Add files using upload-large-folder tool

.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

config.json

@@ -0,0 +1,30 @@
+{
+  "_name_or_path": "deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B",
+  "architectures": [
+    "Qwen2ForCausalLM"
+  ],
+  "attention_dropout": 0.0,
+  "bos_token_id": 151643,
+  "eos_token_id": 151643,
+  "hidden_act": "silu",
+  "hidden_size": 1536,
+  "initializer_range": 0.02,
+  "intermediate_size": 8960,
+  "max_position_embeddings": 131072,
+  "max_window_layers": 21,
+  "model_type": "qwen2",
+  "num_attention_heads": 12,
+  "num_hidden_layers": 28,
+  "num_key_value_heads": 2,
+  "rms_norm_eps": 1e-06,
+  "rope_scaling": null,
+  "rope_theta": 10000,
+  "sliding_window": 4096,
+  "tie_word_embeddings": false,
+  "torch_dtype": "bfloat16",
+  "transformers_version": "4.49.0",
+  "use_cache": false,
+  "use_mrope": false,
+  "use_sliding_window": false,
+  "vocab_size": 151936
+}

generation_config.json

@@ -0,0 +1,9 @@
+{
+  "_from_model_config": true,
+  "bos_token_id": 151646,
+  "do_sample": true,
+  "eos_token_id": 151643,
+  "temperature": 0.6,
+  "top_p": 0.95,
+  "transformers_version": "4.49.0"
+}

latest

@@ -0,0 +1 @@
+global_step50

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dd7794f885763ea54610093c6434279c892012211173a5a43c7b3eed67d81831
+size 3554214752

rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5fdf5dab088e40643817c4729a171b87d19f30b4693074f67498e787ec3ebf99
+size 14960

rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:93534a5e1a72750a0a250a8b7f453be2d25abf03a7ddd59220f8f95670323176
+size 14960

rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:15e7056d80551f130219ac8e710c1cdaf71949805baf2be9ab493dd3954bd3dc
+size 14960

scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2b1eb06383f9abe538d3bacb9e95a3d1b19c21cae1dfb62d8567723fcc11b74a
+size 1064

special_tokens_map.json

@@ -0,0 +1,23 @@
+{
+  "bos_token": {
+    "content": "<｜begin▁of▁sentence｜>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "eos_token": {
+    "content": "<｜end▁of▁sentence｜>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": {
+    "content": "<｜end▁of▁sentence｜>",
+    "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:a4256422650d141f228fe954acee98679da412984c29a569877eefd3af69315a
+size 11422959

tokenizer_config.json

@@ -0,0 +1,195 @@
+{
+  "add_bos_token": true,
+  "add_eos_token": false,
+  "add_prefix_space": null,
+  "added_tokens_decoder": {
+    "151643": {
+      "content": "<｜end▁of▁sentence｜>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151644": {
+      "content": "<｜User｜>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151645": {
+      "content": "<｜Assistant｜>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151646": {
+      "content": "<｜begin▁of▁sentence｜>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151647": {
+      "content": "<|EOT|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151648": {
+      "content": "<think>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151649": {
+      "content": "</think>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "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
+    }
+  },
+  "bos_token": "<｜begin▁of▁sentence｜>",
+  "chat_template": "{% if not add_generation_prompt is defined %}{% set add_generation_prompt = false %}{% endif %}{% set ns = namespace(is_first=false, is_tool=false, is_output_first=true, system_prompt='') %}{%- for message in messages %}{%- if message['role'] == 'system' %}{% set ns.system_prompt = message['content'] %}{%- endif %}{%- endfor %}{{bos_token}}{{ns.system_prompt}}{%- for message in messages %}{%- if message['role'] == 'user' %}{%- set ns.is_tool = false -%}{{'<｜User｜>' + message['content']}}{%- endif %}{%- if message['role'] == 'assistant' and message['content'] is none %}{%- set ns.is_tool = false -%}{%- for tool in message['tool_calls']%}{%- if not ns.is_first %}{{'<｜Assistant｜><｜tool▁calls▁begin｜><｜tool▁call▁begin��>' + tool['type'] + '<｜tool▁sep｜>' + tool['function']['name'] + '\\n' + '```json' + '\\n' + tool['function']['arguments'] + '\\n' + '```' + '<｜tool▁call▁end｜>'}}{%- set ns.is_first = true -%}{%- else %}{{'\\n' + '<｜tool▁call▁begin｜>' + tool['type'] + '<｜tool▁sep｜>' + tool['function']['name'] + '\\n' + '```json' + '\\n' + tool['function']['arguments'] + '\\n' + '```' + '<｜tool▁call▁end｜>'}}{{'<｜tool▁calls▁end｜><｜end▁of▁sentence｜>'}}{%- endif %}{%- endfor %}{%- endif %}{%- if message['role'] == 'assistant' and message['content'] is not none %}{%- if ns.is_tool %}{{'<｜tool▁outputs▁end｜>' + message['content'] + '<｜end▁of▁sentence｜>'}}{%- set ns.is_tool = false -%}{%- else %}{% set content = message['content'] %}{% if '</think>' in content %}{% set content = content.split('</think>')[-1] %}{% endif %}{{'<｜Assistant｜>' + content + '<｜end▁of▁sentence｜>'}}{%- endif %}{%- endif %}{%- if message['role'] == 'tool' %}{%- set ns.is_tool = true -%}{%- if ns.is_output_first %}{{'<｜tool▁outputs▁begin｜><｜tool▁output▁begin｜>' + message['content'] + '<｜tool▁output▁end｜>'}}{%- set ns.is_output_first = false %}{%- else %}{{'\\n<｜tool▁output▁begin｜>' + message['content'] + '<｜tool▁output▁end｜>'}}{%- endif %}{%- endif %}{%- endfor -%}{% if ns.is_tool %}{{'<｜tool▁outputs▁end｜>'}}{% endif %}{% if add_generation_prompt and not ns.is_tool %}{{'<｜Assistant｜><think>\\n'}}{% endif %}",
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<｜end▁of▁sentence｜>",
+  "extra_special_tokens": {},
+  "legacy": true,
+  "model_max_length": 16384,
+  "pad_token": "<｜end▁of▁sentence｜>",
+  "sp_model_kwargs": {},
+  "tokenizer_class": "LlamaTokenizerFast",
+  "unk_token": null,
+  "use_default_system_prompt": false
+}

trainer_state.json

@@ -0,0 +1,733 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.0856898029134533,
+  "eval_steps": 500,
+  "global_step": 50,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 2862.5695190429688,
+      "epoch": 0.001713796058269066,
+      "grad_norm": 0.16917119920253754,
+      "kl": 0.0,
+      "learning_rate": 2e-08,
+      "loss": 0.0467,
+      "reward": 0.12026740610599518,
+      "reward_std": 0.47210293635725975,
+      "rewards/cosine_scaled_reward": -0.1343107339926064,
+      "rewards/format_reward": 0.3888888917863369,
+      "step": 1
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 2739.5,
+      "epoch": 0.003427592116538132,
+      "grad_norm": 0.18499086797237396,
+      "kl": 0.0,
+      "learning_rate": 4e-08,
+      "loss": 0.0391,
+      "reward": -0.05314926430583,
+      "reward_std": 0.36226021870970726,
+      "rewards/cosine_scaled_reward": -0.21407463820651174,
+      "rewards/format_reward": 0.3750000111758709,
+      "step": 2
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 2830.77783203125,
+      "epoch": 0.005141388174807198,
+      "grad_norm": 0.18050141632556915,
+      "kl": 3.692507743835449e-05,
+      "learning_rate": 6e-08,
+      "loss": -0.0099,
+      "reward": 0.09248740691691637,
+      "reward_std": 0.6634868904948235,
+      "rewards/cosine_scaled_reward": -0.15514518693089485,
+      "rewards/format_reward": 0.4027777798473835,
+      "step": 3
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 2805.625,
+      "epoch": 0.006855184233076264,
+      "grad_norm": 0.16725227236747742,
+      "kl": 3.123283386230469e-05,
+      "learning_rate": 8e-08,
+      "loss": 0.0535,
+      "reward": 0.29482341557741165,
+      "reward_std": 0.7357868105173111,
+      "rewards/cosine_scaled_reward": -0.08869940415024757,
+      "rewards/format_reward": 0.4722222313284874,
+      "step": 4
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 2503.4305419921875,
+      "epoch": 0.00856898029134533,
+      "grad_norm": 0.1689172238111496,
+      "kl": 2.6106834411621094e-05,
+      "learning_rate": 1e-07,
+      "loss": 0.0737,
+      "reward": 0.5638985820114613,
+      "reward_std": 0.5826280601322651,
+      "rewards/cosine_scaled_reward": -0.009717360138893127,
+      "rewards/format_reward": 0.5833333283662796,
+      "step": 5
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 3151.4305419921875,
+      "epoch": 0.010282776349614395,
+      "grad_norm": 0.20181749761104584,
+      "kl": 4.225969314575195e-05,
+      "learning_rate": 1.2e-07,
+      "loss": 0.0636,
+      "reward": -0.12377724796533585,
+      "reward_std": 0.5666400231420994,
+      "rewards/cosine_scaled_reward": -0.19383307732641697,
+      "rewards/format_reward": 0.2638888889923692,
+      "step": 6
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 2207.652801513672,
+      "epoch": 0.011996572407883462,
+      "grad_norm": 0.26183807849884033,
+      "kl": 3.224611282348633e-05,
+      "learning_rate": 1.4e-07,
+      "loss": 0.0306,
+      "reward": 0.588454547803849,
+      "reward_std": 0.710302334278822,
+      "rewards/cosine_scaled_reward": -0.03910606319550425,
+      "rewards/format_reward": 0.6666666716337204,
+      "step": 7
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 2880.6388549804688,
+      "epoch": 0.013710368466152529,
+      "grad_norm": 0.1484246701002121,
+      "kl": 2.5212764739990234e-05,
+      "learning_rate": 1.6e-07,
+      "loss": 0.0465,
+      "reward": 0.5758275026455522,
+      "reward_std": 0.8158164247870445,
+      "rewards/cosine_scaled_reward": 0.01708040665835142,
+      "rewards/format_reward": 0.5416666716337204,
+      "step": 8
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 2553.8611450195312,
+      "epoch": 0.015424164524421594,
+      "grad_norm": 0.29594743251800537,
+      "kl": 2.3365020751953125e-05,
+      "learning_rate": 1.8e-07,
+      "loss": 0.0821,
+      "reward": 0.235587022267282,
+      "reward_std": 0.7132463157176971,
+      "rewards/cosine_scaled_reward": -0.11137315654195845,
+      "rewards/format_reward": 0.45833334419876337,
+      "step": 9
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 2864.3333740234375,
+      "epoch": 0.01713796058269066,
+      "grad_norm": 0.18843963742256165,
+      "kl": 3.55839729309082e-05,
+      "learning_rate": 2e-07,
+      "loss": 0.0595,
+      "reward": 0.3688787994906306,
+      "reward_std": 0.9138147085905075,
+      "rewards/cosine_scaled_reward": -0.016949500888586044,
+      "rewards/format_reward": 0.4027777761220932,
+      "step": 10
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 2650.374969482422,
+      "epoch": 0.018851756640959727,
+      "grad_norm": 0.35286611318588257,
+      "kl": 3.769993782043457e-05,
+      "learning_rate": 2.1999999999999998e-07,
+      "loss": 0.1724,
+      "reward": 0.1046360395848751,
+      "reward_std": 0.672421894967556,
+      "rewards/cosine_scaled_reward": -0.12129308935254812,
+      "rewards/format_reward": 0.34722223225980997,
+      "step": 11
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 2968.8750610351562,
+      "epoch": 0.02056555269922879,
+      "grad_norm": 0.33867669105529785,
+      "kl": 3.49581241607666e-05,
+      "learning_rate": 2.4e-07,
+      "loss": 0.1097,
+      "reward": -0.1256260685622692,
+      "reward_std": 0.5329847931861877,
+      "rewards/cosine_scaled_reward": -0.22253524931147695,
+      "rewards/format_reward": 0.3194444477558136,
+      "step": 12
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 2797.5555419921875,
+      "epoch": 0.022279348757497857,
+      "grad_norm": 0.18638171255588531,
+      "kl": 3.197789192199707e-05,
+      "learning_rate": 2.6e-07,
+      "loss": -0.0238,
+      "reward": 0.3018228765577078,
+      "reward_std": 0.7313368320465088,
+      "rewards/cosine_scaled_reward": -0.06436635553836823,
+      "rewards/format_reward": 0.43055556155741215,
+      "step": 13
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 2933.9444580078125,
+      "epoch": 0.023993144815766924,
+      "grad_norm": 0.24482618272304535,
+      "kl": 3.820657730102539e-05,
+      "learning_rate": 2.8e-07,
+      "loss": 0.0235,
+      "reward": 0.058196403086185455,
+      "reward_std": 0.5374786257743835,
+      "rewards/cosine_scaled_reward": -0.12367959041148424,
+      "rewards/format_reward": 0.3055555634200573,
+      "step": 14
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 3445.888916015625,
+      "epoch": 0.02570694087403599,
+      "grad_norm": 0.12928012013435364,
+      "kl": 4.57763671875e-05,
+      "learning_rate": 3e-07,
+      "loss": 0.036,
+      "reward": -0.03555959090590477,
+      "reward_std": 0.8275671154260635,
+      "rewards/cosine_scaled_reward": -0.1358353467658162,
+      "rewards/format_reward": 0.23611111752688885,
+      "step": 15
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 2473.013916015625,
+      "epoch": 0.027420736932305057,
+      "grad_norm": 0.18912632763385773,
+      "kl": 3.026425838470459e-05,
+      "learning_rate": 3.2e-07,
+      "loss": 0.0402,
+      "reward": 0.7541989013552666,
+      "reward_std": 0.7829795032739639,
+      "rewards/cosine_scaled_reward": 0.07848833221942186,
+      "rewards/format_reward": 0.5972222313284874,
+      "step": 16
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 2747.5694580078125,
+      "epoch": 0.02913453299057412,
+      "grad_norm": 0.28827619552612305,
+      "kl": 3.106892108917236e-05,
+      "learning_rate": 3.4000000000000003e-07,
+      "loss": 0.009,
+      "reward": 0.4498763456940651,
+      "reward_std": 0.5080211237072945,
+      "rewards/cosine_scaled_reward": 0.03743817750364542,
+      "rewards/format_reward": 0.37500000186264515,
+      "step": 17
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 3216.0694580078125,
+      "epoch": 0.030848329048843187,
+      "grad_norm": 0.13800697028636932,
+      "kl": 3.427267074584961e-05,
+      "learning_rate": 3.6e-07,
+      "loss": -0.0062,
+      "reward": -0.10277504308032803,
+      "reward_std": 0.5791455134749413,
+      "rewards/cosine_scaled_reward": -0.17638752236962318,
+      "rewards/format_reward": 0.2500000074505806,
+      "step": 18
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 3065.2916870117188,
+      "epoch": 0.032562125107112254,
+      "grad_norm": 0.1491522639989853,
+      "kl": 3.0159950256347656e-05,
+      "learning_rate": 3.7999999999999996e-07,
+      "loss": 0.0265,
+      "reward": 0.019092530012130737,
+      "reward_std": 0.7753017023205757,
+      "rewards/cosine_scaled_reward": -0.14323152415454388,
+      "rewards/format_reward": 0.3055555606260896,
+      "step": 19
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 3379.2777709960938,
+      "epoch": 0.03427592116538132,
+      "grad_norm": 0.2114741951227188,
+      "kl": 2.86102294921875e-05,
+      "learning_rate": 4e-07,
+      "loss": 0.047,
+      "reward": -0.053781009279191494,
+      "reward_std": 0.6423763372004032,
+      "rewards/cosine_scaled_reward": -0.13800161890685558,
+      "rewards/format_reward": 0.22222222946584225,
+      "step": 20
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 2875.291717529297,
+      "epoch": 0.03598971722365039,
+      "grad_norm": 0.2242029458284378,
+      "kl": 4.082918167114258e-05,
+      "learning_rate": 4.1999999999999995e-07,
+      "loss": 0.0932,
+      "reward": 0.013707960490137339,
+      "reward_std": 0.6616120561957359,
+      "rewards/cosine_scaled_reward": -0.1528682578355074,
+      "rewards/format_reward": 0.3194444552063942,
+      "step": 21
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 3119.0833129882812,
+      "epoch": 0.037703513281919454,
+      "grad_norm": 0.16363810002803802,
+      "kl": 2.7298927307128906e-05,
+      "learning_rate": 4.3999999999999997e-07,
+      "loss": 0.0862,
+      "reward": 0.10138127207756042,
+      "reward_std": 0.662372387945652,
+      "rewards/cosine_scaled_reward": -0.11597602907568216,
+      "rewards/format_reward": 0.3333333358168602,
+      "step": 22
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 3114.138916015625,
+      "epoch": 0.03941730934018852,
+      "grad_norm": 0.14078961312770844,
+      "kl": 2.0310282707214355e-05,
+      "learning_rate": 4.6e-07,
+      "loss": 0.0087,
+      "reward": -0.18106067180633545,
+      "reward_std": 0.5297312885522842,
+      "rewards/cosine_scaled_reward": -0.22247478552162647,
+      "rewards/format_reward": 0.2638888917863369,
+      "step": 23
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 2742.736083984375,
+      "epoch": 0.04113110539845758,
+      "grad_norm": 0.17268158495426178,
+      "kl": 2.4259090423583984e-05,
+      "learning_rate": 4.8e-07,
+      "loss": 0.0427,
+      "reward": 0.21900439262390137,
+      "reward_std": 0.4178895503282547,
+      "rewards/cosine_scaled_reward": -0.07105334987863898,
+      "rewards/format_reward": 0.361111119389534,
+      "step": 24
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 2691.263916015625,
+      "epoch": 0.04284490145672665,
+      "grad_norm": 0.31574270129203796,
+      "kl": 3.209710121154785e-05,
+      "learning_rate": 5e-07,
+      "loss": 0.055,
+      "reward": 0.1675301417708397,
+      "reward_std": 0.8015742003917694,
+      "rewards/cosine_scaled_reward": -0.1454015839844942,
+      "rewards/format_reward": 0.4583333395421505,
+      "step": 25
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 3124.7361450195312,
+      "epoch": 0.044558697514995714,
+      "grad_norm": 0.18195514380931854,
+      "kl": 3.510713577270508e-05,
+      "learning_rate": 5.2e-07,
+      "loss": 0.1146,
+      "reward": 0.12784775905311108,
+      "reward_std": 0.7148727774620056,
+      "rewards/cosine_scaled_reward": -0.10968722542747855,
+      "rewards/format_reward": 0.3472222276031971,
+      "step": 26
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 3161.9306030273438,
+      "epoch": 0.04627249357326478,
+      "grad_norm": 0.1664053052663803,
+      "kl": 1.7210841178894043e-05,
+      "learning_rate": 5.4e-07,
+      "loss": 0.066,
+      "reward": 0.3565527945756912,
+      "reward_std": 0.7730190977454185,
+      "rewards/cosine_scaled_reward": 0.01160972099751234,
+      "rewards/format_reward": 0.3333333432674408,
+      "step": 27
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 2473.9444580078125,
+      "epoch": 0.04798628963153385,
+      "grad_norm": 0.3084779679775238,
+      "kl": 3.291666507720947e-05,
+      "learning_rate": 5.6e-07,
+      "loss": 0.0691,
+      "reward": 0.5224915966391563,
+      "reward_std": 0.6627854257822037,
+      "rewards/cosine_scaled_reward": -0.0026430808939039707,
+      "rewards/format_reward": 0.5277777910232544,
+      "step": 28
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 3082.3333740234375,
+      "epoch": 0.049700085689802914,
+      "grad_norm": 0.19051846861839294,
+      "kl": 3.807246685028076e-05,
+      "learning_rate": 5.8e-07,
+      "loss": 0.0483,
+      "reward": 0.3683222606778145,
+      "reward_std": 1.0155043303966522,
+      "rewards/cosine_scaled_reward": -0.024172214791178703,
+      "rewards/format_reward": 0.41666666977107525,
+      "step": 29
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 2674.8750610351562,
+      "epoch": 0.05141388174807198,
+      "grad_norm": 0.2294771671295166,
+      "kl": 3.553926944732666e-05,
+      "learning_rate": 6e-07,
+      "loss": 0.094,
+      "reward": 0.3091086260974407,
+      "reward_std": 0.7602805271744728,
+      "rewards/cosine_scaled_reward": -0.10239012539386749,
+      "rewards/format_reward": 0.5138888880610466,
+      "step": 30
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 2638.4583129882812,
+      "epoch": 0.05312767780634105,
+      "grad_norm": 0.1822758913040161,
+      "kl": 2.135336399078369e-05,
+      "learning_rate": 6.2e-07,
+      "loss": 0.0591,
+      "reward": 0.41959726088680327,
+      "reward_std": 0.5659373477101326,
+      "rewards/cosine_scaled_reward": 0.0014653010293841362,
+      "rewards/format_reward": 0.4166666716337204,
+      "step": 31
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 2679.65283203125,
+      "epoch": 0.054841473864610114,
+      "grad_norm": 0.26114267110824585,
+      "kl": 2.42539681494236e-05,
+      "learning_rate": 6.4e-07,
+      "loss": 0.1342,
+      "reward": 0.39799272175878286,
+      "reward_std": 0.8707453161478043,
+      "rewards/cosine_scaled_reward": -0.08572587324306369,
+      "rewards/format_reward": 0.5694444514811039,
+      "step": 32
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 2365.5556030273438,
+      "epoch": 0.056555269922879174,
+      "grad_norm": 0.2686753571033478,
+      "kl": 6.015598773956299e-05,
+      "learning_rate": 6.6e-07,
+      "loss": 0.096,
+      "reward": 0.6290882844477892,
+      "reward_std": 0.9406441152095795,
+      "rewards/cosine_scaled_reward": 0.05065522436052561,
+      "rewards/format_reward": 0.5277777798473835,
+      "step": 33
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 2769.9722900390625,
+      "epoch": 0.05826906598114824,
+      "grad_norm": 0.26905301213264465,
+      "kl": 9.942054748535156e-05,
+      "learning_rate": 6.800000000000001e-07,
+      "loss": 0.1084,
+      "reward": 0.12581600830890238,
+      "reward_std": 0.6685186251997948,
+      "rewards/cosine_scaled_reward": -0.13848087890073657,
+      "rewards/format_reward": 0.4027777835726738,
+      "step": 34
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 3135.5694580078125,
+      "epoch": 0.05998286203941731,
+      "grad_norm": 0.1376740038394928,
+      "kl": 8.040666580200195e-05,
+      "learning_rate": 7e-07,
+      "loss": 0.0542,
+      "reward": -0.20180033892393112,
+      "reward_std": 0.46953874826431274,
+      "rewards/cosine_scaled_reward": -0.2328446013852954,
+      "rewards/format_reward": 0.2638888927176595,
+      "step": 35
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 2743.3055725097656,
+      "epoch": 0.061696658097686374,
+      "grad_norm": 0.19778397679328918,
+      "kl": 0.00010716915130615234,
+      "learning_rate": 7.2e-07,
+      "loss": 0.0158,
+      "reward": 0.12446275702677667,
+      "reward_std": 0.5396032929420471,
+      "rewards/cosine_scaled_reward": -0.1391575187444687,
+      "rewards/format_reward": 0.40277778543531895,
+      "step": 36
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 3021.9861450195312,
+      "epoch": 0.06341045415595545,
+      "grad_norm": 0.16646842658519745,
+      "kl": 4.979968070983887e-05,
+      "learning_rate": 7.4e-07,
+      "loss": 0.0452,
+      "reward": -0.27577453199774027,
+      "reward_std": 0.3186095915734768,
+      "rewards/cosine_scaled_reward": -0.26288727298378944,
+      "rewards/format_reward": 0.25000000838190317,
+      "step": 37
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 2999.0694580078125,
+      "epoch": 0.06512425021422451,
+      "grad_norm": 0.25001204013824463,
+      "kl": 0.00015616416931152344,
+      "learning_rate": 7.599999999999999e-07,
+      "loss": 0.114,
+      "reward": 0.10032026097178459,
+      "reward_std": 0.7015992030501366,
+      "rewards/cosine_scaled_reward": -0.10261764749884605,
+      "rewards/format_reward": 0.3055555634200573,
+      "step": 38
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 3170.4027709960938,
+      "epoch": 0.06683804627249357,
+      "grad_norm": 0.13187235593795776,
+      "kl": 0.00024265050888061523,
+      "learning_rate": 7.799999999999999e-07,
+      "loss": 0.023,
+      "reward": 0.032449623104184866,
+      "reward_std": 0.7292564660310745,
+      "rewards/cosine_scaled_reward": -0.12960851565003395,
+      "rewards/format_reward": 0.2916666679084301,
+      "step": 39
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 2721.5694580078125,
+      "epoch": 0.06855184233076264,
+      "grad_norm": 0.24242877960205078,
+      "kl": 0.00039273500442504883,
+      "learning_rate": 8e-07,
+      "loss": 0.1056,
+      "reward": 0.2101573906838894,
+      "reward_std": 0.5202245805412531,
+      "rewards/cosine_scaled_reward": -0.07547687107580714,
+      "rewards/format_reward": 0.3611111156642437,
+      "step": 40
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 2850.65283203125,
+      "epoch": 0.0702656383890317,
+      "grad_norm": 0.1599811613559723,
+      "kl": 0.00010573863983154297,
+      "learning_rate": 8.199999999999999e-07,
+      "loss": 0.0787,
+      "reward": 0.3573916293680668,
+      "reward_std": 0.7150286287069321,
+      "rewards/cosine_scaled_reward": -0.022693071980029345,
+      "rewards/format_reward": 0.4027777835726738,
+      "step": 41
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 2629.402801513672,
+      "epoch": 0.07197943444730077,
+      "grad_norm": 0.2574199140071869,
+      "kl": 0.0005946159362792969,
+      "learning_rate": 8.399999999999999e-07,
+      "loss": 0.0996,
+      "reward": 0.3184794522821903,
+      "reward_std": 0.7374207675457001,
+      "rewards/cosine_scaled_reward": -0.0699269392844144,
+      "rewards/format_reward": 0.45833333022892475,
+      "step": 42
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 2797.75,
+      "epoch": 0.07369323050556983,
+      "grad_norm": 0.1763925850391388,
+      "kl": 0.00013440847396850586,
+      "learning_rate": 8.599999999999999e-07,
+      "loss": 0.081,
+      "reward": 0.460858479142189,
+      "reward_std": 0.6768517717719078,
+      "rewards/cosine_scaled_reward": 0.015151458326727152,
+      "rewards/format_reward": 0.4305555671453476,
+      "step": 43
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 2764.15283203125,
+      "epoch": 0.07540702656383891,
+      "grad_norm": 0.14260780811309814,
+      "kl": 0.0010910332202911377,
+      "learning_rate": 8.799999999999999e-07,
+      "loss": -0.0101,
+      "reward": 0.24442699179053307,
+      "reward_std": 0.7257742322981358,
+      "rewards/cosine_scaled_reward": -0.14861984690651298,
+      "rewards/format_reward": 0.5416666716337204,
+      "step": 44
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 2988.3055419921875,
+      "epoch": 0.07712082262210797,
+      "grad_norm": 0.14807860553264618,
+      "kl": 0.0002259761095046997,
+      "learning_rate": 9e-07,
+      "loss": 0.0058,
+      "reward": 0.1327950283885002,
+      "reward_std": 0.6829644069075584,
+      "rewards/cosine_scaled_reward": -0.07943581556901336,
+      "rewards/format_reward": 0.2916666679084301,
+      "step": 45
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 2406.750030517578,
+      "epoch": 0.07883461868037704,
+      "grad_norm": 0.3411652743816376,
+      "kl": 0.0010082721710205078,
+      "learning_rate": 9.2e-07,
+      "loss": 0.0325,
+      "reward": 0.37563128490000963,
+      "reward_std": 0.5480553992092609,
+      "rewards/cosine_scaled_reward": -0.07607323676347733,
+      "rewards/format_reward": 0.5277777835726738,
+      "step": 46
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 2894.4306030273438,
+      "epoch": 0.0805484147386461,
+      "grad_norm": 0.1801064908504486,
+      "kl": 0.0010660886764526367,
+      "learning_rate": 9.399999999999999e-07,
+      "loss": 0.0585,
+      "reward": 0.2889724150300026,
+      "reward_std": 0.7576765939593315,
+      "rewards/cosine_scaled_reward": -0.08468045573681593,
+      "rewards/format_reward": 0.4583333358168602,
+      "step": 47
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 2787.6111450195312,
+      "epoch": 0.08226221079691516,
+      "grad_norm": 0.2200099229812622,
+      "kl": 0.0008892565965652466,
+      "learning_rate": 9.6e-07,
+      "loss": 0.1153,
+      "reward": 0.6353000029921532,
+      "reward_std": 0.7432012036442757,
+      "rewards/cosine_scaled_reward": 0.07459445751737803,
+      "rewards/format_reward": 0.4861111268401146,
+      "step": 48
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 2355.763946533203,
+      "epoch": 0.08397600685518423,
+      "grad_norm": 0.22268813848495483,
+      "kl": 0.002144336700439453,
+      "learning_rate": 9.8e-07,
+      "loss": 0.0718,
+      "reward": 0.5641769841313362,
+      "reward_std": 0.697506844997406,
+      "rewards/cosine_scaled_reward": -0.016522612422704697,
+      "rewards/format_reward": 0.5972222313284874,
+      "step": 49
+    },
+    {
+      "clip_ratio": 0.0,
+      "completion_length": 2625.7083129882812,
+      "epoch": 0.0856898029134533,
+      "grad_norm": 0.18417640030384064,
+      "kl": 0.0020694732666015625,
+      "learning_rate": 1e-06,
+      "loss": 0.0792,
+      "reward": 0.09065666515380144,
+      "reward_std": 0.6994658857584,
+      "rewards/cosine_scaled_reward": -0.16300500556826591,
+      "rewards/format_reward": 0.4166666753590107,
+      "step": 50
+    }
+  ],
+  "logging_steps": 1,
+  "max_steps": 500,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 1,
+  "save_steps": 25,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 0.0,
+  "train_batch_size": 6,
+  "trial_name": null,
+  "trial_params": null
+}

training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ef5e1578db57a0a30c2f11ec7d5ad1212c33cd4e69b490df6a3b1fe0a5dbd90f
+size 8248

zero_to_fp32.py

@@ -0,0 +1,674 @@
+#!/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 json
+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)
+
+        if BUFFER_NAMES not in state_dict:
+            raise ValueError(f"{file} is not a model state checkpoint")
+        buffer_names = state_dict[BUFFER_NAMES]
+        if debug:
+            print("Found buffers:", buffer_names)
+
+        # recover just the buffers while restoring them to fp32 if they were saved in fp16
+        buffers = {k: v.float() for k, v in state_dict["module"].items() if k in buffer_names}
+        param_shapes = state_dict[PARAM_SHAPES]
+
+        # collect parameters that are included in param_shapes
+        param_names = []
+        for s in param_shapes:
+            for name in s.keys():
+                param_names.append(name)
+
+        # update with frozen parameters
+        frozen_param_shapes = state_dict.get(FROZEN_PARAM_SHAPES, None)
+        if frozen_param_shapes is not None:
+            if debug:
+                print(f"Found frozen_param_shapes: {frozen_param_shapes}")
+            param_names += list(frozen_param_shapes.keys())
+
+        # handle shared params
+        shared_params = [[k, v] for k, v in state_dict["shared_params"].items()]
+
+        ds_version = state_dict.get(DS_VERSION, None)
+
+        frozen_param_fragments = state_dict.get(FROZEN_PARAM_FRAGMENTS, None)
+
+        z_model_state = zero_model_state(buffers=buffers,
+                                         param_shapes=param_shapes,
+                                         shared_params=shared_params,
+                                         ds_version=ds_version,
+                                         frozen_param_shapes=frozen_param_shapes,
+                                         frozen_param_fragments=frozen_param_fragments)
+        zero_model_states.append(z_model_state)
+
+    return zero_model_states
+
+
+def parse_optim_states(files, ds_checkpoint_dir):
+    total_files = len(files)
+    state_dicts = []
+    for f in files:
+        state_dict = torch.load(f, map_location=device)
+        # immediately discard the potentially huge 2 optimizer states as we only care for fp32 master weights
+        # and also handle the case where it was already removed by another helper script
+        state_dict["optimizer_state_dict"].pop("optimizer_state_dict", None)
+        state_dicts.append(state_dict)
+
+    if not ZERO_STAGE in state_dicts[0][OPTIMIZER_STATE_DICT]:
+        raise ValueError(f"{files[0]} is not a zero checkpoint")
+    zero_stage = state_dicts[0][OPTIMIZER_STATE_DICT][ZERO_STAGE]
+    world_size = state_dicts[0][OPTIMIZER_STATE_DICT][PARTITION_COUNT]
+
+    # For ZeRO-2 each param group can have different partition_count as data parallelism for expert
+    # parameters can be different from data parallelism for non-expert parameters. So we can just
+    # use the max of the partition_count to get the dp world_size.
+
+    if type(world_size) is list:
+        world_size = max(world_size)
+
+    if world_size != total_files:
+        raise ValueError(
+            f"Expected {world_size} of '*_optim_states.pt' under '{ds_checkpoint_dir}' but found {total_files} files. "
+            "Possibly due to an overwrite of an old checkpoint, or a checkpoint didn't get saved by one or more processes."
+        )
+
+    # the groups are named differently in each stage
+    if zero_stage <= 2:
+        fp32_groups_key = SINGLE_PARTITION_OF_FP32_GROUPS
+    elif zero_stage == 3:
+        fp32_groups_key = FP32_FLAT_GROUPS
+    else:
+        raise ValueError(f"unknown zero stage {zero_stage}")
+
+    if zero_stage <= 2:
+        fp32_flat_groups = [state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key] for i in range(len(state_dicts))]
+    elif zero_stage == 3:
+        # if there is more than one param group, there will be multiple flattened tensors - one
+        # flattened tensor per group - for simplicity merge them into a single tensor
+        #
+        # XXX: could make the script more memory efficient for when there are multiple groups - it
+        # will require matching the sub-lists of param_shapes for each param group flattened tensor
+
+        fp32_flat_groups = [
+            torch.cat(state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key], 0) for i in range(len(state_dicts))
+        ]
+
+    return zero_stage, world_size, fp32_flat_groups
+
+
+def _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters):
+    """
+    Returns fp32 state_dict reconstructed from ds checkpoint
+
+    Args:
+        - ``ds_checkpoint_dir``: path to the deepspeed checkpoint folder (where the optimizer files are)
+
+    """
+    print(f"Processing zero checkpoint '{ds_checkpoint_dir}'")
+
+    optim_files = get_optim_files(ds_checkpoint_dir)
+    zero_stage, world_size, fp32_flat_groups = parse_optim_states(optim_files, ds_checkpoint_dir)
+    print(f"Detected checkpoint of type zero stage {zero_stage}, world_size: {world_size}")
+
+    model_files = get_model_state_files(ds_checkpoint_dir)
+
+    zero_model_states = parse_model_states(model_files)
+    print(f'Parsing checkpoint created by deepspeed=={zero_model_states[0].ds_version}')
+
+    if zero_stage <= 2:
+        return _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                                          exclude_frozen_parameters)
+    elif zero_stage == 3:
+        return _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                                          exclude_frozen_parameters)
+
+
+def _zero2_merge_frozen_params(state_dict, zero_model_states):
+    if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
+        return
+
+    frozen_param_shapes = zero_model_states[0].frozen_param_shapes
+    frozen_param_fragments = zero_model_states[0].frozen_param_fragments
+
+    if debug:
+        num_elem = sum(s.numel() for s in frozen_param_shapes.values())
+        print(f'rank 0: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
+
+        wanted_params = len(frozen_param_shapes)
+        wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
+        avail_numel = sum([p.numel() for p in frozen_param_fragments.values()])
+        print(f'Frozen params: Have {avail_numel} numels to process.')
+        print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
+
+    total_params = 0
+    total_numel = 0
+    for name, shape in frozen_param_shapes.items():
+        total_params += 1
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+
+        state_dict[name] = frozen_param_fragments[name]
+
+        if debug:
+            print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
+
+    print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _has_callable(obj, fn):
+    attr = getattr(obj, fn, None)
+    return callable(attr)
+
+
+def _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
+    param_shapes = zero_model_states[0].param_shapes
+
+    # Reconstruction protocol:
+    #
+    # XXX: document this
+
+    if debug:
+        for i in range(world_size):
+            for j in range(len(fp32_flat_groups[0])):
+                print(f"{FP32_FLAT_GROUPS}[{i}][{j}].shape={fp32_flat_groups[i][j].shape}")
+
+    # XXX: memory usage doubles here (zero2)
+    num_param_groups = len(fp32_flat_groups[0])
+    merged_single_partition_of_fp32_groups = []
+    for i in range(num_param_groups):
+        merged_partitions = [sd[i] for sd in fp32_flat_groups]
+        full_single_fp32_vector = torch.cat(merged_partitions, 0)
+        merged_single_partition_of_fp32_groups.append(full_single_fp32_vector)
+    avail_numel = sum(
+        [full_single_fp32_vector.numel() for full_single_fp32_vector in merged_single_partition_of_fp32_groups])
+
+    if debug:
+        wanted_params = sum([len(shapes) for shapes in param_shapes])
+        wanted_numel = sum([sum(shape.numel() for shape in shapes.values()) for shapes in param_shapes])
+        # not asserting if there is a mismatch due to possible padding
+        print(f"Have {avail_numel} numels to process.")
+        print(f"Need {wanted_numel} numels in {wanted_params} params.")
+
+    # params
+    # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
+    # out-of-core computing solution
+    total_numel = 0
+    total_params = 0
+    for shapes, full_single_fp32_vector in zip(param_shapes, merged_single_partition_of_fp32_groups):
+        offset = 0
+        avail_numel = full_single_fp32_vector.numel()
+        for name, shape in shapes.items():
+
+            unpartitioned_numel = shape.numel() if _has_callable(shape, 'numel') else math.prod(shape)
+            total_numel += unpartitioned_numel
+            total_params += 1
+
+            if debug:
+                print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
+            state_dict[name] = full_single_fp32_vector.narrow(0, offset, unpartitioned_numel).view(shape)
+            offset += unpartitioned_numel
+
+        # Z2 started to align to 2*world_size to improve nccl performance. Therefore both offset and
+        # avail_numel can differ by anywhere between 0..2*world_size. Due to two unrelated complex
+        # paddings performed in the code it's almost impossible to predict the exact numbers w/o the
+        # live optimizer object, so we are checking that the numbers are within the right range
+        align_to = 2 * world_size
+
+        def zero2_align(x):
+            return align_to * math.ceil(x / align_to)
+
+        if debug:
+            print(f"original offset={offset}, avail_numel={avail_numel}")
+
+        offset = zero2_align(offset)
+        avail_numel = zero2_align(avail_numel)
+
+        if debug:
+            print(f"aligned  offset={offset}, avail_numel={avail_numel}")
+
+        # Sanity check
+        if offset != avail_numel:
+            raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
+
+    print(f"Reconstructed fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                               exclude_frozen_parameters):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    if not exclude_frozen_parameters:
+        _zero2_merge_frozen_params(state_dict, zero_model_states)
+
+    _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
+
+    # recover shared parameters
+    for pair in zero_model_states[0].shared_params:
+        if pair[1] in state_dict:
+            state_dict[pair[0]] = state_dict[pair[1]]
+
+    return state_dict
+
+
+def zero3_partitioned_param_info(unpartitioned_numel, world_size):
+    remainder = unpartitioned_numel % world_size
+    padding_numel = (world_size - remainder) if remainder else 0
+    partitioned_numel = math.ceil(unpartitioned_numel / world_size)
+    return partitioned_numel, padding_numel
+
+
+def _zero3_merge_frozen_params(state_dict, world_size, zero_model_states):
+    if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
+        return
+
+    if debug:
+        for i in range(world_size):
+            num_elem = sum(s.numel() for s in zero_model_states[i].frozen_param_fragments.values())
+            print(f'rank {i}: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
+
+        frozen_param_shapes = zero_model_states[0].frozen_param_shapes
+        wanted_params = len(frozen_param_shapes)
+        wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
+        avail_numel = sum([p.numel() for p in zero_model_states[0].frozen_param_fragments.values()]) * world_size
+        print(f'Frozen params: Have {avail_numel} numels to process.')
+        print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
+
+    total_params = 0
+    total_numel = 0
+    for name, shape in zero_model_states[0].frozen_param_shapes.items():
+        total_params += 1
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+
+        param_frags = tuple(model_state.frozen_param_fragments[name] for model_state in zero_model_states)
+        state_dict[name] = torch.cat(param_frags, 0).narrow(0, 0, unpartitioned_numel).view(shape)
+
+        partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
+
+        if debug:
+            print(
+                f"Frozen params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
+            )
+
+    print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
+    param_shapes = zero_model_states[0].param_shapes
+    avail_numel = fp32_flat_groups[0].numel() * world_size
+    # Reconstruction protocol: For zero3 we need to zip the partitions together at boundary of each
+    # param, re-consolidating each param, while dealing with padding if any
+
+    # merge list of dicts, preserving order
+    param_shapes = {k: v for d in param_shapes for k, v in d.items()}
+
+    if debug:
+        for i in range(world_size):
+            print(f"{FP32_FLAT_GROUPS}[{i}].shape={fp32_flat_groups[i].shape}")
+
+        wanted_params = len(param_shapes)
+        wanted_numel = sum(shape.numel() for shape in param_shapes.values())
+        # not asserting if there is a mismatch due to possible padding
+        avail_numel = fp32_flat_groups[0].numel() * world_size
+        print(f"Trainable params: Have {avail_numel} numels to process.")
+        print(f"Trainable params: Need {wanted_numel} numels in {wanted_params} params.")
+
+    # params
+    # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
+    # out-of-core computing solution
+    offset = 0
+    total_numel = 0
+    total_params = 0
+    for name, shape in 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}"
+            )
+
+        # XXX: memory usage doubles here
+        state_dict[name] = torch.cat(
+            tuple(fp32_flat_groups[i].narrow(0, offset, partitioned_numel) for i in range(world_size)),
+            0).narrow(0, 0, unpartitioned_numel).view(shape)
+        offset += partitioned_numel
+
+    offset *= world_size
+
+    # Sanity check
+    if offset != avail_numel:
+        raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
+
+    print(f"Reconstructed Trainable fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                               exclude_frozen_parameters):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    if not exclude_frozen_parameters:
+        _zero3_merge_frozen_params(state_dict, world_size, zero_model_states)
+
+    _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
+
+    # recover shared parameters
+    for pair in zero_model_states[0].shared_params:
+        if pair[1] in state_dict:
+            state_dict[pair[0]] = state_dict[pair[1]]
+
+    return state_dict
+
+
+def get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag=None, exclude_frozen_parameters=False):
+    """
+    Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated state_dict that can be loaded with
+    ``load_state_dict()`` and used for training without DeepSpeed or shared with others, for example
+    via a model hub.
+
+    Args:
+        - ``checkpoint_dir``: path to the desired checkpoint folder
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in 'latest' file. e.g., ``global_step14``
+        - ``exclude_frozen_parameters``: exclude frozen parameters
+
+    Returns:
+        - pytorch ``state_dict``
+
+    Note: this approach may not work if your application doesn't have sufficient free CPU memory and
+    you may need to use the offline approach using the ``zero_to_fp32.py`` script that is saved with
+    the checkpoint.
+
+    A typical usage might be ::
+
+        from deepspeed.utils.zero_to_fp32 import get_fp32_state_dict_from_zero_checkpoint
+        # do the training and checkpoint saving
+        state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir) # already on cpu
+        model = model.cpu() # move to cpu
+        model.load_state_dict(state_dict)
+        # submit to model hub or save the model to share with others
+
+    In this example the ``model`` will no longer be usable in the deepspeed context of the same
+    application. i.e. you will need to re-initialize the deepspeed engine, since
+    ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
+
+    If you want it all done for you, use ``load_state_dict_from_zero_checkpoint`` instead.
+
+    """
+    if tag is None:
+        latest_path = os.path.join(checkpoint_dir, 'latest')
+        if os.path.isfile(latest_path):
+            with open(latest_path, 'r') as fd:
+                tag = fd.read().strip()
+        else:
+            raise ValueError(f"Unable to find 'latest' file at {latest_path}")
+
+    ds_checkpoint_dir = os.path.join(checkpoint_dir, tag)
+
+    if not os.path.isdir(ds_checkpoint_dir):
+        raise FileNotFoundError(f"Directory '{ds_checkpoint_dir}' doesn't exist")
+
+    return _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters)
+
+
+def convert_zero_checkpoint_to_fp32_state_dict(checkpoint_dir,
+                                               output_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)
+
+    # 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")
+        state_dict_split = split_torch_state_dict_into_shards(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
+    filename_to_tensors = state_dict_split.filename_to_tensors.items()
+    for shard_file, tensors in tqdm(filename_to_tensors, desc="Saving checkpoint shards"):
+        shard = {tensor: state_dict[tensor].contiguous() for tensor in tensors}
+        output_path = os.path.join(output_dir, shard_file)
+        if safe_serialization:
+            save_file(shard, output_path, metadata={"format": "pt"})
+        else:
+            torch.save(shard, output_path)
+
+    # 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)