Upload folder using huggingface_hub

.gitattributes

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

added_tokens.json

@@ -0,0 +1,24 @@
+{
+  "</tool_call>": 151658,
+  "<tool_call>": 151657,
+  "<|box_end|>": 151649,
+  "<|box_start|>": 151648,
+  "<|endoftext|>": 151643,
+  "<|file_sep|>": 151664,
+  "<|fim_middle|>": 151660,
+  "<|fim_pad|>": 151662,
+  "<|fim_prefix|>": 151659,
+  "<|fim_suffix|>": 151661,
+  "<|im_end|>": 151645,
+  "<|im_start|>": 151644,
+  "<|image_pad|>": 151655,
+  "<|object_ref_end|>": 151647,
+  "<|object_ref_start|>": 151646,
+  "<|quad_end|>": 151651,
+  "<|quad_start|>": 151650,
+  "<|repo_name|>": 151663,
+  "<|video_pad|>": 151656,
+  "<|vision_end|>": 151653,
+  "<|vision_pad|>": 151654,
+  "<|vision_start|>": 151652
+}

config.json

@@ -0,0 +1,29 @@
+{
+  "architectures": [
+    "Qwen2ForCausalLM"
+  ],
+  "attention_dropout": 0.0,
+  "eos_token_id": 151645,
+  "hidden_act": "silu",
+  "hidden_size": 896,
+  "initializer_range": 0.02,
+  "intermediate_size": 4864,
+  "max_position_embeddings": 32768,
+  "max_window_layers": 21,
+  "model_type": "qwen2",
+  "num_attention_heads": 14,
+  "num_hidden_layers": 24,
+  "num_key_value_heads": 2,
+  "pad_token_id": 151654,
+  "rms_norm_eps": 1e-06,
+  "rope_scaling": null,
+  "rope_theta": 1000000.0,
+  "sliding_window": null,
+  "tie_word_embeddings": true,
+  "torch_dtype": "bfloat16",
+  "transformers_version": "4.51.3",
+  "unsloth_fixed": true,
+  "use_cache": false,
+  "use_sliding_window": false,
+  "vocab_size": 151936
+}

generation_config.json

@@ -0,0 +1,15 @@
+{
+  "bos_token_id": 151643,
+  "do_sample": true,
+  "eos_token_id": [
+    151645,
+    151643
+  ],
+  "max_length": 32768,
+  "pad_token_id": 151654,
+  "repetition_penalty": 1.1,
+  "temperature": 0.7,
+  "top_k": 20,
+  "top_p": 0.8,
+  "transformers_version": "4.51.3"
+}

global_step7000/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:c0852d1fbff16422886263593984c42479de79563c30e9657704a99170ba0aad
+size 1482103621

global_step7000/bf16_zero_pp_rank_1_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:903383d27744021bfe73816fc4e9492dc00d25ea9cf56adcf0344478976e2559
+size 1482103621

global_step7000/bf16_zero_pp_rank_2_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6f8ed521b5f01a15a4d97666f4041bc32c498777bce82b9b12b3ebdff8379a3b
+size 1482103621

global_step7000/bf16_zero_pp_rank_3_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:458060de5e82f671d2d76485c86ed2a2e63cd1c72b2d51d7d2c96c166b948231
+size 1482103621

global_step7000/zero_pp_rank_0_mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:754e68fc6e83a53cfbe61e3f715249a22778fd051166234714eed25a932631d4
+size 143059

global_step7000/zero_pp_rank_1_mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a31b45f44056f70d8be3b7c623112b13c96bf967dc215e8d91db97717afc262a
+size 142995

global_step7000/zero_pp_rank_2_mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dbd19837ad97e20a7ed9f7b640a347732c417b977e9c85b9702288a8ae175626
+size 142995

global_step7000/zero_pp_rank_3_mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:feb1eab726fdfda9f2cd5f9dffaae579f5b8ce6eabd287421d00e6e2c01d1805
+size 142995

latest

@@ -0,0 +1 @@
+global_step7000

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:da75b70e7761305896f20ed419b1ce89f132e4dc41c9dbfff095105888bf3488
+size 988097824

rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fac59f8685f9987cb2f2c0813fb063d93985dae8ca7b3e3348bbd3b9962e8e8e
+size 15429

rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b2c41082d86bdc69b01102c0ecce9462330f04cdaa135203d57af14ff58941c3
+size 15365

rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:845609fc37929e5a312464583328e4fd9732b30e9807636ea1f368cfe52393e2
+size 15429

rng_state_3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7c5052e84f057d012da47071b672bc71bcf008104c8d05356cd8037609551ecd
+size 15429

scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:412035590d0899313ef5cc734e77cc09a9b1b46dab5e47596b0bbf94ebdd86e4
+size 1465

special_tokens_map.json

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

tokenizer.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:64e71213db910f5cafa86d35091f37393dcc344b1bbc34091d1b3eed4cca01d5
+size 11422064

tokenizer_config.json

@@ -0,0 +1,212 @@
+{
+  "add_bos_token": false,
+  "add_prefix_space": false,
+  "added_tokens_decoder": {
+    "151643": {
+      "content": "<|endoftext|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151644": {
+      "content": "<|im_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151645": {
+      "content": "<|im_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151646": {
+      "content": "<|object_ref_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151647": {
+      "content": "<|object_ref_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151648": {
+      "content": "<|box_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151649": {
+      "content": "<|box_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151650": {
+      "content": "<|quad_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151651": {
+      "content": "<|quad_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151652": {
+      "content": "<|vision_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151653": {
+      "content": "<|vision_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151654": {
+      "content": "<|vision_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151655": {
+      "content": "<|image_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151656": {
+      "content": "<|video_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151657": {
+      "content": "<tool_call>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151658": {
+      "content": "</tool_call>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151659": {
+      "content": "<|fim_prefix|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151660": {
+      "content": "<|fim_middle|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151661": {
+      "content": "<|fim_suffix|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151662": {
+      "content": "<|fim_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151663": {
+      "content": "<|repo_name|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151664": {
+      "content": "<|file_sep|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    }
+  },
+  "additional_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "bos_token": null,
+  "chat_template": "{%- if tools %}\n    {{- '<|im_start|>system\\n' }}\n    {%- if messages[0]['role'] == 'system' %}\n        {{- messages[0]['content'] }}\n    {%- else %}\n        {{- 'You are Qwen, created by Alibaba Cloud. You are a helpful assistant.' }}\n    {%- endif %}\n    {{- \"\\n\\n# Tools\\n\\nYou may call one or more functions to assist with the user query.\\n\\nYou are provided with function signatures within <tools></tools> XML tags:\\n<tools>\" }}\n    {%- for tool in tools %}\n        {{- \"\\n\" }}\n        {{- tool | tojson }}\n    {%- endfor %}\n    {{- \"\\n</tools>\\n\\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\\n<tool_call>\\n{\\\"name\\\": <function-name>, \\\"arguments\\\": <args-json-object>}\\n</tool_call><|im_end|>\\n\" }}\n{%- else %}\n    {%- if messages[0]['role'] == 'system' %}\n        {{- '<|im_start|>system\\n' + messages[0]['content'] + '<|im_end|>\\n' }}\n    {%- else %}\n        {{- '<|im_start|>system\\nYou are Qwen, created by Alibaba Cloud. You are a helpful assistant.<|im_end|>\\n' }}\n    {%- endif %}\n{%- endif %}\n{%- for message in messages %}\n    {%- if (message.role == \"user\") or (message.role == \"system\" and not loop.first) or (message.role == \"assistant\" and not message.tool_calls) %}\n        {{- '<|im_start|>' + message.role + '\\n' + message.content + '<|im_end|>' + '\\n' }}\n    {%- elif message.role == \"assistant\" %}\n        {{- '<|im_start|>' + message.role }}\n        {%- if message.content %}\n            {{- '\\n' + message.content }}\n        {%- endif %}\n        {%- for tool_call in message.tool_calls %}\n            {%- if tool_call.function is defined %}\n                {%- set tool_call = tool_call.function %}\n            {%- endif %}\n            {{- '\\n<tool_call>\\n{\"name\": \"' }}\n            {{- tool_call.name }}\n            {{- '\", \"arguments\": ' }}\n            {{- tool_call.arguments | tojson }}\n            {{- '}\\n</tool_call>' }}\n        {%- endfor %}\n        {{- '<|im_end|>\\n' }}\n    {%- elif message.role == \"tool\" %}\n        {%- if (loop.index0 == 0) or (messages[loop.index0 - 1].role != \"tool\") %}\n            {{- '<|im_start|>user' }}\n        {%- endif %}\n        {{- '\\n<tool_response>\\n' }}\n        {{- message.content }}\n        {{- '\\n</tool_response>' }}\n        {%- if loop.last or (messages[loop.index0 + 1].role != \"tool\") %}\n            {{- '<|im_end|>\\n' }}\n        {%- endif %}\n    {%- endif %}\n{%- endfor %}\n{%- if add_generation_prompt %}\n    {{- '<|im_start|>assistant\\n' }}\n{%- endif %}\n",
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|im_end|>",
+  "errors": "replace",
+  "extra_special_tokens": {},
+  "max_length": null,
+  "model_max_length": 32768,
+  "pad_to_multiple_of": null,
+  "pad_token": "<|vision_pad|>",
+  "pad_token_type_id": 0,
+  "padding_side": "right",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}

trainer_state.json

@@ -0,0 +1,1994 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.1678065054211844,
+  "eval_steps": 500,
+  "global_step": 7000,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "clip_ratio": 0.00699951171875,
+      "completion_length": 1024.0,
+      "epoch": 0.008340283569641367,
+      "grad_norm": 14.120885699967895,
+      "kl": 2.8944024658203125,
+      "learning_rate": 1.9945510147345007e-06,
+      "loss": 0.0458,
+      "num_tokens": 608832.0,
+      "reward": -3.234580533504486,
+      "reward_std": 3.1962914752960203,
+      "rewards/generate_all_rewards": -3.234580533504486,
+      "step": 50
+    },
+    {
+      "clip_ratio": 0.005478515625,
+      "completion_length": 1024.0,
+      "epoch": 0.016680567139282735,
+      "grad_norm": 5.506146217443755,
+      "kl": 4.24171875,
+      "learning_rate": 1.9889908256880732e-06,
+      "loss": 0.0764,
+      "num_tokens": 1207352.0,
+      "reward": -1.5938849544525147,
+      "reward_std": 3.6290176486968995,
+      "rewards/generate_all_rewards": -1.5938849544525147,
+      "step": 100
+    },
+    {
+      "clip_ratio": 0.006314697265625,
+      "completion_length": 1024.0,
+      "epoch": 0.025020850708924104,
+      "grad_norm": 14.06130768102936,
+      "kl": 6.4025,
+      "learning_rate": 1.9834306366416458e-06,
+      "loss": 0.121,
+      "num_tokens": 1817552.0,
+      "reward": -0.14887901425361633,
+      "reward_std": 3.041101009249687,
+      "rewards/generate_all_rewards": -0.14887901425361633,
+      "step": 150
+    },
+    {
+      "clip_ratio": 0.008756103515625,
+      "completion_length": 1024.0,
+      "epoch": 0.03336113427856547,
+      "grad_norm": 3.8576112038400194,
+      "kl": 6.95140625,
+      "learning_rate": 1.9778704475952183e-06,
+      "loss": 0.1366,
+      "num_tokens": 2408832.0,
+      "reward": 0.1574733567237854,
+      "reward_std": 3.0124819111824035,
+      "rewards/generate_all_rewards": 0.1574733567237854,
+      "step": 200
+    },
+    {
+      "clip_ratio": 0.00777587890625,
+      "completion_length": 1024.0,
+      "epoch": 0.041701417848206836,
+      "grad_norm": 8.583233886890447,
+      "kl": 4.32921875,
+      "learning_rate": 1.9723102585487904e-06,
+      "loss": 0.0816,
+      "num_tokens": 3012520.0,
+      "reward": 1.650909082889557,
+      "reward_std": 2.1874516403675077,
+      "rewards/generate_all_rewards": 1.650909082889557,
+      "step": 250
+    },
+    {
+      "clip_ratio": 0.0061767578125,
+      "completion_length": 1024.0,
+      "epoch": 0.05004170141784821,
+      "grad_norm": 9.521656811246332,
+      "kl": 2.9395703125,
+      "learning_rate": 1.966750069502363e-06,
+      "loss": 0.0528,
+      "num_tokens": 3635196.0,
+      "reward": 2.0889700829982756,
+      "reward_std": 1.676103963404894,
+      "rewards/generate_all_rewards": 2.0889700829982756,
+      "step": 300
+    },
+    {
+      "clip_ratio": 0.00710205078125,
+      "completion_length": 1024.0,
+      "epoch": 0.058381984987489574,
+      "grad_norm": 3.2056926065848876,
+      "kl": 3.523515625,
+      "learning_rate": 1.9611898804559355e-06,
+      "loss": 0.0655,
+      "num_tokens": 4244948.0,
+      "reward": 2.5839235186576843,
+      "reward_std": 1.675377692580223,
+      "rewards/generate_all_rewards": 2.5839235186576843,
+      "step": 350
+    },
+    {
+      "clip_ratio": 0.0070361328125,
+      "completion_length": 1024.0,
+      "epoch": 0.06672226855713094,
+      "grad_norm": 5.870128860699156,
+      "kl": 3.76171875,
+      "learning_rate": 1.955629691409508e-06,
+      "loss": 0.07,
+      "num_tokens": 4843180.0,
+      "reward": 1.3626538455486297,
+      "reward_std": 1.5572759065032005,
+      "rewards/generate_all_rewards": 1.3626538455486297,
+      "step": 400
+    },
+    {
+      "clip_ratio": 0.007911376953125,
+      "completion_length": 1024.0,
+      "epoch": 0.0750625521267723,
+      "grad_norm": 3.5520731952805487,
+      "kl": 3.944765625,
+      "learning_rate": 1.95006950236308e-06,
+      "loss": 0.0754,
+      "num_tokens": 5447660.0,
+      "reward": 2.8184473848342897,
+      "reward_std": 1.537802910655737,
+      "rewards/generate_all_rewards": 2.8184473848342897,
+      "step": 450
+    },
+    {
+      "clip_ratio": 0.013525390625,
+      "completion_length": 1024.0,
+      "epoch": 0.08340283569641367,
+      "grad_norm": 7.328124794124316,
+      "kl": 3.373125,
+      "learning_rate": 1.9445093133166527e-06,
+      "loss": 0.0614,
+      "num_tokens": 6051428.0,
+      "reward": 2.4807280468940736,
+      "reward_std": 0.9742915752530098,
+      "rewards/generate_all_rewards": 2.4807280468940736,
+      "step": 500
+    },
+    {
+      "clip_ratio": 0.010689697265625,
+      "completion_length": 1024.0,
+      "epoch": 0.09174311926605505,
+      "grad_norm": 8.879346331019656,
+      "kl": 3.567578125,
+      "learning_rate": 1.938949124270225e-06,
+      "loss": 0.0664,
+      "num_tokens": 6652180.0,
+      "reward": 2.4438777142763137,
+      "reward_std": 1.1059524276852608,
+      "rewards/generate_all_rewards": 2.4438777142763137,
+      "step": 550
+    },
+    {
+      "clip_ratio": 0.006470947265625,
+      "completion_length": 1024.0,
+      "epoch": 0.10008340283569642,
+      "grad_norm": 15.87593845075513,
+      "kl": 1.68625,
+      "learning_rate": 1.9333889352237977e-06,
+      "loss": 0.0261,
+      "num_tokens": 7262336.0,
+      "reward": 3.7379005312919618,
+      "reward_std": 0.9547000896930694,
+      "rewards/generate_all_rewards": 3.7379005312919618,
+      "step": 600
+    },
+    {
+      "clip_ratio": 0.008651123046875,
+      "completion_length": 1024.0,
+      "epoch": 0.10842368640533778,
+      "grad_norm": 16.184176794809037,
+      "kl": 1.2030859375,
+      "learning_rate": 1.92782874617737e-06,
+      "loss": 0.0164,
+      "num_tokens": 7853484.0,
+      "reward": 3.1192458760738373,
+      "reward_std": 0.7382513232529163,
+      "rewards/generate_all_rewards": 3.1192458760738373,
+      "step": 650
+    },
+    {
+      "clip_ratio": 0.007640380859375,
+      "completion_length": 1024.0,
+      "epoch": 0.11676396997497915,
+      "grad_norm": 12.281047071498108,
+      "kl": 3.4060546875,
+      "learning_rate": 1.9222685571309424e-06,
+      "loss": 0.0604,
+      "num_tokens": 8459596.0,
+      "reward": 3.791183285713196,
+      "reward_std": 1.1176388543844222,
+      "rewards/generate_all_rewards": 3.791183285713196,
+      "step": 700
+    },
+    {
+      "clip_ratio": 0.00601806640625,
+      "completion_length": 1024.0,
+      "epoch": 0.12510425354462051,
+      "grad_norm": 10.59835515773422,
+      "kl": 1.3465234375,
+      "learning_rate": 1.916708368084515e-06,
+      "loss": 0.0197,
+      "num_tokens": 9062924.0,
+      "reward": 3.9991130876541137,
+      "reward_std": 0.933748829215765,
+      "rewards/generate_all_rewards": 3.9991130876541137,
+      "step": 750
+    },
+    {
+      "clip_ratio": 0.008304443359375,
+      "completion_length": 1024.0,
+      "epoch": 0.13344453711426188,
+      "grad_norm": 4.399868599163885,
+      "kl": 2.779375,
+      "learning_rate": 1.911148179038087e-06,
+      "loss": 0.0503,
+      "num_tokens": 9676376.0,
+      "reward": 3.1121265506744384,
+      "reward_std": 0.8076464046537876,
+      "rewards/generate_all_rewards": 3.1121265506744384,
+      "step": 800
+    },
+    {
+      "clip_ratio": 0.00646484375,
+      "completion_length": 1024.0,
+      "epoch": 0.14178482068390325,
+      "grad_norm": 12.33402253801181,
+      "kl": 2.50671875,
+      "learning_rate": 1.9055879899916596e-06,
+      "loss": 0.0423,
+      "num_tokens": 10291400.0,
+      "reward": 2.333748939037323,
+      "reward_std": 0.9894427044689655,
+      "rewards/generate_all_rewards": 2.333748939037323,
+      "step": 850
+    },
+    {
+      "clip_ratio": 0.01214111328125,
+      "completion_length": 1024.0,
+      "epoch": 0.1501251042535446,
+      "grad_norm": 14.289114263028262,
+      "kl": 1.8778515625,
+      "learning_rate": 1.9000278009452319e-06,
+      "loss": 0.0295,
+      "num_tokens": 10887736.0,
+      "reward": 2.980680326223373,
+      "reward_std": 0.6083575973659754,
+      "rewards/generate_all_rewards": 2.980680326223373,
+      "step": 900
+    },
+    {
+      "clip_ratio": 0.009613037109375,
+      "completion_length": 1024.0,
+      "epoch": 0.15846538782318598,
+      "grad_norm": 10.14315612638421,
+      "kl": 2.92171875,
+      "learning_rate": 1.8944676118988046e-06,
+      "loss": 0.0497,
+      "num_tokens": 11494176.0,
+      "reward": 2.402811622619629,
+      "reward_std": 1.0233210255205631,
+      "rewards/generate_all_rewards": 2.402811622619629,
+      "step": 950
+    },
+    {
+      "clip_ratio": 0.00775146484375,
+      "completion_length": 1024.0,
+      "epoch": 0.16680567139282734,
+      "grad_norm": 3.95911283332978,
+      "kl": 1.9640625,
+      "learning_rate": 1.888907422852377e-06,
+      "loss": 0.0329,
+      "num_tokens": 12103840.0,
+      "reward": 3.709632108211517,
+      "reward_std": 0.8795299279689789,
+      "rewards/generate_all_rewards": 3.709632108211517,
+      "step": 1000
+    },
+    {
+      "clip_ratio": 0.009173583984375,
+      "completion_length": 1024.0,
+      "epoch": 0.17514595496246874,
+      "grad_norm": 2.404205261708653,
+      "kl": 2.656328125,
+      "learning_rate": 1.8833472338059493e-06,
+      "loss": 0.0456,
+      "num_tokens": 12706496.0,
+      "reward": 3.0651380997896194,
+      "reward_std": 0.8471601485460997,
+      "rewards/generate_all_rewards": 3.0651380997896194,
+      "step": 1050
+    },
+    {
+      "clip_ratio": 0.00503662109375,
+      "completion_length": 1024.0,
+      "epoch": 0.1834862385321101,
+      "grad_norm": 16.0574388547409,
+      "kl": 2.2871875,
+      "learning_rate": 1.8777870447595216e-06,
+      "loss": 0.0421,
+      "num_tokens": 13320876.0,
+      "reward": 3.456456989645958,
+      "reward_std": 0.8847015166282653,
+      "rewards/generate_all_rewards": 3.456456989645958,
+      "step": 1100
+    },
+    {
+      "clip_ratio": 0.004837646484375,
+      "completion_length": 1024.0,
+      "epoch": 0.19182652210175147,
+      "grad_norm": 7.653139462749305,
+      "kl": 2.241796875,
+      "learning_rate": 1.8722268557130943e-06,
+      "loss": 0.0392,
+      "num_tokens": 13915940.0,
+      "reward": 2.1091002190113066,
+      "reward_std": 0.8859185457229615,
+      "rewards/generate_all_rewards": 2.1091002190113066,
+      "step": 1150
+    },
+    {
+      "clip_ratio": 0.0070654296875,
+      "completion_length": 1024.0,
+      "epoch": 0.20016680567139283,
+      "grad_norm": 7.678037556240038,
+      "kl": 2.1728125,
+      "learning_rate": 1.8666666666666667e-06,
+      "loss": 0.0371,
+      "num_tokens": 14512636.0,
+      "reward": 1.9523184645175933,
+      "reward_std": 1.1099144089221955,
+      "rewards/generate_all_rewards": 1.9523184645175933,
+      "step": 1200
+    },
+    {
+      "clip_ratio": 0.0082373046875,
+      "completion_length": 1024.0,
+      "epoch": 0.2085070892410342,
+      "grad_norm": 20.776332057103467,
+      "kl": 2.02796875,
+      "learning_rate": 1.861106477620239e-06,
+      "loss": 0.0341,
+      "num_tokens": 15119036.0,
+      "reward": 3.578354719877243,
+      "reward_std": 0.9086613065004349,
+      "rewards/generate_all_rewards": 3.578354719877243,
+      "step": 1250
+    },
+    {
+      "clip_ratio": 0.00704345703125,
+      "completion_length": 1024.0,
+      "epoch": 0.21684737281067556,
+      "grad_norm": 2.912945916637192,
+      "kl": 2.1746875,
+      "learning_rate": 1.8555462885738113e-06,
+      "loss": 0.0361,
+      "num_tokens": 15734840.0,
+      "reward": 3.4535324451327325,
+      "reward_std": 1.1092214401811362,
+      "rewards/generate_all_rewards": 3.4535324451327325,
+      "step": 1300
+    },
+    {
+      "clip_ratio": 0.005543212890625,
+      "completion_length": 1024.0,
+      "epoch": 0.22518765638031693,
+      "grad_norm": 4.363074357591489,
+      "kl": 1.61734375,
+      "learning_rate": 1.849986099527384e-06,
+      "loss": 0.0271,
+      "num_tokens": 16342192.0,
+      "reward": 3.3417653107643126,
+      "reward_std": 0.829094213321805,
+      "rewards/generate_all_rewards": 3.3417653107643126,
+      "step": 1350
+    },
+    {
+      "clip_ratio": 0.007327880859375,
+      "completion_length": 1024.0,
+      "epoch": 0.2335279399499583,
+      "grad_norm": 51.189874692363496,
+      "kl": 3.0484375,
+      "learning_rate": 1.8444259104809564e-06,
+      "loss": 0.053,
+      "num_tokens": 16957324.0,
+      "reward": 3.9473286485671997,
+      "reward_std": 0.7120842409878969,
+      "rewards/generate_all_rewards": 3.9473286485671997,
+      "step": 1400
+    },
+    {
+      "clip_ratio": 0.006270751953125,
+      "completion_length": 1024.0,
+      "epoch": 0.24186822351959966,
+      "grad_norm": 7.8429753301262695,
+      "kl": 2.339921875,
+      "learning_rate": 1.8388657214345287e-06,
+      "loss": 0.0396,
+      "num_tokens": 17558900.0,
+      "reward": 2.530026806592941,
+      "reward_std": 0.9393242979049683,
+      "rewards/generate_all_rewards": 2.530026806592941,
+      "step": 1450
+    },
+    {
+      "clip_ratio": 0.0064697265625,
+      "completion_length": 1024.0,
+      "epoch": 0.25020850708924103,
+      "grad_norm": 5.769136809903878,
+      "kl": 2.0521875,
+      "learning_rate": 1.833305532388101e-06,
+      "loss": 0.0349,
+      "num_tokens": 18161068.0,
+      "reward": 3.6563946413993835,
+      "reward_std": 0.5981371226906776,
+      "rewards/generate_all_rewards": 3.6563946413993835,
+      "step": 1500
+    },
+    {
+      "clip_ratio": 0.0059423828125,
+      "completion_length": 1024.0,
+      "epoch": 0.2585487906588824,
+      "grad_norm": 4.37800770904507,
+      "kl": 2.79046875,
+      "learning_rate": 1.8277453433416735e-06,
+      "loss": 0.0483,
+      "num_tokens": 18776456.0,
+      "reward": 2.3797482299804686,
+      "reward_std": 1.1877363938093186,
+      "rewards/generate_all_rewards": 2.3797482299804686,
+      "step": 1550
+    },
+    {
+      "clip_ratio": 0.009705810546875,
+      "completion_length": 1024.0,
+      "epoch": 0.26688907422852376,
+      "grad_norm": 5.849705216576927,
+      "kl": 2.444140625,
+      "learning_rate": 1.822185154295246e-06,
+      "loss": 0.0422,
+      "num_tokens": 19375140.0,
+      "reward": 3.1630379295349123,
+      "reward_std": 0.7082004471123219,
+      "rewards/generate_all_rewards": 3.1630379295349123,
+      "step": 1600
+    },
+    {
+      "clip_ratio": 0.006885986328125,
+      "completion_length": 1024.0,
+      "epoch": 0.27522935779816515,
+      "grad_norm": 11.341122516574023,
+      "kl": 1.932421875,
+      "learning_rate": 1.8166249652488184e-06,
+      "loss": 0.0311,
+      "num_tokens": 19975452.0,
+      "reward": 2.987669861316681,
+      "reward_std": 0.7325755050033331,
+      "rewards/generate_all_rewards": 2.987669861316681,
+      "step": 1650
+    },
+    {
+      "clip_ratio": 0.0052978515625,
+      "completion_length": 1024.0,
+      "epoch": 0.2835696413678065,
+      "grad_norm": 16.88086046670062,
+      "kl": 2.680234375,
+      "learning_rate": 1.8110647762023907e-06,
+      "loss": 0.0472,
+      "num_tokens": 20587572.0,
+      "reward": 3.527438926696777,
+      "reward_std": 1.062203018963337,
+      "rewards/generate_all_rewards": 3.527438926696777,
+      "step": 1700
+    },
+    {
+      "clip_ratio": 0.007686767578125,
+      "completion_length": 1024.0,
+      "epoch": 0.2919099249374479,
+      "grad_norm": 3.020717228432076,
+      "kl": 2.73734375,
+      "learning_rate": 1.8055045871559633e-06,
+      "loss": 0.0487,
+      "num_tokens": 21194080.0,
+      "reward": 3.477699854373932,
+      "reward_std": 0.9699093826115132,
+      "rewards/generate_all_rewards": 3.477699854373932,
+      "step": 1750
+    },
+    {
+      "clip_ratio": 0.005350341796875,
+      "completion_length": 1024.0,
+      "epoch": 0.3002502085070892,
+      "grad_norm": 10.250138946853658,
+      "kl": 1.2153125,
+      "learning_rate": 1.7999443981095358e-06,
+      "loss": 0.0179,
+      "num_tokens": 21802912.0,
+      "reward": 4.211458885669709,
+      "reward_std": 0.5974260994791984,
+      "rewards/generate_all_rewards": 4.211458885669709,
+      "step": 1800
+    },
+    {
+      "clip_ratio": 0.00499755859375,
+      "completion_length": 1024.0,
+      "epoch": 0.3085904920767306,
+      "grad_norm": 2.728898524811883,
+      "kl": 2.489296875,
+      "learning_rate": 1.7943842090631081e-06,
+      "loss": 0.0422,
+      "num_tokens": 22416820.0,
+      "reward": 3.9941040158271788,
+      "reward_std": 0.9688579052686691,
+      "rewards/generate_all_rewards": 3.9941040158271788,
+      "step": 1850
+    },
+    {
+      "clip_ratio": 0.004656982421875,
+      "completion_length": 1024.0,
+      "epoch": 0.31693077564637195,
+      "grad_norm": 3.263929854853948,
+      "kl": 1.40859375,
+      "learning_rate": 1.7888240200166804e-06,
+      "loss": 0.0225,
+      "num_tokens": 23012076.0,
+      "reward": 3.5541777658462523,
+      "reward_std": 0.6293248501420021,
+      "rewards/generate_all_rewards": 3.5541777658462523,
+      "step": 1900
+    },
+    {
+      "clip_ratio": 0.00759521484375,
+      "completion_length": 1024.0,
+      "epoch": 0.32527105921601335,
+      "grad_norm": 43.80415642034278,
+      "kl": 2.095546875,
+      "learning_rate": 1.783263830970253e-06,
+      "loss": 0.0352,
+      "num_tokens": 23611824.0,
+      "reward": 3.3262206852436065,
+      "reward_std": 0.669879068210721,
+      "rewards/generate_all_rewards": 3.3262206852436065,
+      "step": 1950
+    },
+    {
+      "clip_ratio": 0.00798095703125,
+      "completion_length": 1024.0,
+      "epoch": 0.3336113427856547,
+      "grad_norm": 19.175386958975828,
+      "kl": 2.42546875,
+      "learning_rate": 1.7777036419238253e-06,
+      "loss": 0.0422,
+      "num_tokens": 24227908.0,
+      "reward": 3.655040820837021,
+      "reward_std": 0.8108076846599579,
+      "rewards/generate_all_rewards": 3.655040820837021,
+      "step": 2000
+    },
+    {
+      "clip_ratio": 0.00546875,
+      "completion_length": 1024.0,
+      "epoch": 0.3419516263552961,
+      "grad_norm": 3.351550298003648,
+      "kl": 2.765546875,
+      "learning_rate": 1.7721434528773978e-06,
+      "loss": 0.0498,
+      "num_tokens": 24836204.0,
+      "reward": 3.816224058866501,
+      "reward_std": 0.7453654401004315,
+      "rewards/generate_all_rewards": 3.816224058866501,
+      "step": 2050
+    },
+    {
+      "clip_ratio": 0.008900146484375,
+      "completion_length": 1024.0,
+      "epoch": 0.3502919099249375,
+      "grad_norm": 3.2293957852618385,
+      "kl": 2.447734375,
+      "learning_rate": 1.7665832638309701e-06,
+      "loss": 0.045,
+      "num_tokens": 25456812.0,
+      "reward": 4.989015902280808,
+      "reward_std": 0.7733614352345467,
+      "rewards/generate_all_rewards": 4.989015902280808,
+      "step": 2100
+    },
+    {
+      "clip_ratio": 0.007430419921875,
+      "completion_length": 1024.0,
+      "epoch": 0.3586321934945788,
+      "grad_norm": 8.737056342531416,
+      "kl": 5.238203125,
+      "learning_rate": 1.7610230747845425e-06,
+      "loss": 0.1001,
+      "num_tokens": 26071304.0,
+      "reward": 2.9227333569526674,
+      "reward_std": 1.5148471367359162,
+      "rewards/generate_all_rewards": 2.9227333569526674,
+      "step": 2150
+    },
+    {
+      "clip_ratio": 0.007855224609375,
+      "completion_length": 1024.0,
+      "epoch": 0.3669724770642202,
+      "grad_norm": 2719.9654735665067,
+      "kl": 2.415546875,
+      "learning_rate": 1.755462885738115e-06,
+      "loss": 0.0428,
+      "num_tokens": 26679488.0,
+      "reward": 3.643260437250137,
+      "reward_std": 0.8649014130234718,
+      "rewards/generate_all_rewards": 3.643260437250137,
+      "step": 2200
+    },
+    {
+      "clip_ratio": 0.006314697265625,
+      "completion_length": 1024.0,
+      "epoch": 0.37531276063386154,
+      "grad_norm": 5.853101521377133,
+      "kl": 3.21375,
+      "learning_rate": 1.7499026966916875e-06,
+      "loss": 0.0575,
+      "num_tokens": 27294504.0,
+      "reward": 2.968059607744217,
+      "reward_std": 1.0330314177647233,
+      "rewards/generate_all_rewards": 2.968059607744217,
+      "step": 2250
+    },
+    {
+      "clip_ratio": 0.005992431640625,
+      "completion_length": 1024.0,
+      "epoch": 0.38365304420350294,
+      "grad_norm": 8.224103504347667,
+      "kl": 5.151015625,
+      "learning_rate": 1.7443425076452599e-06,
+      "loss": 0.0975,
+      "num_tokens": 27914656.0,
+      "reward": 2.8436513328552246,
+      "reward_std": 2.013354176878929,
+      "rewards/generate_all_rewards": 2.8436513328552246,
+      "step": 2300
+    },
+    {
+      "clip_ratio": 0.0057470703125,
+      "completion_length": 1024.0,
+      "epoch": 0.3919933277731443,
+      "grad_norm": 7.8051782744106415,
+      "kl": 4.03734375,
+      "learning_rate": 1.7387823185988322e-06,
+      "loss": 0.0742,
+      "num_tokens": 28516504.0,
+      "reward": 2.9896522617340087,
+      "reward_std": 1.8563957458920777,
+      "rewards/generate_all_rewards": 2.9896522617340087,
+      "step": 2350
+    },
+    {
+      "clip_ratio": 0.004320068359375,
+      "completion_length": 1024.0,
+      "epoch": 0.40033361134278567,
+      "grad_norm": 3.3006539122678276,
+      "kl": 2.824375,
+      "learning_rate": 1.7332221295524047e-06,
+      "loss": 0.0525,
+      "num_tokens": 29117264.0,
+      "reward": 3.0978617167472837,
+      "reward_std": 1.175471960231662,
+      "rewards/generate_all_rewards": 3.0978617167472837,
+      "step": 2400
+    },
+    {
+      "clip_ratio": 0.0046337890625,
+      "completion_length": 1024.0,
+      "epoch": 0.408673894912427,
+      "grad_norm": 12.460451063363156,
+      "kl": 2.73203125,
+      "learning_rate": 1.727661940505977e-06,
+      "loss": 0.0495,
+      "num_tokens": 29726312.0,
+      "reward": 3.126603194475174,
+      "reward_std": 1.1947118404507637,
+      "rewards/generate_all_rewards": 3.126603194475174,
+      "step": 2450
+    },
+    {
+      "clip_ratio": 0.007073974609375,
+      "completion_length": 1024.0,
+      "epoch": 0.4170141784820684,
+      "grad_norm": 6.23861285525636,
+      "kl": 3.635703125,
+      "learning_rate": 1.7221017514595496e-06,
+      "loss": 0.0665,
+      "num_tokens": 30314692.0,
+      "reward": 2.6027573454380035,
+      "reward_std": 0.7329472535848618,
+      "rewards/generate_all_rewards": 2.6027573454380035,
+      "step": 2500
+    },
+    {
+      "clip_ratio": 0.005120849609375,
+      "completion_length": 1024.0,
+      "epoch": 0.42535446205170974,
+      "grad_norm": 12.279697999562723,
+      "kl": 2.185078125,
+      "learning_rate": 1.7165415624131219e-06,
+      "loss": 0.0379,
+      "num_tokens": 30929740.0,
+      "reward": 4.00673865199089,
+      "reward_std": 0.6778046156093478,
+      "rewards/generate_all_rewards": 4.00673865199089,
+      "step": 2550
+    },
+    {
+      "clip_ratio": 0.00871337890625,
+      "completion_length": 1024.0,
+      "epoch": 0.43369474562135113,
+      "grad_norm": 12.736320988063575,
+      "kl": 2.5959375,
+      "learning_rate": 1.7109813733666944e-06,
+      "loss": 0.0452,
+      "num_tokens": 31523968.0,
+      "reward": 3.6301355397701265,
+      "reward_std": 0.6656330497562886,
+      "rewards/generate_all_rewards": 3.6301355397701265,
+      "step": 2600
+    },
+    {
+      "clip_ratio": 0.0074755859375,
+      "completion_length": 1024.0,
+      "epoch": 0.44203502919099247,
+      "grad_norm": 2.6222378119323952,
+      "kl": 3.994296875,
+      "learning_rate": 1.7054211843202667e-06,
+      "loss": 0.0724,
+      "num_tokens": 32128080.0,
+      "reward": 2.5648028159141543,
+      "reward_std": 1.231038143262267,
+      "rewards/generate_all_rewards": 2.5648028159141543,
+      "step": 2650
+    },
+    {
+      "clip_ratio": 0.005047607421875,
+      "completion_length": 1024.0,
+      "epoch": 0.45037531276063386,
+      "grad_norm": 9.275686783269586,
+      "kl": 2.709140625,
+      "learning_rate": 1.6998609952738393e-06,
+      "loss": 0.0454,
+      "num_tokens": 32720372.0,
+      "reward": 3.4994620156288145,
+      "reward_std": 0.4822941809147596,
+      "rewards/generate_all_rewards": 3.4994620156288145,
+      "step": 2700
+    },
+    {
+      "clip_ratio": 0.0049267578125,
+      "completion_length": 1024.0,
+      "epoch": 0.45871559633027525,
+      "grad_norm": 8.541221675792642,
+      "kl": 2.709296875,
+      "learning_rate": 1.6943008062274116e-06,
+      "loss": 0.0486,
+      "num_tokens": 33336552.0,
+      "reward": 3.0300825768709183,
+      "reward_std": 0.5665881184674799,
+      "rewards/generate_all_rewards": 3.0300825768709183,
+      "step": 2750
+    },
+    {
+      "clip_ratio": 0.0037109375,
+      "completion_length": 1024.0,
+      "epoch": 0.4670558798999166,
+      "grad_norm": 5.125419890585354,
+      "kl": 1.825859375,
+      "learning_rate": 1.6887406171809841e-06,
+      "loss": 0.0302,
+      "num_tokens": 33945248.0,
+      "reward": 3.4999762004613877,
+      "reward_std": 0.5717377527058125,
+      "rewards/generate_all_rewards": 3.4999762004613877,
+      "step": 2800
+    },
+    {
+      "clip_ratio": 0.0053125,
+      "completion_length": 1024.0,
+      "epoch": 0.475396163469558,
+      "grad_norm": 10.202348905573427,
+      "kl": 2.3403125,
+      "learning_rate": 1.6831804281345565e-06,
+      "loss": 0.0395,
+      "num_tokens": 34550132.0,
+      "reward": 2.1186443191766737,
+      "reward_std": 0.3818393291532993,
+      "rewards/generate_all_rewards": 2.1186443191766737,
+      "step": 2850
+    },
+    {
+      "clip_ratio": 0.00617431640625,
+      "completion_length": 1024.0,
+      "epoch": 0.4837364470391993,
+      "grad_norm": 2.246128805290571,
+      "kl": 2.578515625,
+      "learning_rate": 1.677620239088129e-06,
+      "loss": 0.0449,
+      "num_tokens": 35151512.0,
+      "reward": 2.3393343752622604,
+      "reward_std": 0.6313513360917569,
+      "rewards/generate_all_rewards": 2.3393343752622604,
+      "step": 2900
+    },
+    {
+      "clip_ratio": 0.004263916015625,
+      "completion_length": 1024.0,
+      "epoch": 0.4920767306088407,
+      "grad_norm": 9.691052501917392,
+      "kl": 1.461484375,
+      "learning_rate": 1.6720600500417013e-06,
+      "loss": 0.0248,
+      "num_tokens": 35754108.0,
+      "reward": 3.9832346379756927,
+      "reward_std": 0.4096560184657574,
+      "rewards/generate_all_rewards": 3.9832346379756927,
+      "step": 2950
+    },
+    {
+      "clip_ratio": 0.0049609375,
+      "completion_length": 1024.0,
+      "epoch": 0.5004170141784821,
+      "grad_norm": 2.452304665586332,
+      "kl": 2.235390625,
+      "learning_rate": 1.6664998609952738e-06,
+      "loss": 0.0385,
+      "num_tokens": 36368004.0,
+      "reward": 4.140755372047424,
+      "reward_std": 0.432945069745183,
+      "rewards/generate_all_rewards": 4.140755372047424,
+      "step": 3000
+    },
+    {
+      "clip_ratio": 0.006944580078125,
+      "completion_length": 1024.0,
+      "epoch": 0.5087572977481234,
+      "grad_norm": 4.598716528073911,
+      "kl": 2.54578125,
+      "learning_rate": 1.6609396719488462e-06,
+      "loss": 0.0442,
+      "num_tokens": 36977484.0,
+      "reward": 3.9808691453933718,
+      "reward_std": 0.4201528353989124,
+      "rewards/generate_all_rewards": 3.9808691453933718,
+      "step": 3050
+    },
+    {
+      "clip_ratio": 0.00439697265625,
+      "completion_length": 1024.0,
+      "epoch": 0.5170975813177648,
+      "grad_norm": 9.080608114471648,
+      "kl": 2.86203125,
+      "learning_rate": 1.6553794829024185e-06,
+      "loss": 0.0512,
+      "num_tokens": 37583980.0,
+      "reward": 2.9114849293231964,
+      "reward_std": 0.6796937258541584,
+      "rewards/generate_all_rewards": 2.9114849293231964,
+      "step": 3100
+    },
+    {
+      "clip_ratio": 0.00376708984375,
+      "completion_length": 1024.0,
+      "epoch": 0.5254378648874062,
+      "grad_norm": 8.167703275815628,
+      "kl": 3.2890625,
+      "learning_rate": 1.649819293855991e-06,
+      "loss": 0.0587,
+      "num_tokens": 38185184.0,
+      "reward": 3.087721600532532,
+      "reward_std": 0.8554429135844112,
+      "rewards/generate_all_rewards": 3.087721600532532,
+      "step": 3150
+    },
+    {
+      "clip_ratio": 0.00428466796875,
+      "completion_length": 1024.0,
+      "epoch": 0.5337781484570475,
+      "grad_norm": 9.265129726935722,
+      "kl": 3.20609375,
+      "learning_rate": 1.6442591048095636e-06,
+      "loss": 0.0587,
+      "num_tokens": 38794784.0,
+      "reward": 3.553533103466034,
+      "reward_std": 0.7431424564123154,
+      "rewards/generate_all_rewards": 3.553533103466034,
+      "step": 3200
+    },
+    {
+      "clip_ratio": 0.006676025390625,
+      "completion_length": 1024.0,
+      "epoch": 0.5421184320266889,
+      "grad_norm": 5.64854377106438,
+      "kl": 2.15875,
+      "learning_rate": 1.6386989157631359e-06,
+      "loss": 0.0361,
+      "num_tokens": 39402240.0,
+      "reward": 2.871965115070343,
+      "reward_std": 0.42509609084576366,
+      "rewards/generate_all_rewards": 2.871965115070343,
+      "step": 3250
+    },
+    {
+      "clip_ratio": 0.002960205078125,
+      "completion_length": 1024.0,
+      "epoch": 0.5504587155963303,
+      "grad_norm": 4.95636107786261,
+      "kl": 2.88828125,
+      "learning_rate": 1.6331387267167082e-06,
+      "loss": 0.0513,
+      "num_tokens": 40009924.0,
+      "reward": 3.8079018712043764,
+      "reward_std": 0.7513183067925274,
+      "rewards/generate_all_rewards": 3.8079018712043764,
+      "step": 3300
+    },
+    {
+      "clip_ratio": 0.004202880859375,
+      "completion_length": 1024.0,
+      "epoch": 0.5587989991659716,
+      "grad_norm": 6.97217793872751,
+      "kl": 3.969921875,
+      "learning_rate": 1.6275785376702807e-06,
+      "loss": 0.0709,
+      "num_tokens": 40625152.0,
+      "reward": 2.3424331378936767,
+      "reward_std": 1.1434777556359768,
+      "rewards/generate_all_rewards": 2.3424331378936767,
+      "step": 3350
+    },
+    {
+      "clip_ratio": 0.005111083984375,
+      "completion_length": 1024.0,
+      "epoch": 0.567139282735613,
+      "grad_norm": 15.565181340971451,
+      "kl": 3.036171875,
+      "learning_rate": 1.6220183486238533e-06,
+      "loss": 0.0562,
+      "num_tokens": 41229704.0,
+      "reward": 3.3502265119552614,
+      "reward_std": 1.13145647123456,
+      "rewards/generate_all_rewards": 3.3502265119552614,
+      "step": 3400
+    },
+    {
+      "clip_ratio": 0.00511962890625,
+      "completion_length": 1024.0,
+      "epoch": 0.5754795663052544,
+      "grad_norm": 7.176359529921464,
+      "kl": 2.051875,
+      "learning_rate": 1.6164581595774256e-06,
+      "loss": 0.0347,
+      "num_tokens": 41825184.0,
+      "reward": 3.6443237620592117,
+      "reward_std": 0.7032080652937293,
+      "rewards/generate_all_rewards": 3.6443237620592117,
+      "step": 3450
+    },
+    {
+      "clip_ratio": 0.011055908203125,
+      "completion_length": 1024.0,
+      "epoch": 0.5838198498748958,
+      "grad_norm": 4.526966823196885,
+      "kl": 2.79703125,
+      "learning_rate": 1.610897970530998e-06,
+      "loss": 0.0513,
+      "num_tokens": 42442268.0,
+      "reward": 4.048895968794823,
+      "reward_std": 0.958944509550929,
+      "rewards/generate_all_rewards": 4.048895968794823,
+      "step": 3500
+    },
+    {
+      "clip_ratio": 0.00597412109375,
+      "completion_length": 1024.0,
+      "epoch": 0.5921601334445371,
+      "grad_norm": 16.586877688863396,
+      "kl": 2.502890625,
+      "learning_rate": 1.6053377814845702e-06,
+      "loss": 0.0449,
+      "num_tokens": 43049748.0,
+      "reward": 3.0637675487995146,
+      "reward_std": 0.749227255731821,
+      "rewards/generate_all_rewards": 3.0637675487995146,
+      "step": 3550
+    },
+    {
+      "clip_ratio": 0.004638671875,
+      "completion_length": 1024.0,
+      "epoch": 0.6005004170141784,
+      "grad_norm": 10.447493313935354,
+      "kl": 3.10890625,
+      "learning_rate": 1.599777592438143e-06,
+      "loss": 0.0547,
+      "num_tokens": 43651800.0,
+      "reward": 2.9153013825416565,
+      "reward_std": 0.824127499461174,
+      "rewards/generate_all_rewards": 2.9153013825416565,
+      "step": 3600
+    },
+    {
+      "clip_ratio": 0.005782470703125,
+      "completion_length": 1024.0,
+      "epoch": 0.6088407005838199,
+      "grad_norm": 32.10338980603829,
+      "kl": 2.6153125,
+      "learning_rate": 1.5942174033917153e-06,
+      "loss": 0.0471,
+      "num_tokens": 44241760.0,
+      "reward": 3.5675123274326324,
+      "reward_std": 0.6166292682575295,
+      "rewards/generate_all_rewards": 3.5675123274326324,
+      "step": 3650
+    },
+    {
+      "clip_ratio": 0.005928955078125,
+      "completion_length": 1024.0,
+      "epoch": 0.6171809841534612,
+      "grad_norm": 6.685408071704869,
+      "kl": 3.515390625,
+      "learning_rate": 1.5886572143452876e-06,
+      "loss": 0.0666,
+      "num_tokens": 44852096.0,
+      "reward": 3.9864674687385557,
+      "reward_std": 0.8102425380796194,
+      "rewards/generate_all_rewards": 3.9864674687385557,
+      "step": 3700
+    },
+    {
+      "clip_ratio": 0.00569580078125,
+      "completion_length": 1024.0,
+      "epoch": 0.6255212677231026,
+      "grad_norm": 3.368030262404586,
+      "kl": 2.3678125,
+      "learning_rate": 1.58309702529886e-06,
+      "loss": 0.0432,
+      "num_tokens": 45460220.0,
+      "reward": 3.25111713051796,
+      "reward_std": 0.9311061368137598,
+      "rewards/generate_all_rewards": 3.25111713051796,
+      "step": 3750
+    },
+    {
+      "clip_ratio": 0.006986083984375,
+      "completion_length": 1024.0,
+      "epoch": 0.6338615512927439,
+      "grad_norm": 7.551706236295235,
+      "kl": 2.4209375,
+      "learning_rate": 1.5775368362524327e-06,
+      "loss": 0.041,
+      "num_tokens": 46072700.0,
+      "reward": 2.695195918381214,
+      "reward_std": 0.8751908247172833,
+      "rewards/generate_all_rewards": 2.695195918381214,
+      "step": 3800
+    },
+    {
+      "clip_ratio": 0.003411865234375,
+      "completion_length": 1024.0,
+      "epoch": 0.6422018348623854,
+      "grad_norm": 18.6347155666052,
+      "kl": 3.246953125,
+      "learning_rate": 1.571976647206005e-06,
+      "loss": 0.0602,
+      "num_tokens": 46667728.0,
+      "reward": 2.4113513553142547,
+      "reward_std": 1.3523825246095658,
+      "rewards/generate_all_rewards": 2.4113513553142547,
+      "step": 3850
+    },
+    {
+      "clip_ratio": 0.004542236328125,
+      "completion_length": 1024.0,
+      "epoch": 0.6505421184320267,
+      "grad_norm": 21.13739448151968,
+      "kl": 2.31109375,
+      "learning_rate": 1.5664164581595773e-06,
+      "loss": 0.0404,
+      "num_tokens": 47270204.0,
+      "reward": 3.915536617040634,
+      "reward_std": 0.6984670387580991,
+      "rewards/generate_all_rewards": 3.915536617040634,
+      "step": 3900
+    },
+    {
+      "clip_ratio": 0.005711669921875,
+      "completion_length": 1024.0,
+      "epoch": 0.658882402001668,
+      "grad_norm": 3.7694883514173156,
+      "kl": 2.70125,
+      "learning_rate": 1.5608562691131497e-06,
+      "loss": 0.047,
+      "num_tokens": 47877444.0,
+      "reward": 4.0312881523370745,
+      "reward_std": 0.6717341633141041,
+      "rewards/generate_all_rewards": 4.0312881523370745,
+      "step": 3950
+    },
+    {
+      "clip_ratio": 0.002728271484375,
+      "completion_length": 1024.0,
+      "epoch": 0.6672226855713094,
+      "grad_norm": 28.629124425382763,
+      "kl": 4.227265625,
+      "learning_rate": 1.5552960800667224e-06,
+      "loss": 0.0783,
+      "num_tokens": 48487624.0,
+      "reward": 3.6814190673828127,
+      "reward_std": 0.98351726539433,
+      "rewards/generate_all_rewards": 3.6814190673828127,
+      "step": 4000
+    },
+    {
+      "clip_ratio": 0.003985595703125,
+      "completion_length": 1024.0,
+      "epoch": 0.6755629691409508,
+      "grad_norm": 2.066191936598565,
+      "kl": 2.4734375,
+      "learning_rate": 1.5497358910202947e-06,
+      "loss": 0.0416,
+      "num_tokens": 49084624.0,
+      "reward": 3.4913946342468263,
+      "reward_std": 0.5357848100364209,
+      "rewards/generate_all_rewards": 3.4913946342468263,
+      "step": 4050
+    },
+    {
+      "clip_ratio": 0.0029345703125,
+      "completion_length": 1024.0,
+      "epoch": 0.6839032527105922,
+      "grad_norm": 2.29887054280825,
+      "kl": 1.980625,
+      "learning_rate": 1.544175701973867e-06,
+      "loss": 0.0351,
+      "num_tokens": 49697284.0,
+      "reward": 3.8557702827453615,
+      "reward_std": 0.6670322266966104,
+      "rewards/generate_all_rewards": 3.8557702827453615,
+      "step": 4100
+    },
+    {
+      "clip_ratio": 0.0025244278740137816,
+      "completion_length": 1023.4825,
+      "epoch": 0.6922435362802335,
+      "grad_norm": 2.273554199254762,
+      "kl": 2.0675,
+      "learning_rate": 1.5386155129274394e-06,
+      "loss": 0.0349,
+      "num_tokens": 50298965.0,
+      "reward": 3.7827609944343568,
+      "reward_std": 0.6611206940561533,
+      "rewards/generate_all_rewards": 3.7827609944343568,
+      "step": 4150
+    },
+    {
+      "clip_ratio": 0.003787841796875,
+      "completion_length": 1024.0,
+      "epoch": 0.700583819849875,
+      "grad_norm": 2.729061403557998,
+      "kl": 2.938828125,
+      "learning_rate": 1.5330553238810117e-06,
+      "loss": 0.0527,
+      "num_tokens": 50900509.0,
+      "reward": 3.8453090608119966,
+      "reward_std": 0.6590460070222616,
+      "rewards/generate_all_rewards": 3.8453090608119966,
+      "step": 4200
+    },
+    {
+      "clip_ratio": 0.003973388671875,
+      "completion_length": 1024.0,
+      "epoch": 0.7089241034195163,
+      "grad_norm": 2.6324042043612716,
+      "kl": 2.7425,
+      "learning_rate": 1.5274951348345844e-06,
+      "loss": 0.0492,
+      "num_tokens": 51503393.0,
+      "reward": 3.854447617530823,
+      "reward_std": 0.7062110313773156,
+      "rewards/generate_all_rewards": 3.854447617530823,
+      "step": 4250
+    },
+    {
+      "clip_ratio": 0.005185546875,
+      "completion_length": 1024.0,
+      "epoch": 0.7172643869891576,
+      "grad_norm": 10.054096567071998,
+      "kl": 2.940234375,
+      "learning_rate": 1.5219349457881568e-06,
+      "loss": 0.0519,
+      "num_tokens": 52114009.0,
+      "reward": 3.3904974472522738,
+      "reward_std": 0.5645231993496418,
+      "rewards/generate_all_rewards": 3.3904974472522738,
+      "step": 4300
+    },
+    {
+      "clip_ratio": 0.0033544921875,
+      "completion_length": 1024.0,
+      "epoch": 0.725604670558799,
+      "grad_norm": 2.216377326696927,
+      "kl": 2.188515625,
+      "learning_rate": 1.516374756741729e-06,
+      "loss": 0.038,
+      "num_tokens": 52716977.0,
+      "reward": 3.649118809700012,
+      "reward_std": 0.7559943076968193,
+      "rewards/generate_all_rewards": 3.649118809700012,
+      "step": 4350
+    },
+    {
+      "clip_ratio": 0.008282470703125,
+      "completion_length": 1024.0,
+      "epoch": 0.7339449541284404,
+      "grad_norm": 19.31919089805816,
+      "kl": 3.878515625,
+      "learning_rate": 1.5108145676953014e-06,
+      "loss": 0.0728,
+      "num_tokens": 53315957.0,
+      "reward": 3.508232383728027,
+      "reward_std": 0.46971126724034545,
+      "rewards/generate_all_rewards": 3.508232383728027,
+      "step": 4400
+    },
+    {
+      "clip_ratio": 0.004521484375,
+      "completion_length": 1024.0,
+      "epoch": 0.7422852376980817,
+      "grad_norm": 4.5243235708567155,
+      "kl": 2.6675,
+      "learning_rate": 1.5052543786488742e-06,
+      "loss": 0.0474,
+      "num_tokens": 53921229.0,
+      "reward": 3.100267617702484,
+      "reward_std": 0.3896624885499477,
+      "rewards/generate_all_rewards": 3.100267617702484,
+      "step": 4450
+    },
+    {
+      "clip_ratio": 0.003692626953125,
+      "completion_length": 1024.0,
+      "epoch": 0.7506255212677231,
+      "grad_norm": 2.355784905819362,
+      "kl": 4.606171875,
+      "learning_rate": 1.4996941896024465e-06,
+      "loss": 0.0863,
+      "num_tokens": 54520793.0,
+      "reward": 2.8280148708820345,
+      "reward_std": 1.2710373802483081,
+      "rewards/generate_all_rewards": 2.8280148708820345,
+      "step": 4500
+    },
+    {
+      "clip_ratio": 0.002569580078125,
+      "completion_length": 1024.0,
+      "epoch": 0.7589658048373644,
+      "grad_norm": 12.17255970472075,
+      "kl": 72.12828125,
+      "learning_rate": 1.4941340005560188e-06,
+      "loss": 1.435,
+      "num_tokens": 55124933.0,
+      "reward": 3.7596522867679596,
+      "reward_std": 0.4142132857814431,
+      "rewards/generate_all_rewards": 3.7596522867679596,
+      "step": 4550
+    },
+    {
+      "clip_ratio": 0.002452392578125,
+      "completion_length": 1024.0,
+      "epoch": 0.7673060884070059,
+      "grad_norm": 34.23055432567639,
+      "kl": 1.603515625,
+      "learning_rate": 1.4885738115095911e-06,
+      "loss": 0.0263,
+      "num_tokens": 55732813.0,
+      "reward": 3.4148662626743316,
+      "reward_std": 0.41280762093607337,
+      "rewards/generate_all_rewards": 3.4148662626743316,
+      "step": 4600
+    },
+    {
+      "clip_ratio": 0.00472900390625,
+      "completion_length": 1024.0,
+      "epoch": 0.7756463719766472,
+      "grad_norm": 7.328812046406036,
+      "kl": 2.65265625,
+      "learning_rate": 1.4830136224631637e-06,
+      "loss": 0.0485,
+      "num_tokens": 56338861.0,
+      "reward": 4.01952663898468,
+      "reward_std": 0.638951450586319,
+      "rewards/generate_all_rewards": 4.01952663898468,
+      "step": 4650
+    },
+    {
+      "clip_ratio": 0.003695068359375,
+      "completion_length": 1024.0,
+      "epoch": 0.7839866555462885,
+      "grad_norm": 11.307922352650342,
+      "kl": 2.89109375,
+      "learning_rate": 1.4774534334167362e-06,
+      "loss": 0.0528,
+      "num_tokens": 56942965.0,
+      "reward": 3.3853393924236297,
+      "reward_std": 0.7829919610917568,
+      "rewards/generate_all_rewards": 3.3853393924236297,
+      "step": 4700
+    },
+    {
+      "clip_ratio": 0.003349609375,
+      "completion_length": 1024.0,
+      "epoch": 0.79232693911593,
+      "grad_norm": 5.0280708015540645,
+      "kl": 2.42765625,
+      "learning_rate": 1.4718932443703085e-06,
+      "loss": 0.043,
+      "num_tokens": 57542753.0,
+      "reward": 3.185204845368862,
+      "reward_std": 0.6598466786742211,
+      "rewards/generate_all_rewards": 3.185204845368862,
+      "step": 4750
+    },
+    {
+      "clip_ratio": 0.00411865234375,
+      "completion_length": 1024.0,
+      "epoch": 0.8006672226855713,
+      "grad_norm": 3.835529281144975,
+      "kl": 2.94375,
+      "learning_rate": 1.4663330553238808e-06,
+      "loss": 0.0548,
+      "num_tokens": 58163693.0,
+      "reward": 4.450564415454864,
+      "reward_std": 1.0702211380563675,
+      "rewards/generate_all_rewards": 4.450564415454864,
+      "step": 4800
+    },
+    {
+      "clip_ratio": 0.002451171875,
+      "completion_length": 1024.0,
+      "epoch": 0.8090075062552127,
+      "grad_norm": 19.94946459088367,
+      "kl": 2.078359375,
+      "learning_rate": 1.4607728662774534e-06,
+      "loss": 0.0365,
+      "num_tokens": 58750017.0,
+      "reward": 2.8765233075618744,
+      "reward_std": 0.7672751601785421,
+      "rewards/generate_all_rewards": 2.8765233075618744,
+      "step": 4850
+    },
+    {
+      "clip_ratio": 0.00357421875,
+      "completion_length": 1024.0,
+      "epoch": 0.817347789824854,
+      "grad_norm": 17.047237053916263,
+      "kl": 3.17078125,
+      "learning_rate": 1.455212677231026e-06,
+      "loss": 0.056,
+      "num_tokens": 59348589.0,
+      "reward": 4.052628271579742,
+      "reward_std": 0.6698121561482548,
+      "rewards/generate_all_rewards": 4.052628271579742,
+      "step": 4900
+    },
+    {
+      "clip_ratio": 0.004931640625,
+      "completion_length": 1024.0,
+      "epoch": 0.8256880733944955,
+      "grad_norm": 4.826417211060371,
+      "kl": 2.607734375,
+      "learning_rate": 1.4496524881845982e-06,
+      "loss": 0.0451,
+      "num_tokens": 59954733.0,
+      "reward": 2.8341499626636506,
+      "reward_std": 0.46431461662054063,
+      "rewards/generate_all_rewards": 2.8341499626636506,
+      "step": 4950
+    },
+    {
+      "clip_ratio": 0.005946044921875,
+      "completion_length": 1024.0,
+      "epoch": 0.8340283569641368,
+      "grad_norm": 10.291202916786846,
+      "kl": 3.267421875,
+      "learning_rate": 1.4440922991381705e-06,
+      "loss": 0.06,
+      "num_tokens": 60574085.0,
+      "reward": 3.4563196295499803,
+      "reward_std": 0.794042921513319,
+      "rewards/generate_all_rewards": 3.4563196295499803,
+      "step": 5000
+    },
+    {
+      "clip_ratio": 0.005137939453125,
+      "completion_length": 1024.0,
+      "epoch": 0.8423686405337781,
+      "grad_norm": 31.864028882209055,
+      "kl": 2.831640625,
+      "learning_rate": 1.438532110091743e-06,
+      "loss": 0.0516,
+      "num_tokens": 61172865.0,
+      "reward": 2.721455451250076,
+      "reward_std": 0.5221429407224059,
+      "rewards/generate_all_rewards": 2.721455451250076,
+      "step": 5050
+    },
+    {
+      "clip_ratio": 0.00337646484375,
+      "completion_length": 1024.0,
+      "epoch": 0.8507089241034195,
+      "grad_norm": 14.39794633355936,
+      "kl": 1.65421875,
+      "learning_rate": 1.4329719210453156e-06,
+      "loss": 0.028,
+      "num_tokens": 61761873.0,
+      "reward": 4.077827769517898,
+      "reward_std": 0.6053008568286896,
+      "rewards/generate_all_rewards": 4.077827769517898,
+      "step": 5100
+    },
+    {
+      "clip_ratio": 0.003028564453125,
+      "completion_length": 1024.0,
+      "epoch": 0.8590492076730609,
+      "grad_norm": 6.864401695816703,
+      "kl": 2.664453125,
+      "learning_rate": 1.427411731998888e-06,
+      "loss": 0.0474,
+      "num_tokens": 62374005.0,
+      "reward": 3.6210562777519226,
+      "reward_std": 0.3803251222521067,
+      "rewards/generate_all_rewards": 3.6210562777519226,
+      "step": 5150
+    },
+    {
+      "clip_ratio": 0.00245361328125,
+      "completion_length": 1024.0,
+      "epoch": 0.8673894912427023,
+      "grad_norm": 3.555441991695044,
+      "kl": 2.218125,
+      "learning_rate": 1.4218515429524603e-06,
+      "loss": 0.0374,
+      "num_tokens": 62985201.0,
+      "reward": 3.7183710026741026,
+      "reward_std": 0.6048373529314994,
+      "rewards/generate_all_rewards": 3.7183710026741026,
+      "step": 5200
+    },
+    {
+      "clip_ratio": 0.00371826171875,
+      "completion_length": 1024.0,
+      "epoch": 0.8757297748123436,
+      "grad_norm": 6.813815164137024,
+      "kl": 3.165859375,
+      "learning_rate": 1.4162913539060328e-06,
+      "loss": 0.0591,
+      "num_tokens": 63587933.0,
+      "reward": 3.470691123008728,
+      "reward_std": 0.8655337832123041,
+      "rewards/generate_all_rewards": 3.470691123008728,
+      "step": 5250
+    },
+    {
+      "clip_ratio": 0.00478271484375,
+      "completion_length": 1024.0,
+      "epoch": 0.8840700583819849,
+      "grad_norm": 3.837611227178622,
+      "kl": 2.496484375,
+      "learning_rate": 1.4107311648596051e-06,
+      "loss": 0.0445,
+      "num_tokens": 64193125.0,
+      "reward": 2.8873752689361574,
+      "reward_std": 0.6322074158862233,
+      "rewards/generate_all_rewards": 2.8873752689361574,
+      "step": 5300
+    },
+    {
+      "clip_ratio": 0.00561767578125,
+      "completion_length": 1024.0,
+      "epoch": 0.8924103419516264,
+      "grad_norm": 10.223757483016007,
+      "kl": 5.140546875,
+      "learning_rate": 1.4051709758131776e-06,
+      "loss": 0.0962,
+      "num_tokens": 64791869.0,
+      "reward": 2.5429132187366488,
+      "reward_std": 0.906138878762722,
+      "rewards/generate_all_rewards": 2.5429132187366488,
+      "step": 5350
+    },
+    {
+      "clip_ratio": 0.00482421875,
+      "completion_length": 1024.0,
+      "epoch": 0.9007506255212677,
+      "grad_norm": 14.701033479824986,
+      "kl": 4.24796875,
+      "learning_rate": 1.39961078676675e-06,
+      "loss": 0.0798,
+      "num_tokens": 65400309.0,
+      "reward": 2.5115936332941056,
+      "reward_std": 0.8834956586360931,
+      "rewards/generate_all_rewards": 2.5115936332941056,
+      "step": 5400
+    },
+    {
+      "clip_ratio": 0.00357421875,
+      "completion_length": 1024.0,
+      "epoch": 0.9090909090909091,
+      "grad_norm": 9.892821812806087,
+      "kl": 5.386484375,
+      "learning_rate": 1.3940505977203225e-06,
+      "loss": 0.1015,
+      "num_tokens": 65993385.0,
+      "reward": 1.6406327021121978,
+      "reward_std": 1.3033610412478447,
+      "rewards/generate_all_rewards": 1.6406327021121978,
+      "step": 5450
+    },
+    {
+      "clip_ratio": 0.003983154296875,
+      "completion_length": 1024.0,
+      "epoch": 0.9174311926605505,
+      "grad_norm": 3.900132285131672,
+      "kl": 2.206015625,
+      "learning_rate": 1.3884904086738948e-06,
+      "loss": 0.0387,
+      "num_tokens": 66600541.0,
+      "reward": 3.18016503572464,
+      "reward_std": 0.8799462950974702,
+      "rewards/generate_all_rewards": 3.18016503572464,
+      "step": 5500
+    },
+    {
+      "clip_ratio": 0.00421142578125,
+      "completion_length": 1024.0,
+      "epoch": 0.9257714762301918,
+      "grad_norm": 2.87190664608335,
+      "kl": 2.29390625,
+      "learning_rate": 1.3829302196274674e-06,
+      "loss": 0.0432,
+      "num_tokens": 67211765.0,
+      "reward": 3.1571250998973848,
+      "reward_std": 0.6987396457791328,
+      "rewards/generate_all_rewards": 3.1571250998973848,
+      "step": 5550
+    },
+    {
+      "clip_ratio": 0.00346923828125,
+      "completion_length": 1024.0,
+      "epoch": 0.9341117597998332,
+      "grad_norm": 4.817047536540244,
+      "kl": 2.0128125,
+      "learning_rate": 1.3773700305810397e-06,
+      "loss": 0.0367,
+      "num_tokens": 67836585.0,
+      "reward": 4.290312564373016,
+      "reward_std": 0.7204281195998192,
+      "rewards/generate_all_rewards": 4.290312564373016,
+      "step": 5600
+    },
+    {
+      "clip_ratio": 0.00431640625,
+      "completion_length": 1024.0,
+      "epoch": 0.9424520433694745,
+      "grad_norm": 14.566705655714621,
+      "kl": 2.095234375,
+      "learning_rate": 1.3718098415346122e-06,
+      "loss": 0.0362,
+      "num_tokens": 68447877.0,
+      "reward": 4.023037674427033,
+      "reward_std": 0.45142842307686804,
+      "rewards/generate_all_rewards": 4.023037674427033,
+      "step": 5650
+    },
+    {
+      "clip_ratio": 0.006263427734375,
+      "completion_length": 1024.0,
+      "epoch": 0.950792326939116,
+      "grad_norm": 7.824949124358837,
+      "kl": 3.860703125,
+      "learning_rate": 1.3662496524881845e-06,
+      "loss": 0.0707,
+      "num_tokens": 69049573.0,
+      "reward": 3.0012189817428587,
+      "reward_std": 0.3978773768246174,
+      "rewards/generate_all_rewards": 3.0012189817428587,
+      "step": 5700
+    },
+    {
+      "clip_ratio": 0.002972412109375,
+      "completion_length": 1024.0,
+      "epoch": 0.9591326105087573,
+      "grad_norm": 3.1286184419710303,
+      "kl": 3.4021875,
+      "learning_rate": 1.3606894634417569e-06,
+      "loss": 0.0635,
+      "num_tokens": 69639193.0,
+      "reward": 3.187228103876114,
+      "reward_std": 0.8414063957333565,
+      "rewards/generate_all_rewards": 3.187228103876114,
+      "step": 5750
+    },
+    {
+      "clip_ratio": 0.004266357421875,
+      "completion_length": 1024.0,
+      "epoch": 0.9674728940783986,
+      "grad_norm": 19.314033277311875,
+      "kl": 2.820625,
+      "learning_rate": 1.3551292743953294e-06,
+      "loss": 0.0517,
+      "num_tokens": 70247585.0,
+      "reward": 3.7847342348098754,
+      "reward_std": 0.6664008083939552,
+      "rewards/generate_all_rewards": 3.7847342348098754,
+      "step": 5800
+    },
+    {
+      "clip_ratio": 0.00450439453125,
+      "completion_length": 1024.0,
+      "epoch": 0.97581317764804,
+      "grad_norm": 7.481107247218354,
+      "kl": 2.52328125,
+      "learning_rate": 1.349569085348902e-06,
+      "loss": 0.0467,
+      "num_tokens": 70850869.0,
+      "reward": 3.71783961057663,
+      "reward_std": 0.6372227993607521,
+      "rewards/generate_all_rewards": 3.71783961057663,
+      "step": 5850
+    },
+    {
+      "clip_ratio": 0.003583984375,
+      "completion_length": 1024.0,
+      "epoch": 0.9841534612176814,
+      "grad_norm": 13.485541281738868,
+      "kl": 2.18,
+      "learning_rate": 1.3440088963024742e-06,
+      "loss": 0.0388,
+      "num_tokens": 71456689.0,
+      "reward": 3.874968693852425,
+      "reward_std": 0.7574485129117966,
+      "rewards/generate_all_rewards": 3.874968693852425,
+      "step": 5900
+    },
+    {
+      "clip_ratio": 0.003751220703125,
+      "completion_length": 1024.0,
+      "epoch": 0.9924937447873228,
+      "grad_norm": 4.232392646106402,
+      "kl": 2.265625,
+      "learning_rate": 1.3384487072560466e-06,
+      "loss": 0.0411,
+      "num_tokens": 72057601.0,
+      "reward": 3.2541480442881583,
+      "reward_std": 0.5004438938200474,
+      "rewards/generate_all_rewards": 3.2541480442881583,
+      "step": 5950
+    },
+    {
+      "clip_ratio": 0.004661865234375,
+      "completion_length": 1024.0,
+      "epoch": 1.001000834028357,
+      "grad_norm": 4.190575159866155,
+      "kl": 2.268203125,
+      "learning_rate": 1.332888518209619e-06,
+      "loss": 0.041,
+      "num_tokens": 72657241.0,
+      "reward": 4.536292546391487,
+      "reward_std": 0.7764492811262608,
+      "rewards/generate_all_rewards": 4.536292546391487,
+      "step": 6000
+    },
+    {
+      "clip_ratio": 0.003243408203125,
+      "completion_length": 1024.0,
+      "epoch": 1.0093411175979983,
+      "grad_norm": 2.73669917409696,
+      "kl": 2.99953125,
+      "learning_rate": 1.3273283291631916e-06,
+      "loss": 0.0554,
+      "num_tokens": 73258405.0,
+      "reward": 3.7951004153490064,
+      "reward_std": 0.5186809635162354,
+      "rewards/generate_all_rewards": 3.7951004153490064,
+      "step": 6050
+    },
+    {
+      "clip_ratio": 0.004434814453125,
+      "completion_length": 1024.0,
+      "epoch": 1.0176814011676396,
+      "grad_norm": 3.2206800159419102,
+      "kl": 3.396875,
+      "learning_rate": 1.321768140116764e-06,
+      "loss": 0.0631,
+      "num_tokens": 73852137.0,
+      "reward": 2.9864656680822375,
+      "reward_std": 0.6515207149833441,
+      "rewards/generate_all_rewards": 2.9864656680822375,
+      "step": 6100
+    },
+    {
+      "clip_ratio": 0.0039599609375,
+      "completion_length": 1024.0,
+      "epoch": 1.0260216847372812,
+      "grad_norm": 5.2073505200985615,
+      "kl": 3.2271875,
+      "learning_rate": 1.3162079510703363e-06,
+      "loss": 0.0602,
+      "num_tokens": 74469161.0,
+      "reward": 3.73267055273056,
+      "reward_std": 0.5677018699049949,
+      "rewards/generate_all_rewards": 3.73267055273056,
+      "step": 6150
+    },
+    {
+      "clip_ratio": 0.002672119140625,
+      "completion_length": 1024.0,
+      "epoch": 1.0343619683069225,
+      "grad_norm": 5.539030446663812,
+      "kl": 3.00671875,
+      "learning_rate": 1.3106477620239086e-06,
+      "loss": 0.0557,
+      "num_tokens": 75077049.0,
+      "reward": 3.6774249446392058,
+      "reward_std": 0.8043338760733605,
+      "rewards/generate_all_rewards": 3.6774249446392058,
+      "step": 6200
+    },
+    {
+      "clip_ratio": 0.004927978515625,
+      "completion_length": 1024.0,
+      "epoch": 1.0427022518765638,
+      "grad_norm": 2.7534484045841787,
+      "kl": 3.223828125,
+      "learning_rate": 1.3050875729774811e-06,
+      "loss": 0.0604,
+      "num_tokens": 75687349.0,
+      "reward": 3.7158334070444106,
+      "reward_std": 0.658657222688198,
+      "rewards/generate_all_rewards": 3.7158334070444106,
+      "step": 6250
+    },
+    {
+      "clip_ratio": 0.005045166015625,
+      "completion_length": 1024.0,
+      "epoch": 1.0510425354462052,
+      "grad_norm": 4.953550380847901,
+      "kl": 2.543203125,
+      "learning_rate": 1.2995273839310537e-06,
+      "loss": 0.0472,
+      "num_tokens": 76300569.0,
+      "reward": 4.2147672295570375,
+      "reward_std": 0.8778787985444069,
+      "rewards/generate_all_rewards": 4.2147672295570375,
+      "step": 6300
+    },
+    {
+      "clip_ratio": 0.003118896484375,
+      "completion_length": 1024.0,
+      "epoch": 1.0593828190158465,
+      "grad_norm": 12.367784692060768,
+      "kl": 2.41859375,
+      "learning_rate": 1.293967194884626e-06,
+      "loss": 0.043,
+      "num_tokens": 76904045.0,
+      "reward": 3.5026879012584686,
+      "reward_std": 0.6680633321404457,
+      "rewards/generate_all_rewards": 3.5026879012584686,
+      "step": 6350
+    },
+    {
+      "clip_ratio": 0.00299072265625,
+      "completion_length": 1024.0,
+      "epoch": 1.0677231025854879,
+      "grad_norm": 58.121185792290575,
+      "kl": 2.092421875,
+      "learning_rate": 1.2884070058381983e-06,
+      "loss": 0.0389,
+      "num_tokens": 77527197.0,
+      "reward": 4.4785698533058165,
+      "reward_std": 0.33464464247226716,
+      "rewards/generate_all_rewards": 4.4785698533058165,
+      "step": 6400
+    },
+    {
+      "clip_ratio": 0.005118408203125,
+      "completion_length": 1024.0,
+      "epoch": 1.0760633861551292,
+      "grad_norm": 24.287293596006418,
+      "kl": 3.210234375,
+      "learning_rate": 1.2828468167917708e-06,
+      "loss": 0.0578,
+      "num_tokens": 78134137.0,
+      "reward": 3.7603088819980623,
+      "reward_std": 0.4170807982981205,
+      "rewards/generate_all_rewards": 3.7603088819980623,
+      "step": 6450
+    },
+    {
+      "clip_ratio": 0.003282470703125,
+      "completion_length": 1024.0,
+      "epoch": 1.0844036697247708,
+      "grad_norm": 15.397400516678681,
+      "kl": 2.402890625,
+      "learning_rate": 1.2772866277453434e-06,
+      "loss": 0.0428,
+      "num_tokens": 78750593.0,
+      "reward": 4.395290724039078,
+      "reward_std": 0.5200403224676847,
+      "rewards/generate_all_rewards": 4.395290724039078,
+      "step": 6500
+    },
+    {
+      "clip_ratio": 0.0048388671875,
+      "completion_length": 1024.0,
+      "epoch": 1.092743953294412,
+      "grad_norm": 4.963221819052845,
+      "kl": 5.079765625,
+      "learning_rate": 1.2717264386989157e-06,
+      "loss": 0.0951,
+      "num_tokens": 79360801.0,
+      "reward": 3.146202702522278,
+      "reward_std": 0.6797874164581299,
+      "rewards/generate_all_rewards": 3.146202702522278,
+      "step": 6550
+    },
+    {
+      "clip_ratio": 0.003487548828125,
+      "completion_length": 1024.0,
+      "epoch": 1.1010842368640534,
+      "grad_norm": 15.023192969035028,
+      "kl": 3.15125,
+      "learning_rate": 1.266166249652488e-06,
+      "loss": 0.0587,
+      "num_tokens": 79952117.0,
+      "reward": 3.741974139213562,
+      "reward_std": 0.4647814616560936,
+      "rewards/generate_all_rewards": 3.741974139213562,
+      "step": 6600
+    },
+    {
+      "clip_ratio": 0.004453125,
+      "completion_length": 1024.0,
+      "epoch": 1.1094245204336948,
+      "grad_norm": 9.103985085349178,
+      "kl": 2.305625,
+      "learning_rate": 1.2606060606060606e-06,
+      "loss": 0.0413,
+      "num_tokens": 80552069.0,
+      "reward": 4.223839626312256,
+      "reward_std": 0.5068751226365567,
+      "rewards/generate_all_rewards": 4.223839626312256,
+      "step": 6650
+    },
+    {
+      "clip_ratio": 0.003953857421875,
+      "completion_length": 1024.0,
+      "epoch": 1.117764804003336,
+      "grad_norm": 3.2493747477518973,
+      "kl": 2.931640625,
+      "learning_rate": 1.255045871559633e-06,
+      "loss": 0.0535,
+      "num_tokens": 81155465.0,
+      "reward": 3.6686401844024656,
+      "reward_std": 0.7149755907058716,
+      "rewards/generate_all_rewards": 3.6686401844024656,
+      "step": 6700
+    },
+    {
+      "clip_ratio": 0.00394287109375,
+      "completion_length": 1024.0,
+      "epoch": 1.1261050875729774,
+      "grad_norm": 3.186846652783417,
+      "kl": 3.184296875,
+      "learning_rate": 1.2494856825132054e-06,
+      "loss": 0.0584,
+      "num_tokens": 81751105.0,
+      "reward": 4.3162576973438265,
+      "reward_std": 0.45037852857261895,
+      "rewards/generate_all_rewards": 4.3162576973438265,
+      "step": 6750
+    },
+    {
+      "clip_ratio": 0.003707275390625,
+      "completion_length": 1024.0,
+      "epoch": 1.1344453711426188,
+      "grad_norm": 3.04261280846647,
+      "kl": 2.888828125,
+      "learning_rate": 1.2439254934667777e-06,
+      "loss": 0.0521,
+      "num_tokens": 82351093.0,
+      "reward": 3.8912365996837615,
+      "reward_std": 0.35723258018493653,
+      "rewards/generate_all_rewards": 3.8912365996837615,
+      "step": 6800
+    },
+    {
+      "clip_ratio": 0.00365478515625,
+      "completion_length": 1024.0,
+      "epoch": 1.1427856547122601,
+      "grad_norm": 4.310881136848889,
+      "kl": 2.07828125,
+      "learning_rate": 1.23836530442035e-06,
+      "loss": 0.0358,
+      "num_tokens": 82958733.0,
+      "reward": 4.600189430117607,
+      "reward_std": 0.4825837394595146,
+      "rewards/generate_all_rewards": 4.600189430117607,
+      "step": 6850
+    },
+    {
+      "clip_ratio": 0.005318603515625,
+      "completion_length": 1024.0,
+      "epoch": 1.1511259382819015,
+      "grad_norm": 5.6721210845224075,
+      "kl": 3.6896875,
+      "learning_rate": 1.2328051153739228e-06,
+      "loss": 0.067,
+      "num_tokens": 83556809.0,
+      "reward": 3.0905766177177427,
+      "reward_std": 0.7636496469378471,
+      "rewards/generate_all_rewards": 3.0905766177177427,
+      "step": 6900
+    },
+    {
+      "clip_ratio": 0.003841552734375,
+      "completion_length": 1024.0,
+      "epoch": 1.159466221851543,
+      "grad_norm": 35.64711800695698,
+      "kl": 2.9359375,
+      "learning_rate": 1.2272449263274951e-06,
+      "loss": 0.0535,
+      "num_tokens": 84165225.0,
+      "reward": 3.255465921163559,
+      "reward_std": 1.003816493228078,
+      "rewards/generate_all_rewards": 3.255465921163559,
+      "step": 6950
+    },
+    {
+      "clip_ratio": 0.00360595703125,
+      "completion_length": 1024.0,
+      "epoch": 1.1678065054211844,
+      "grad_norm": 5.510807439460425,
+      "kl": 2.919296875,
+      "learning_rate": 1.2216847372810674e-06,
+      "loss": 0.0537,
+      "num_tokens": 84781149.0,
+      "reward": 3.9430871558189393,
+      "reward_std": 0.46672826692461966,
+      "rewards/generate_all_rewards": 3.9430871558189393,
+      "step": 7000
+    }
+  ],
+  "logging_steps": 50,
+  "max_steps": 17985,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 3,
+  "save_steps": 1000,
+  "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": 4,
+  "trial_name": null,
+  "trial_params": null
+}

training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ced00c0dc42390164944b48c1567a5cf07bbb0603e64c349f2106e674c6d7365
+size 7825

zero_to_fp32.py

@@ -0,0 +1,760 @@
+#!/usr/bin/env python
+
+# Copyright (c) Microsoft Corporation.
+# SPDX-License-Identifier: Apache-2.0
+
+# DeepSpeed Team
+
+# This script extracts fp32 consolidated weights from a zero 1, 2 and 3 DeepSpeed checkpoints. It gets
+# copied into the top level checkpoint dir, so the user can easily do the conversion at any point in
+# the future. Once extracted, the weights don't require DeepSpeed and can be used in any
+# application.
+#
+# example:
+#   python zero_to_fp32.py . output_dir/
+#   or
+#   python zero_to_fp32.py . output_dir/ --safe_serialization
+
+import argparse
+import torch
+import glob
+import math
+import os
+import re
+import gc
+import json
+import numpy as np
+from tqdm import tqdm
+from collections import OrderedDict
+from dataclasses import dataclass
+
+# while this script doesn't use deepspeed to recover data, since the checkpoints are pickled with
+# DeepSpeed data structures it has to be available in the current python environment.
+from deepspeed.utils import logger
+from deepspeed.checkpoint.constants import (DS_VERSION, OPTIMIZER_STATE_DICT, SINGLE_PARTITION_OF_FP32_GROUPS,
+                                            FP32_FLAT_GROUPS, ZERO_STAGE, PARTITION_COUNT, PARAM_SHAPES, BUFFER_NAMES,
+                                            FROZEN_PARAM_SHAPES, FROZEN_PARAM_FRAGMENTS)
+
+
+@dataclass
+class zero_model_state:
+    buffers: dict()
+    param_shapes: dict()
+    shared_params: list
+    ds_version: int
+    frozen_param_shapes: dict()
+    frozen_param_fragments: dict()
+
+
+debug = 0
+
+# load to cpu
+device = torch.device('cpu')
+
+
+def atoi(text):
+    return int(text) if text.isdigit() else text
+
+
+def natural_keys(text):
+    '''
+    alist.sort(key=natural_keys) sorts in human order
+    http://nedbatchelder.com/blog/200712/human_sorting.html
+    (See Toothy's implementation in the comments)
+    '''
+    return [atoi(c) for c in re.split(r'(\d+)', text)]
+
+
+def get_model_state_file(checkpoint_dir, zero_stage):
+    if not os.path.isdir(checkpoint_dir):
+        raise FileNotFoundError(f"Directory '{checkpoint_dir}' doesn't exist")
+
+    # there should be only one file
+    if zero_stage <= 2:
+        file = os.path.join(checkpoint_dir, "mp_rank_00_model_states.pt")
+    elif zero_stage == 3:
+        file = os.path.join(checkpoint_dir, "zero_pp_rank_0_mp_rank_00_model_states.pt")
+
+    if not os.path.exists(file):
+        raise FileNotFoundError(f"can't find model states file at '{file}'")
+
+    return file
+
+
+def get_checkpoint_files(checkpoint_dir, glob_pattern):
+    # XXX: need to test that this simple glob rule works for multi-node setup too
+    ckpt_files = sorted(glob.glob(os.path.join(checkpoint_dir, glob_pattern)), key=natural_keys)
+
+    if len(ckpt_files) == 0:
+        raise FileNotFoundError(f"can't find {glob_pattern} files in directory '{checkpoint_dir}'")
+
+    return ckpt_files
+
+
+def get_optim_files(checkpoint_dir):
+    return get_checkpoint_files(checkpoint_dir, "*_optim_states.pt")
+
+
+def get_model_state_files(checkpoint_dir):
+    return get_checkpoint_files(checkpoint_dir, "*_model_states.pt")
+
+
+def parse_model_states(files):
+    zero_model_states = []
+    for file in files:
+        state_dict = torch.load(file, map_location=device, weights_only=False)
+
+        if BUFFER_NAMES not in state_dict:
+            raise ValueError(f"{file} is not a model state checkpoint")
+        buffer_names = state_dict[BUFFER_NAMES]
+        if debug:
+            print("Found buffers:", buffer_names)
+
+        # recover just the buffers while restoring them to fp32 if they were saved in fp16
+        buffers = {k: v.float() for k, v in state_dict["module"].items() if k in buffer_names}
+        param_shapes = state_dict[PARAM_SHAPES]
+
+        # collect parameters that are included in param_shapes
+        param_names = []
+        for s in param_shapes:
+            for name in s.keys():
+                param_names.append(name)
+
+        # update with frozen parameters
+        frozen_param_shapes = state_dict.get(FROZEN_PARAM_SHAPES, None)
+        if frozen_param_shapes is not None:
+            if debug:
+                print(f"Found frozen_param_shapes: {frozen_param_shapes}")
+            param_names += list(frozen_param_shapes.keys())
+
+        # handle shared params
+        shared_params = [[k, v] for k, v in state_dict["shared_params"].items()]
+
+        ds_version = state_dict.get(DS_VERSION, None)
+
+        frozen_param_fragments = state_dict.get(FROZEN_PARAM_FRAGMENTS, None)
+
+        z_model_state = zero_model_state(buffers=buffers,
+                                         param_shapes=param_shapes,
+                                         shared_params=shared_params,
+                                         ds_version=ds_version,
+                                         frozen_param_shapes=frozen_param_shapes,
+                                         frozen_param_fragments=frozen_param_fragments)
+        zero_model_states.append(z_model_state)
+
+    return zero_model_states
+
+
+def parse_optim_states(files, ds_checkpoint_dir):
+    total_files = len(files)
+    state_dicts = []
+    for f in tqdm(files, desc='Loading checkpoint shards'):
+        state_dict = torch.load(f, map_location=device, mmap=True, weights_only=False)
+        # immediately discard the potentially huge 2 optimizer states as we only care for fp32 master weights
+        # and also handle the case where it was already removed by another helper script
+        state_dict["optimizer_state_dict"].pop("optimizer_state_dict", None)
+        state_dicts.append(state_dict)
+
+    if not ZERO_STAGE in state_dicts[0][OPTIMIZER_STATE_DICT]:
+        raise ValueError(f"{files[0]} is not a zero checkpoint")
+    zero_stage = state_dicts[0][OPTIMIZER_STATE_DICT][ZERO_STAGE]
+    world_size = state_dicts[0][OPTIMIZER_STATE_DICT][PARTITION_COUNT]
+
+    # For ZeRO-2 each param group can have different partition_count as data parallelism for expert
+    # parameters can be different from data parallelism for non-expert parameters. So we can just
+    # use the max of the partition_count to get the dp world_size.
+
+    if type(world_size) is list:
+        world_size = max(world_size)
+
+    if world_size != total_files:
+        raise ValueError(
+            f"Expected {world_size} of '*_optim_states.pt' under '{ds_checkpoint_dir}' but found {total_files} files. "
+            "Possibly due to an overwrite of an old checkpoint, or a checkpoint didn't get saved by one or more processes."
+        )
+
+    # the groups are named differently in each stage
+    if zero_stage <= 2:
+        fp32_groups_key = SINGLE_PARTITION_OF_FP32_GROUPS
+    elif zero_stage == 3:
+        fp32_groups_key = FP32_FLAT_GROUPS
+    else:
+        raise ValueError(f"unknown zero stage {zero_stage}")
+
+    fp32_flat_groups = [state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key] for i in range(len(state_dicts))]
+    return zero_stage, world_size, fp32_flat_groups
+
+
+def _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters):
+    """
+    Returns fp32 state_dict reconstructed from ds checkpoint
+
+    Args:
+        - ``ds_checkpoint_dir``: path to the deepspeed checkpoint folder (where the optimizer files are)
+
+    """
+    print(f"Processing zero checkpoint '{ds_checkpoint_dir}'")
+
+    optim_files = get_optim_files(ds_checkpoint_dir)
+    zero_stage, world_size, fp32_flat_groups = parse_optim_states(optim_files, ds_checkpoint_dir)
+    print(f"Detected checkpoint of type zero stage {zero_stage}, world_size: {world_size}")
+
+    model_files = get_model_state_files(ds_checkpoint_dir)
+
+    zero_model_states = parse_model_states(model_files)
+    print(f'Parsing checkpoint created by deepspeed=={zero_model_states[0].ds_version}')
+
+    if zero_stage <= 2:
+        return _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                                          exclude_frozen_parameters)
+    elif zero_stage == 3:
+        return _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                                          exclude_frozen_parameters)
+
+
+def _zero2_merge_frozen_params(state_dict, zero_model_states):
+    if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
+        return
+
+    frozen_param_shapes = zero_model_states[0].frozen_param_shapes
+    frozen_param_fragments = zero_model_states[0].frozen_param_fragments
+
+    if debug:
+        num_elem = sum(s.numel() for s in frozen_param_shapes.values())
+        print(f'rank 0: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
+
+        wanted_params = len(frozen_param_shapes)
+        wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
+        avail_numel = sum([p.numel() for p in frozen_param_fragments.values()])
+        print(f'Frozen params: Have {avail_numel} numels to process.')
+        print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
+
+    total_params = 0
+    total_numel = 0
+    for name, shape in frozen_param_shapes.items():
+        total_params += 1
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+
+        state_dict[name] = frozen_param_fragments[name]
+
+        if debug:
+            print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
+
+    print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _has_callable(obj, fn):
+    attr = getattr(obj, fn, None)
+    return callable(attr)
+
+
+def _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
+    param_shapes = zero_model_states[0].param_shapes
+
+    # Reconstruction protocol:
+    #
+    # XXX: document this
+
+    if debug:
+        for i in range(world_size):
+            for j in range(len(fp32_flat_groups[0])):
+                print(f"{FP32_FLAT_GROUPS}[{i}][{j}].shape={fp32_flat_groups[i][j].shape}")
+
+    # XXX: memory usage doubles here (zero2)
+    num_param_groups = len(fp32_flat_groups[0])
+    merged_single_partition_of_fp32_groups = []
+    for i in range(num_param_groups):
+        merged_partitions = [sd[i] for sd in fp32_flat_groups]
+        full_single_fp32_vector = torch.cat(merged_partitions, 0)
+        merged_single_partition_of_fp32_groups.append(full_single_fp32_vector)
+    avail_numel = sum(
+        [full_single_fp32_vector.numel() for full_single_fp32_vector in merged_single_partition_of_fp32_groups])
+
+    if debug:
+        wanted_params = sum([len(shapes) for shapes in param_shapes])
+        wanted_numel = sum([sum(shape.numel() for shape in shapes.values()) for shapes in param_shapes])
+        # not asserting if there is a mismatch due to possible padding
+        print(f"Have {avail_numel} numels to process.")
+        print(f"Need {wanted_numel} numels in {wanted_params} params.")
+
+    # params
+    # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
+    # out-of-core computing solution
+    total_numel = 0
+    total_params = 0
+    for shapes, full_single_fp32_vector in zip(param_shapes, merged_single_partition_of_fp32_groups):
+        offset = 0
+        avail_numel = full_single_fp32_vector.numel()
+        for name, shape in shapes.items():
+
+            unpartitioned_numel = shape.numel() if _has_callable(shape, 'numel') else math.prod(shape)
+            total_numel += unpartitioned_numel
+            total_params += 1
+
+            if debug:
+                print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
+            state_dict[name] = full_single_fp32_vector.narrow(0, offset, unpartitioned_numel).view(shape)
+            offset += unpartitioned_numel
+
+        # Z2 started to align to 2*world_size to improve nccl performance. Therefore both offset and
+        # avail_numel can differ by anywhere between 0..2*world_size. Due to two unrelated complex
+        # paddings performed in the code it's almost impossible to predict the exact numbers w/o the
+        # live optimizer object, so we are checking that the numbers are within the right range
+        align_to = 2 * world_size
+
+        def zero2_align(x):
+            return align_to * math.ceil(x / align_to)
+
+        if debug:
+            print(f"original offset={offset}, avail_numel={avail_numel}")
+
+        offset = zero2_align(offset)
+        avail_numel = zero2_align(avail_numel)
+
+        if debug:
+            print(f"aligned  offset={offset}, avail_numel={avail_numel}")
+
+        # Sanity check
+        if offset != avail_numel:
+            raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
+
+    print(f"Reconstructed fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                               exclude_frozen_parameters):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    if not exclude_frozen_parameters:
+        _zero2_merge_frozen_params(state_dict, zero_model_states)
+
+    _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
+
+    # recover shared parameters
+    for pair in zero_model_states[0].shared_params:
+        if pair[1] in state_dict:
+            state_dict[pair[0]] = state_dict[pair[1]]
+
+    return state_dict
+
+
+def zero3_partitioned_param_info(unpartitioned_numel, world_size):
+    remainder = unpartitioned_numel % world_size
+    padding_numel = (world_size - remainder) if remainder else 0
+    partitioned_numel = math.ceil(unpartitioned_numel / world_size)
+    return partitioned_numel, padding_numel
+
+
+def _zero3_merge_frozen_params(state_dict, world_size, zero_model_states):
+    if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
+        return
+
+    if debug:
+        for i in range(world_size):
+            num_elem = sum(s.numel() for s in zero_model_states[i].frozen_param_fragments.values())
+            print(f'rank {i}: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
+
+        frozen_param_shapes = zero_model_states[0].frozen_param_shapes
+        wanted_params = len(frozen_param_shapes)
+        wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
+        avail_numel = sum([p.numel() for p in zero_model_states[0].frozen_param_fragments.values()]) * world_size
+        print(f'Frozen params: Have {avail_numel} numels to process.')
+        print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
+
+    total_params = 0
+    total_numel = 0
+    for name, shape in zero_model_states[0].frozen_param_shapes.items():
+        total_params += 1
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+
+        param_frags = tuple(model_state.frozen_param_fragments[name] for model_state in zero_model_states)
+        state_dict[name] = torch.cat(param_frags, 0).narrow(0, 0, unpartitioned_numel).view(shape)
+
+        partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
+
+        if debug:
+            print(
+                f"Frozen params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
+            )
+
+    print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
+
+
+class GatheredTensor:
+    """
+    A pseudo tensor that collects partitioned weights.
+    It is more memory efficient when there are multiple groups.
+    """
+
+    def __init__(self, flat_groups, flat_groups_offset, offset, partitioned_numel, shape):
+        self.flat_groups = flat_groups
+        self.flat_groups_offset = flat_groups_offset
+        self.offset = offset
+        self.partitioned_numel = partitioned_numel
+        self.shape = shape
+        self.dtype = self.flat_groups[0][0].dtype
+
+    def contiguous(self):
+        """
+        Merge partitioned weights from flat_groups into a single tensor.
+        """
+        end_idx = self.offset + self.partitioned_numel
+        world_size = len(self.flat_groups)
+        pad_flat_param_chunks = []
+
+        for rank_i in range(world_size):
+            # for each rank, we need to collect weights from related group/groups
+            flat_groups_at_rank_i = self.flat_groups[rank_i]
+            start_group_id = None
+            end_group_id = None
+            for group_id in range(len(self.flat_groups_offset)):
+                if self.flat_groups_offset[group_id] <= self.offset < self.flat_groups_offset[group_id + 1]:
+                    start_group_id = group_id
+                if self.flat_groups_offset[group_id] < end_idx <= self.flat_groups_offset[group_id + 1]:
+                    end_group_id = group_id
+                    break
+            # collect weights from related group/groups
+            for group_id in range(start_group_id, end_group_id + 1):
+                flat_tensor = flat_groups_at_rank_i[group_id]
+                start_offset = self.offset - self.flat_groups_offset[group_id]
+                end_offset = min(end_idx, self.flat_groups_offset[group_id + 1]) - self.flat_groups_offset[group_id]
+                pad_flat_param_chunks.append(flat_tensor[start_offset:end_offset])
+
+        # collect weights from all ranks
+        pad_flat_param = torch.cat(pad_flat_param_chunks, dim=0)
+        param = pad_flat_param[:self.shape.numel()].view(self.shape).contiguous()
+        return param
+
+
+def _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
+    param_shapes = zero_model_states[0].param_shapes
+    avail_numel = sum([flat_group.numel() for flat_group in fp32_flat_groups[0]]) * world_size
+
+    # Reconstruction protocol: For zero3 we need to zip the partitions together at boundary of each
+    # param, re-consolidating each param, while dealing with padding if any
+
+    # merge list of dicts, preserving order
+    param_shapes = {k: v for d in param_shapes for k, v in d.items()}
+
+    if debug:
+        for i in range(world_size):
+            print(f"{FP32_FLAT_GROUPS}[{i}].shape={fp32_flat_groups[i].shape}")
+
+        wanted_params = len(param_shapes)
+        wanted_numel = sum(shape.numel() for shape in param_shapes.values())
+        # not asserting if there is a mismatch due to possible padding
+        avail_numel = fp32_flat_groups[0].numel() * world_size
+        print(f"Trainable params: Have {avail_numel} numels to process.")
+        print(f"Trainable params: Need {wanted_numel} numels in {wanted_params} params.")
+
+    # params
+    # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
+    # out-of-core computing solution
+    offset = 0
+    total_numel = 0
+    total_params = 0
+    flat_groups_offset = [0] + list(np.cumsum([flat_tensor.numel() for flat_tensor in fp32_flat_groups[0]]))
+    for name, shape in tqdm(param_shapes.items(), desc='Gathering sharded weights'):
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+        total_params += 1
+        partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
+
+        if debug:
+            print(
+                f"Trainable params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
+            )
+
+        # memory efficient tensor
+        tensor = GatheredTensor(fp32_flat_groups, flat_groups_offset, offset, partitioned_numel, shape)
+        state_dict[name] = tensor
+        offset += partitioned_numel
+
+    offset *= world_size
+
+    # Sanity check
+    if offset != avail_numel:
+        raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
+
+    print(f"Reconstructed Trainable fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                               exclude_frozen_parameters):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    if not exclude_frozen_parameters:
+        _zero3_merge_frozen_params(state_dict, world_size, zero_model_states)
+
+    _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
+
+    # recover shared parameters
+    for pair in zero_model_states[0].shared_params:
+        if pair[1] in state_dict:
+            state_dict[pair[0]] = state_dict[pair[1]]
+
+    return state_dict
+
+
+def to_torch_tensor(state_dict, return_empty_tensor=False):
+    """
+    Convert state_dict of GatheredTensor to torch tensor
+    """
+    torch_state_dict = {}
+    converted_tensors = {}
+    for name, tensor in state_dict.items():
+        tensor_id = id(tensor)
+        if tensor_id in converted_tensors:  # shared tensors
+            shared_tensor = torch_state_dict[converted_tensors[tensor_id]]
+            torch_state_dict[name] = shared_tensor
+        else:
+            converted_tensors[tensor_id] = name
+            if return_empty_tensor:
+                torch_state_dict[name] = torch.empty(tensor.shape, dtype=tensor.dtype)
+            else:
+                torch_state_dict[name] = tensor.contiguous()
+    return torch_state_dict
+
+
+def get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir,
+                                             tag=None,
+                                             exclude_frozen_parameters=False,
+                                             lazy_mode=False):
+    """
+    Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated state_dict that can be loaded with
+    ``load_state_dict()`` and used for training without DeepSpeed or shared with others, for example
+    via a model hub.
+
+    Args:
+        - ``checkpoint_dir``: path to the desired checkpoint folder
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in 'latest' file. e.g., ``global_step14``
+        - ``exclude_frozen_parameters``: exclude frozen parameters
+        - ``lazy_mode``: get state_dict in lazy mode. It returns a dict of pesduo tensor instead of torch tensor, which is more memory efficient.
+          Convert the pesduo tensor to torch tensor by ``.contiguous()``
+
+    Returns:
+        - pytorch ``state_dict``
+
+    A typical usage might be ::
+
+        from deepspeed.utils.zero_to_fp32 import get_fp32_state_dict_from_zero_checkpoint
+        # do the training and checkpoint saving
+        state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir) # already on cpu
+        model = model.cpu() # move to cpu
+        model.load_state_dict(state_dict)
+        # submit to model hub or save the model to share with others
+
+    In this example the ``model`` will no longer be usable in the deepspeed context of the same
+    application. i.e. you will need to re-initialize the deepspeed engine, since
+    ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
+
+    If you want it all done for you, use ``load_state_dict_from_zero_checkpoint`` instead.
+
+    Note: the above usage may not work if your application doesn't have sufficient free CPU memory.
+    You may need to use the offline approach using the ``zero_to_fp32.py`` script that is saved with
+    the checkpoint. Or you can load state_dict in lazy mode ::
+
+        from deepspeed.utils.zero_to_fp32 import get_fp32_state_dict_from_zero_checkpoint
+        state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, lazy_mode=True) # not on cpu
+        for name, lazy_tensor in state_dict.item():
+            tensor = lazy_tensor.contiguous()  # to cpu
+            print(name, tensor)
+            # del tensor to release memory if it no longer in use
+    """
+    if tag is None:
+        latest_path = os.path.join(checkpoint_dir, 'latest')
+        if os.path.isfile(latest_path):
+            with open(latest_path, 'r') as fd:
+                tag = fd.read().strip()
+        else:
+            raise ValueError(f"Unable to find 'latest' file at {latest_path}")
+
+    ds_checkpoint_dir = os.path.join(checkpoint_dir, tag)
+
+    if not os.path.isdir(ds_checkpoint_dir):
+        raise FileNotFoundError(f"Directory '{ds_checkpoint_dir}' doesn't exist")
+
+    state_dict = _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters)
+    if lazy_mode:
+        return state_dict
+    else:
+        return to_torch_tensor(state_dict)
+
+
+def convert_zero_checkpoint_to_fp32_state_dict(checkpoint_dir,
+                                               output_dir,
+                                               max_shard_size="5GB",
+                                               safe_serialization=False,
+                                               tag=None,
+                                               exclude_frozen_parameters=False):
+    """
+    Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict`` file that can be
+    loaded with ``torch.load(file)`` + ``load_state_dict()`` and used for training without DeepSpeed.
+
+    Args:
+        - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
+        - ``output_dir``: directory to the pytorch fp32 state_dict output files
+        - ``max_shard_size``: the maximum size for a checkpoint before being sharded, default value is 5GB
+        - ``safe_serialization``:  whether to save the model using `safetensors` or the traditional PyTorch way (that uses `pickle`).
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
+        - ``exclude_frozen_parameters``: exclude frozen parameters
+    """
+
+    # Dependency pre-check
+    if safe_serialization:
+        try:
+            from safetensors.torch import save_file
+        except ImportError:
+            print('If you want to use `safe_serialization`, please `pip install safetensors`')
+            raise
+    if max_shard_size is not None:
+        try:
+            from huggingface_hub import split_torch_state_dict_into_shards
+        except ImportError:
+            print('If you want to use `max_shard_size`, please `pip install huggingface_hub`')
+            raise
+
+    # Convert zero checkpoint to state_dict
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir,
+                                                          tag,
+                                                          exclude_frozen_parameters,
+                                                          lazy_mode=True)
+
+    # Shard the model if it is too big.
+    weights_name = "model.safetensors" if safe_serialization else "pytorch_model.bin"
+    if max_shard_size is not None:
+        filename_pattern = weights_name.replace(".bin", "{suffix}.bin").replace(".safetensors", "{suffix}.safetensors")
+        # an memory-efficient approach for sharding
+        empty_state_dict = to_torch_tensor(state_dict, return_empty_tensor=True)
+        state_dict_split = split_torch_state_dict_into_shards(empty_state_dict,
+                                                              filename_pattern=filename_pattern,
+                                                              max_shard_size=max_shard_size)
+    else:
+        from collections import namedtuple
+        StateDictSplit = namedtuple("StateDictSplit", ["is_sharded", "filename_to_tensors"])
+        state_dict_split = StateDictSplit(is_sharded=False,
+                                          filename_to_tensors={weights_name: list(state_dict.keys())})
+
+    # Save the model by shard
+    os.makedirs(output_dir, exist_ok=True)
+    filename_to_tensors = state_dict_split.filename_to_tensors.items()
+    for shard_file, tensors in tqdm(filename_to_tensors, desc="Saving checkpoint shards"):
+        shard_state_dict = {tensor_name: state_dict[tensor_name] for tensor_name in tensors}
+        shard_state_dict = to_torch_tensor(shard_state_dict)
+        output_path = os.path.join(output_dir, shard_file)
+        if safe_serialization:
+            save_file(shard_state_dict, output_path, metadata={"format": "pt"})
+        else:
+            torch.save(shard_state_dict, output_path)
+        # release the memory of current shard
+        for tensor_name in list(shard_state_dict.keys()):
+            del state_dict[tensor_name]
+            del shard_state_dict[tensor_name]
+        del shard_state_dict
+        gc.collect()
+
+    # Save index if sharded
+    if state_dict_split.is_sharded:
+        index = {
+            "metadata": state_dict_split.metadata,
+            "weight_map": state_dict_split.tensor_to_filename,
+        }
+        save_index_file = "model.safetensors.index.json" if safe_serialization else "pytorch_model.bin.index.json"
+        save_index_file = os.path.join(output_dir, save_index_file)
+        with open(save_index_file, "w", encoding="utf-8") as f:
+            content = json.dumps(index, indent=2, sort_keys=True) + "\n"
+            f.write(content)
+
+
+def load_state_dict_from_zero_checkpoint(model, checkpoint_dir, tag=None):
+    """
+    1. Put the provided model to cpu
+    2. Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict``
+    3. Load it into the provided model
+
+    Args:
+        - ``model``: the model object to update
+        - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
+
+    Returns:
+        - ``model`: modified model
+
+    Make sure you have plenty of CPU memory available before you call this function. If you don't
+    have enough use the ``zero_to_fp32.py`` utility to do the conversion. You will find it
+    conveniently placed for you in the checkpoint folder.
+
+    A typical usage might be ::
+
+        from deepspeed.utils.zero_to_fp32 import load_state_dict_from_zero_checkpoint
+        model = load_state_dict_from_zero_checkpoint(trainer.model, checkpoint_dir)
+        # submit to model hub or save the model to share with others
+
+    Note, that once this was run, the ``model`` will no longer be usable in the deepspeed context
+    of the same application. i.e. you will need to re-initialize the deepspeed engine, since
+    ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
+
+    """
+    logger.info(f"Extracting fp32 weights")
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag)
+
+    logger.info(f"Overwriting model with fp32 weights")
+    model = model.cpu()
+    model.load_state_dict(state_dict, strict=False)
+
+    return model
+
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument("checkpoint_dir",
+                        type=str,
+                        help="path to the desired checkpoint folder, e.g., path/checkpoint-12")
+    parser.add_argument("output_dir",
+                        type=str,
+                        help="directory to the pytorch fp32 state_dict output files"
+                        "(e.g. path/checkpoint-12-output/)")
+    parser.add_argument(
+        "--max_shard_size",
+        type=str,
+        default="5GB",
+        help="The maximum size for a checkpoint before being sharded. Checkpoints shard will then be each of size"
+        "lower than this size. If expressed as a string, needs to be digits followed by a unit (like `5MB`"
+        "We default it to 5GB in order for models to be able to run easily on free-tier google colab instances"
+        "without CPU OOM issues.")
+    parser.add_argument(
+        "--safe_serialization",
+        default=False,
+        action='store_true',
+        help="Whether to save the model using `safetensors` or the traditional PyTorch way (that uses `pickle`).")
+    parser.add_argument("-t",
+                        "--tag",
+                        type=str,
+                        default=None,
+                        help="checkpoint tag used as a unique identifier for checkpoint. e.g., global_step1")
+    parser.add_argument("--exclude_frozen_parameters", action='store_true', help="exclude frozen parameters")
+    parser.add_argument("-d", "--debug", action='store_true', help="enable debug")
+    args = parser.parse_args()
+
+    debug = args.debug
+
+    convert_zero_checkpoint_to_fp32_state_dict(args.checkpoint_dir,
+                                               args.output_dir,
+                                               max_shard_size=args.max_shard_size,
+                                               safe_serialization=args.safe_serialization,
+                                               tag=args.tag,
+                                               exclude_frozen_parameters=args.exclude_frozen_parameters)