+    },
+    {
+      "epoch": 1.1184,
+      "grad_norm": 0.01458608079701662,
+      "learning_rate": 0.0001,
+      "loss": 0.1817,
+      "step": 350
+    },
+    {
+      "epoch": 1.1344,
+      "grad_norm": 0.01765730045735836,
+      "learning_rate": 0.0001,
+      "loss": 0.181,
+      "step": 355
+    },
+    {
+      "epoch": 1.1504,
+      "grad_norm": 0.015403369441628456,
+      "learning_rate": 0.0001,
+      "loss": 0.1541,
+      "step": 360
+    },
+    {
+      "epoch": 1.1663999999999999,
+      "grad_norm": 0.014304639771580696,
+      "learning_rate": 0.0001,
+      "loss": 0.163,
+      "step": 365
+    },
+    {
+      "epoch": 1.1824,
+      "grad_norm": 0.017449308186769485,
+      "learning_rate": 0.0001,
+      "loss": 0.1917,
+      "step": 370
+    },
+    {
+      "epoch": 1.1984,
+      "grad_norm": 0.01425285916775465,
+      "learning_rate": 0.0001,
+      "loss": 0.1441,
+      "step": 375
+    },
+    {
+      "epoch": 1.2144,
+      "grad_norm": 0.021493326872587204,
+      "learning_rate": 0.0001,
+      "loss": 0.1656,
+      "step": 380
+    },
+    {
+      "epoch": 1.2304,
+      "grad_norm": 0.017522413283586502,
+      "learning_rate": 0.0001,
+      "loss": 0.1826,
+      "step": 385
+    },
+    {
+      "epoch": 1.2464,
+      "grad_norm": 0.016096558421850204,
+      "learning_rate": 0.0001,
+      "loss": 0.1562,
+      "step": 390
+    },
+    {
+      "epoch": 1.2624,
+      "grad_norm": 0.014397405087947845,
+      "learning_rate": 0.0001,
+      "loss": 0.1497,
+      "step": 395
+    },
+    {
+      "epoch": 1.2784,
+      "grad_norm": 0.015052678994834423,
+      "learning_rate": 0.0001,
+      "loss": 0.1653,
+      "step": 400
+    },
+    {
+      "epoch": 1.2944,
+      "grad_norm": 0.015351289883255959,
+      "learning_rate": 0.0001,
+      "loss": 0.1533,
+      "step": 405
+    },
+    {
+      "epoch": 1.3104,
+      "grad_norm": 0.013600205071270466,
+      "learning_rate": 0.0001,
+      "loss": 0.1359,
+      "step": 410
+    },
+    {
+      "epoch": 1.3264,
+      "grad_norm": 0.019016163423657417,
+      "learning_rate": 0.0001,
+      "loss": 0.1481,
+      "step": 415
+    },
+    {
+      "epoch": 1.3424,
+      "grad_norm": 0.01852385513484478,
+      "learning_rate": 0.0001,
+      "loss": 0.1594,
+      "step": 420
+    },
+    {
+      "epoch": 1.3584,
+      "grad_norm": 0.024075742810964584,
+      "learning_rate": 0.0001,
+      "loss": 0.1355,
+      "step": 425
+    },
+    {
+      "epoch": 1.3744,
+      "grad_norm": 0.015297528356313705,
+      "learning_rate": 0.0001,
+      "loss": 0.1471,
+      "step": 430
+    },
+    {
+      "epoch": 1.3904,
+      "grad_norm": 0.015145765617489815,
+      "learning_rate": 0.0001,
+      "loss": 0.1542,
+      "step": 435
+    },
+    {
+      "epoch": 1.4064,
+      "grad_norm": 0.01604115217924118,
+      "learning_rate": 0.0001,
+      "loss": 0.1421,
+      "step": 440
+    },
+    {
+      "epoch": 1.4224,
+      "grad_norm": 0.013240514323115349,
+      "learning_rate": 0.0001,
+      "loss": 0.127,
+      "step": 445
+    },
+    {
+      "epoch": 1.4384000000000001,
+      "grad_norm": 0.013415472581982613,
+      "learning_rate": 0.0001,
+      "loss": 0.1253,
+      "step": 450
+    },
+    {
+      "epoch": 1.4544000000000001,
+      "grad_norm": 0.017024055123329163,
+      "learning_rate": 0.0001,
+      "loss": 0.1431,
+      "step": 455
+    },
+    {
+      "epoch": 1.4704,
+      "grad_norm": 0.014443612657487392,
+      "learning_rate": 0.0001,
+      "loss": 0.1326,
+      "step": 460
+    },
+    {
+      "epoch": 1.4864,
+      "grad_norm": 0.015230610966682434,
+      "learning_rate": 0.0001,
+      "loss": 0.1374,
+      "step": 465
+    },
+    {
+      "epoch": 1.5024,
+      "grad_norm": 0.014205357991158962,
+      "learning_rate": 0.0001,
+      "loss": 0.1499,
+      "step": 470
+    },
+    {
+      "epoch": 1.5184,
+      "grad_norm": 0.01602974534034729,
+      "learning_rate": 0.0001,
+      "loss": 0.1214,
+      "step": 475
+    },
+    {
+      "epoch": 1.5344,
+      "grad_norm": 0.015949787572026253,
+      "learning_rate": 0.0001,
+      "loss": 0.1487,
+      "step": 480
+    },
+    {
+      "epoch": 1.5504,
+      "grad_norm": 0.0149220060557127,
+      "learning_rate": 0.0001,
+      "loss": 0.137,
+      "step": 485
+    },
+    {
+      "epoch": 1.5664,
+      "grad_norm": 0.016212308779358864,
+      "learning_rate": 0.0001,
+      "loss": 0.1574,
+      "step": 490
+    },
+    {
+      "epoch": 1.5824,
+      "grad_norm": 0.013444504700601101,
+      "learning_rate": 0.0001,
+      "loss": 0.1311,
+      "step": 495
+    },
+    {
+      "epoch": 1.5984,
+      "grad_norm": 0.016068238765001297,
+      "learning_rate": 0.0001,
+      "loss": 0.1343,
+      "step": 500
+    },
+    {
+      "epoch": 1.6143999999999998,
+      "grad_norm": 0.016453510150313377,
+      "learning_rate": 0.0001,
+      "loss": 0.1477,
+      "step": 505
+    },
+    {
+      "epoch": 1.6303999999999998,
+      "grad_norm": 0.018164295703172684,
+      "learning_rate": 0.0001,
+      "loss": 0.1415,
+      "step": 510
+    },
+    {
+      "epoch": 1.6463999999999999,
+      "grad_norm": 0.01658988557755947,
+      "learning_rate": 0.0001,
+      "loss": 0.1456,
+      "step": 515
+    },
+    {
+      "epoch": 1.6623999999999999,
+      "grad_norm": 0.014328578487038612,
+      "learning_rate": 0.0001,
+      "loss": 0.1426,
+      "step": 520
+    },
+    {
+      "epoch": 1.6784,
+      "grad_norm": 0.0176787618547678,
+      "learning_rate": 0.0001,
+      "loss": 0.1372,
+      "step": 525
+    },
+    {
+      "epoch": 1.6944,
+      "grad_norm": 0.016148056834936142,
+      "learning_rate": 0.0001,
+      "loss": 0.1408,
+      "step": 530
+    },
+    {
+      "epoch": 1.7104,
+      "grad_norm": 0.01808507926762104,
+      "learning_rate": 0.0001,
+      "loss": 0.1335,
+      "step": 535
+    },
+    {
+      "epoch": 1.7264,
+      "grad_norm": 0.01613294892013073,
+      "learning_rate": 0.0001,
+      "loss": 0.1391,
+      "step": 540
+    },
+    {
+      "epoch": 1.7424,
+      "grad_norm": 0.015190095640718937,
+      "learning_rate": 0.0001,
+      "loss": 0.1445,
+      "step": 545
+    },
+    {
+      "epoch": 1.7584,
+      "grad_norm": 0.014399568550288677,
+      "learning_rate": 0.0001,
+      "loss": 0.1301,
+      "step": 550
+    },
+    {
+      "epoch": 1.7744,
+      "grad_norm": 0.013615796342492104,
+      "learning_rate": 0.0001,
+      "loss": 0.1182,
+      "step": 555
+    },
+    {
+      "epoch": 1.7904,
+      "grad_norm": 0.013540414161980152,
+      "learning_rate": 0.0001,
+      "loss": 0.1487,
+      "step": 560
+    },
+    {
+      "epoch": 1.8064,
+      "grad_norm": 0.013552350923418999,
+      "learning_rate": 0.0001,
+      "loss": 0.1372,
+      "step": 565
+    },
+    {
+      "epoch": 1.8224,
+      "grad_norm": 0.014097240753471851,
+      "learning_rate": 0.0001,
+      "loss": 0.1411,
+      "step": 570
+    },
+    {
+      "epoch": 1.8384,
+      "grad_norm": 0.01584635116159916,
+      "learning_rate": 0.0001,
+      "loss": 0.1331,
+      "step": 575
+    },
+    {
+      "epoch": 1.8544,
+      "grad_norm": 0.015123779885470867,
+      "learning_rate": 0.0001,
+      "loss": 0.1357,
+      "step": 580
+    },
+    {
+      "epoch": 1.8704,
+      "grad_norm": 0.01653185486793518,
+      "learning_rate": 0.0001,
+      "loss": 0.1533,
+      "step": 585
+    },
+    {
+      "epoch": 1.8864,
+      "grad_norm": 0.014900973998010159,
+      "learning_rate": 0.0001,
+      "loss": 0.1421,
+      "step": 590
+    },
+    {
+      "epoch": 1.9024,
+      "grad_norm": 0.014678435400128365,
+      "learning_rate": 0.0001,
+      "loss": 0.1422,
+      "step": 595
+    },
+    {
+      "epoch": 1.9184,
+      "grad_norm": 0.01366644911468029,
+      "learning_rate": 0.0001,
+      "loss": 0.1492,
+      "step": 600
+    },
+    {
+      "epoch": 1.9344000000000001,
+      "grad_norm": 0.013686192221939564,
+      "learning_rate": 0.0001,
+      "loss": 0.1362,
+      "step": 605
+    },
+    {
+      "epoch": 1.9504000000000001,
+      "grad_norm": 0.02521859109401703,
+      "learning_rate": 0.0001,
+      "loss": 0.1581,
+      "step": 610
+    },
+    {
+      "epoch": 1.9664000000000001,
+      "grad_norm": 0.014916641637682915,
+      "learning_rate": 0.0001,
+      "loss": 0.143,
+      "step": 615
+    },
+    {
+      "epoch": 1.9824000000000002,
+      "grad_norm": 0.015870360657572746,
+      "learning_rate": 0.0001,
+      "loss": 0.1486,
+      "step": 620
+    },
+    {
+      "epoch": 1.9984,
+      "grad_norm": 0.014020348899066448,
+      "learning_rate": 0.0001,
+      "loss": 0.1512,
+      "step": 625
+    },
+    {
+      "epoch": 2.0128,
+      "grad_norm": 0.016190385445952415,
+      "learning_rate": 0.0001,
+      "loss": 0.138,
+      "step": 630
+    },
+    {
+      "epoch": 2.0288,
+      "grad_norm": 0.017395799979567528,
+      "learning_rate": 0.0001,
+      "loss": 0.1455,
+      "step": 635
+    },
+    {
+      "epoch": 2.0448,
+      "grad_norm": 0.01749918796122074,
+      "learning_rate": 0.0001,
+      "loss": 0.1264,
+      "step": 640
+    },
+    {
+      "epoch": 2.0608,
+      "grad_norm": 0.01680697873234749,
+      "learning_rate": 0.0001,
+      "loss": 0.121,
+      "step": 645
+    },
+    {
+      "epoch": 2.0768,
+      "grad_norm": 0.015204112976789474,
+      "learning_rate": 0.0001,
+      "loss": 0.1049,
+      "step": 650
+    },
+    {
+      "epoch": 2.0928,
+      "grad_norm": 0.01588457264006138,
+      "learning_rate": 0.0001,
+      "loss": 0.1185,
+      "step": 655
+    },
+    {
+      "epoch": 2.1088,
+      "grad_norm": 0.015460017137229443,
+      "learning_rate": 0.0001,
+      "loss": 0.1135,
+      "step": 660
+    },
+    {
+      "epoch": 2.1248,
+      "grad_norm": 0.019131796434521675,
+      "learning_rate": 0.0001,
+      "loss": 0.1157,
+      "step": 665
+    },
+    {
+      "epoch": 2.1408,
+      "grad_norm": 0.018924979493021965,
+      "learning_rate": 0.0001,
+      "loss": 0.0988,
+      "step": 670
+    },
+    {
+      "epoch": 2.1568,
+      "grad_norm": 0.01657225377857685,
+      "learning_rate": 0.0001,
+      "loss": 0.1001,
+      "step": 675
+    },
+    {
+      "epoch": 2.1728,
+      "grad_norm": 0.017764024436473846,
+      "learning_rate": 0.0001,
+      "loss": 0.1027,
+      "step": 680
+    },
+    {
+      "epoch": 2.1888,
+      "grad_norm": 0.016718421131372452,
+      "learning_rate": 0.0001,
+      "loss": 0.1106,
+      "step": 685
+    },
+    {
+      "epoch": 2.2048,
+      "grad_norm": 0.016799675300717354,
+      "learning_rate": 0.0001,
+      "loss": 0.0857,
+      "step": 690
+    },
+    {
+      "epoch": 2.2208,
+      "grad_norm": 0.01616666652262211,
+      "learning_rate": 0.0001,
+      "loss": 0.1063,
+      "step": 695
+    },
+    {
+      "epoch": 2.2368,
+      "grad_norm": 0.016462933272123337,
+      "learning_rate": 0.0001,
+      "loss": 0.1089,
+      "step": 700
+    },
+    {
+      "epoch": 2.2528,
+      "grad_norm": 0.01738760806620121,
+      "learning_rate": 0.0001,
+      "loss": 0.0941,
+      "step": 705
+    },
+    {
+      "epoch": 2.2688,
+      "grad_norm": 0.019756687805056572,
+      "learning_rate": 0.0001,
+      "loss": 0.0915,
+      "step": 710
+    },
+    {
+      "epoch": 2.2848,
+      "grad_norm": 0.018566954880952835,
+      "learning_rate": 0.0001,
+      "loss": 0.1035,
+      "step": 715
+    },
+    {
+      "epoch": 2.3008,
+      "grad_norm": 0.015480290167033672,
+      "learning_rate": 0.0001,
+      "loss": 0.0876,
+      "step": 720
+    },
+    {
+      "epoch": 2.3168,
+      "grad_norm": 0.018031220883131027,
+      "learning_rate": 0.0001,
+      "loss": 0.087,
+      "step": 725
+    },
+    {
+      "epoch": 2.3327999999999998,
+      "grad_norm": 0.017226027324795723,
+      "learning_rate": 0.0001,
+      "loss": 0.094,
+      "step": 730
+    },
+    {
+      "epoch": 2.3487999999999998,
+      "grad_norm": 0.016831068322062492,
+      "learning_rate": 0.0001,
+      "loss": 0.0921,
+      "step": 735
+    },
+    {
+      "epoch": 2.3648,
+      "grad_norm": 0.01735970750451088,
+      "learning_rate": 0.0001,
+      "loss": 0.0881,
+      "step": 740
+    },
+    {
+      "epoch": 2.3808,
+      "grad_norm": 0.019527316093444824,
+      "learning_rate": 0.0001,
+      "loss": 0.0951,
+      "step": 745
+    },
+    {
+      "epoch": 2.3968,
+      "grad_norm": 0.016558416187763214,
+      "learning_rate": 0.0001,
+      "loss": 0.0946,
+      "step": 750
+    },
+    {
+      "epoch": 2.4128,
+      "grad_norm": 0.015810256823897362,
+      "learning_rate": 0.0001,
+      "loss": 0.0807,
+      "step": 755
+    },
+    {
+      "epoch": 2.4288,
+      "grad_norm": 0.018082542344927788,
+      "learning_rate": 0.0001,
+      "loss": 0.0812,
+      "step": 760
+    },
+    {
+      "epoch": 2.4448,
+      "grad_norm": 0.018662385642528534,
+      "learning_rate": 0.0001,
+      "loss": 0.0784,
+      "step": 765
+    },
+    {
+      "epoch": 2.4608,
+      "grad_norm": 0.01677830144762993,
+      "learning_rate": 0.0001,
+      "loss": 0.0837,
+      "step": 770
+    },
+    {
+      "epoch": 2.4768,
+      "grad_norm": 0.01804305613040924,
+      "learning_rate": 0.0001,
+      "loss": 0.0794,
+      "step": 775
+    },
+    {
+      "epoch": 2.4928,
+      "grad_norm": 0.020595306530594826,
+      "learning_rate": 0.0001,
+      "loss": 0.0971,
+      "step": 780
+    },
+    {
+      "epoch": 2.5088,
+      "grad_norm": 0.012743465602397919,
+      "learning_rate": 0.0001,
+      "loss": 0.0795,
+      "step": 785
+    },
+    {
+      "epoch": 2.5248,
+      "grad_norm": 0.02038273587822914,
+      "learning_rate": 0.0001,
+      "loss": 0.0814,
+      "step": 790
+    },
+    {
+      "epoch": 2.5408,
+      "grad_norm": 0.018038209527730942,
+      "learning_rate": 0.0001,
+      "loss": 0.0963,
+      "step": 795
+    },
+    {
+      "epoch": 2.5568,
+      "grad_norm": 0.018109917640686035,
+      "learning_rate": 0.0001,
+      "loss": 0.0869,
+      "step": 800
+    },
+    {
+      "epoch": 2.5728,
+      "grad_norm": 0.02017812617123127,
+      "learning_rate": 0.0001,
+      "loss": 0.0973,
+      "step": 805
+    },
+    {
+      "epoch": 2.5888,
+      "grad_norm": 0.015534596517682076,
+      "learning_rate": 0.0001,
+      "loss": 0.0794,
+      "step": 810
+    },
+    {
+      "epoch": 2.6048,
+      "grad_norm": 0.01944139413535595,
+      "learning_rate": 0.0001,
+      "loss": 0.0912,
+      "step": 815
+    },
+    {
+      "epoch": 2.6208,
+      "grad_norm": 0.01757260598242283,
+      "learning_rate": 0.0001,
+      "loss": 0.0885,
+      "step": 820
+    },
+    {
+      "epoch": 2.6368,
+      "grad_norm": 0.019334355369210243,
+      "learning_rate": 0.0001,
+      "loss": 0.0936,
+      "step": 825
+    },
+    {
+      "epoch": 2.6528,
+      "grad_norm": 0.019028466194868088,
+      "learning_rate": 0.0001,
+      "loss": 0.0928,
+      "step": 830
+    },
+    {
+      "epoch": 2.6688,
+      "grad_norm": 0.017533207312226295,
+      "learning_rate": 0.0001,
+      "loss": 0.0796,
+      "step": 835
+    },
+    {
+      "epoch": 2.6848,
+      "grad_norm": 0.020642371848225594,
+      "learning_rate": 0.0001,
+      "loss": 0.0862,
+      "step": 840
+    },
+    {
+      "epoch": 2.7008,
+      "grad_norm": 0.016191577538847923,
+      "learning_rate": 0.0001,
+      "loss": 0.0851,
+      "step": 845
+    },
+    {
+      "epoch": 2.7168,
+      "grad_norm": 0.019956272095441818,
+      "learning_rate": 0.0001,
+      "loss": 0.094,
+      "step": 850
+    },
+    {
+      "epoch": 2.7328,
+      "grad_norm": 0.023722851648926735,
+      "learning_rate": 0.0001,
+      "loss": 0.0848,
+      "step": 855
+    },
+    {
+      "epoch": 2.7488,
+      "grad_norm": 0.01931099221110344,
+      "learning_rate": 0.0001,
+      "loss": 0.095,
+      "step": 860
+    },
+    {
+      "epoch": 2.7648,
+      "grad_norm": 0.015546354465186596,
+      "learning_rate": 0.0001,
+      "loss": 0.0732,
+      "step": 865
+    },
+    {
+      "epoch": 2.7808,
+      "grad_norm": 0.019629893824458122,
+      "learning_rate": 0.0001,
+      "loss": 0.0812,
+      "step": 870
+    },
+    {
+      "epoch": 2.7968,
+      "grad_norm": 0.020365415140986443,
+      "learning_rate": 0.0001,
+      "loss": 0.0951,
+      "step": 875
+    },
+    {
+      "epoch": 2.8128,
+      "grad_norm": 0.015614895150065422,
+      "learning_rate": 0.0001,
+      "loss": 0.0746,
+      "step": 880
+    },
+    {
+      "epoch": 2.8288,
+      "grad_norm": 0.014506829902529716,
+      "learning_rate": 0.0001,
+      "loss": 0.0968,
+      "step": 885
+    },
+    {
+      "epoch": 2.8448,
+      "grad_norm": 0.017022233456373215,
+      "learning_rate": 0.0001,
+      "loss": 0.0816,
+      "step": 890
+    },
+    {
+      "epoch": 2.8608000000000002,
+      "grad_norm": 0.02071644738316536,
+      "learning_rate": 0.0001,
+      "loss": 0.0932,
+      "step": 895
+    },
+    {
+      "epoch": 2.8768000000000002,
+      "grad_norm": 0.019489502534270287,
+      "learning_rate": 0.0001,
+      "loss": 0.0883,
+      "step": 900
+    }
+  ],
+  "logging_steps": 5,
+  "max_steps": 939,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 3,
+  "save_steps": 900,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 4.632069443587932e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}

sft_train_Qwen2.5-1.5B/output/checkpoint-900/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:503b7d3d1ad6b65eaf54005a8325b401779032ed351cb8d0522268c0b562601d
+size 8081

sft_train_Qwen2.5-1.5B/output/checkpoint-900/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 ZERO_STAGE not 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("Extracting fp32 weights")
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag)
+
+    logger.info("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)