init

.gitattributes

@@ -28,6 +28,7 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.tar filter=lfs diff=lfs merge=lfs -text
 *.tflite filter=lfs diff=lfs merge=lfs -text
 *.tgz filter=lfs diff=lfs merge=lfs -text
+tokenizer.json filter=lfs diff=lfs merge=lfs -text
 *.wasm filter=lfs diff=lfs merge=lfs -text
 *.xz filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text

added_tokens.json

@@ -0,0 +1,24 @@
+{
+  "</tool_call>": 151658,
+  "<tool_call>": 151657,
+  "<|AUDIO|>": 151646,
+  "<|IMAGE|>": 151655,
+  "<|VIDEO|>": 151656,
+  "<|audio_bos|>": 151647,
+  "<|audio_eos|>": 151648,
+  "<|box_end|>": 151649,
+  "<|endoftext|>": 151643,
+  "<|file_sep|>": 151664,
+  "<|fim_middle|>": 151660,
+  "<|fim_pad|>": 151662,
+  "<|fim_prefix|>": 151659,
+  "<|fim_suffix|>": 151661,
+  "<|im_end|>": 151645,
+  "<|im_start|>": 151644,
+  "<|quad_end|>": 151651,
+  "<|quad_start|>": 151650,
+  "<|repo_name|>": 151663,
+  "<|vision_bos|>": 151652,
+  "<|vision_eos|>": 151653,
+  "<|vision_pad|>": 151654
+}

chat_template.jinja

@@ -0,0 +1,7 @@
+{% set audio_count = namespace(value=0) %}{% set image_count = namespace(value=0) %}{% set video_count = namespace(value=0) %}{% for message in messages %}{% if loop.first and message['role'] != 'system' %}<|im_start|>system
+You are a helpful assistant.<|im_end|>
+{% endif %}<|im_start|>{{ message['role'] }}
+{% if message['content'] is string %}{{ message['content'] }}<|im_end|>
+{% else %}{% for content in message['content'] %}{% if content['type'] == 'image' or 'image' in content or 'image_url' in content %}{% set image_count.value = image_count.value + 1 %}{% if add_vision_id %}Picture {{ image_count.value }}: {% endif %}<|vision_bos|><|IMAGE|><|vision_eos|>{% elif content['type'] == 'audio' or 'audio' in content or 'audio_url' in content %}{% set audio_count.value = audio_count.value + 1 %}{% if add_audio_id %}Audio {{ audio_count.value }}: {% endif %}<|audio_bos|><|AUDIO|><|audio_eos|>{% elif content['type'] == 'video' or 'video' in content %}{% set video_count.value = video_count.value + 1 %}{% if add_vision_id %}Video {{ video_count.value }}: {% endif %}<|vision_bos|><|VIDEO|><|vision_eos|>{% elif 'text' in content %}{{ content['text'] }}{% endif %}{% endfor %}<|im_end|>
+{% endif %}{% endfor %}{% if add_generation_prompt %}<|im_start|>assistant
+{% endif %}

config.json

@@ -0,0 +1,573 @@
+{
+  "architectures": [
+    "Qwen2_5OmniForConditionalGeneration"
+  ],
+  "enable_audio_output": true,
+  "enable_talker": true,
+  "model_type": "qwen2_5_omni",
+  "talker_config": {
+    "_attn_implementation_autoset": true,
+    "_name_or_path": "Qwen2.5-Omni-7B/talker",
+    "architectures": [
+      "Qwen2OmniTalkerForConditionalGeneration"
+    ],
+    "attention_dropout": 0.0,
+    "audio_end_token_id": 151648,
+    "audio_start_token_id": 151647,
+    "audio_token_index": 151646,
+    "embedding_size": 3584,
+    "head_dim": 128,
+    "hidden_act": "silu",
+    "hidden_size": 896,
+    "image_token_index": 151655,
+    "init_std": 0.02,
+    "initializer_range": 0.02,
+    "intermediate_size": 18944,
+    "max_position_embeddings": 32768,
+    "max_window_layers": 28,
+    "model_type": "qwen2_5_omni_talker",
+    "num_attention_heads": 12,
+    "num_hidden_layers": 24,
+    "num_key_value_heads": 4,
+    "position_id_per_seconds": 25,
+    "rms_norm_eps": 1e-06,
+    "rope_scaling": {
+      "mrope_section": [
+        16,
+        24,
+        24
+      ],
+      "rope_type": "default",
+      "type": "default"
+    },
+    "rope_theta": 1000000.0,
+    "seconds_per_chunk": 2,
+    "sliding_window": 32768,
+    "spatial_merge_size": 2,
+    "torch_dtype": "float32",
+    "tts_codec_end_token_id": 8294,
+    "tts_codec_mask_token_id": 8296,
+    "tts_codec_pad_token_id": 8292,
+    "tts_codec_start_token_id": 8293,
+    "tts_text_end_token_id": 151861,
+    "tts_text_pad_token_id": 151859,
+    "tts_text_start_token_id": 151860,
+    "use_cache": true,
+    "use_sliding_window": false,
+    "video_token_index": 151656,
+    "vision_end_token_id": 151653,
+    "vision_start_token_id": 151652,
+    "vocab_size": 8448
+  },
+  "thinker_config": {
+    "_attn_implementation_autoset": true,
+    "_name_or_path": "Qwen2.5-Omni-7B/thinker",
+    "architectures": [
+      "Qwen2OmniNaViTThinkerForConditionalGeneration"
+    ],
+    "audio_config": {
+      "_attn_implementation_autoset": true,
+      "_name_or_path": "",
+      "activation_dropout": 0.0,
+      "activation_function": "gelu",
+      "add_cross_attention": false,
+      "architectures": null,
+      "attention_dropout": 0.0,
+      "bad_words_ids": null,
+      "begin_suppress_tokens": null,
+      "bos_token_id": null,
+      "chunk_size_feed_forward": 0,
+      "cross_attention_hidden_size": null,
+      "d_model": 1280,
+      "decoder_start_token_id": null,
+      "diversity_penalty": 0.0,
+      "do_sample": false,
+      "dropout": 0.0,
+      "early_stopping": false,
+      "encoder_attention_heads": 20,
+      "encoder_ffn_dim": 5120,
+      "encoder_layerdrop": 0.0,
+      "encoder_layers": 32,
+      "encoder_no_repeat_ngram_size": 0,
+      "eos_token_id": null,
+      "exponential_decay_length_penalty": null,
+      "finetuning_task": null,
+      "forced_bos_token_id": null,
+      "forced_eos_token_id": null,
+      "id2label": {
+        "0": "LABEL_0",
+        "1": "LABEL_1"
+      },
+      "init_std": 0.02,
+      "initializer_range": 0.02,
+      "is_decoder": false,
+      "is_encoder_decoder": false,
+      "label2id": {
+        "LABEL_0": 0,
+        "LABEL_1": 1
+      },
+      "length_penalty": 1.0,
+      "max_length": 20,
+      "max_source_positions": 1500,
+      "min_length": 0,
+      "model_type": "qwen2_5_omni_audio_encoder",
+      "n_window": 100,
+      "no_repeat_ngram_size": 0,
+      "num_beam_groups": 1,
+      "num_beams": 1,
+      "num_hidden_layers": 32,
+      "num_mel_bins": 128,
+      "num_return_sequences": 1,
+      "output_attentions": false,
+      "output_dim": 3584,
+      "output_hidden_states": false,
+      "output_scores": false,
+      "pad_token_id": null,
+      "prefix": null,
+      "problem_type": null,
+      "pruned_heads": {},
+      "remove_invalid_values": false,
+      "repetition_penalty": 1.0,
+      "return_dict": true,
+      "return_dict_in_generate": false,
+      "scale_embedding": false,
+      "sep_token_id": null,
+      "suppress_tokens": null,
+      "task_specific_params": null,
+      "temperature": 1.0,
+      "tf_legacy_loss": false,
+      "tie_encoder_decoder": false,
+      "tie_word_embeddings": true,
+      "tokenizer_class": null,
+      "top_k": 50,
+      "top_p": 1.0,
+      "torch_dtype": null,
+      "torchscript": false,
+      "typical_p": 1.0,
+      "use_bfloat16": false
+    },
+    "audio_end_token_id": 151648,
+    "audio_start_token_id": 151647,
+    "audio_token_index": 151646,
+    "bos_token_id": 151644,
+    "eos_token_id": 151645,
+    "ignore_index": -100,
+    "image_token_index": 151655,
+    "init_std": 0.02,
+    "initializer_range": 0.02,
+    "model_type": "qwen2_5_omni_thinker",
+    "pad_token_id": 151643,
+    "position_id_per_seconds": 25,
+    "seconds_per_chunk": 2,
+    "text_config": {
+      "_attn_implementation_autoset": false,
+      "_name_or_path": "",
+      "add_cross_attention": false,
+      "architectures": null,
+      "attention_dropout": 0.0,
+      "bad_words_ids": null,
+      "begin_suppress_tokens": null,
+      "bos_token_id": null,
+      "chunk_size_feed_forward": 0,
+      "cross_attention_hidden_size": null,
+      "decoder_start_token_id": null,
+      "diversity_penalty": 0.0,
+      "do_sample": false,
+      "early_stopping": false,
+      "encoder_no_repeat_ngram_size": 0,
+      "eos_token_id": null,
+      "exponential_decay_length_penalty": null,
+      "finetuning_task": null,
+      "forced_bos_token_id": null,
+      "forced_eos_token_id": null,
+      "hidden_act": "silu",
+      "hidden_size": 3584,
+      "id2label": {
+        "0": "LABEL_0",
+        "1": "LABEL_1"
+      },
+      "init_std": 0.02,
+      "initializer_range": 0.02,
+      "intermediate_size": 18944,
+      "is_decoder": false,
+      "is_encoder_decoder": false,
+      "label2id": {
+        "LABEL_0": 0,
+        "LABEL_1": 1
+      },
+      "length_penalty": 1.0,
+      "max_length": 20,
+      "max_position_embeddings": 32768,
+      "max_window_layers": 28,
+      "min_length": 0,
+      "model_type": "qwen2_5_omni_text",
+      "no_repeat_ngram_size": 0,
+      "num_attention_heads": 28,
+      "num_beam_groups": 1,
+      "num_beams": 1,
+      "num_hidden_layers": 28,
+      "num_key_value_heads": 4,
+      "num_return_sequences": 1,
+      "output_attentions": false,
+      "output_hidden_states": false,
+      "output_scores": false,
+      "pad_token_id": null,
+      "prefix": null,
+      "problem_type": null,
+      "pruned_heads": {},
+      "remove_invalid_values": false,
+      "repetition_penalty": 1.0,
+      "return_dict": true,
+      "return_dict_in_generate": false,
+      "rms_norm_eps": 1e-06,
+      "rope_scaling": {
+        "mrope_section": [
+          16,
+          24,
+          24
+        ],
+        "rope_type": "default",
+        "type": "default"
+      },
+      "rope_theta": 1000000.0,
+      "sep_token_id": null,
+      "sliding_window": 32768,
+      "suppress_tokens": null,
+      "task_specific_params": null,
+      "temperature": 1.0,
+      "tf_legacy_loss": false,
+      "tie_encoder_decoder": false,
+      "tie_word_embeddings": false,
+      "tokenizer_class": null,
+      "top_k": 50,
+      "top_p": 1.0,
+      "torch_dtype": null,
+      "torchscript": false,
+      "typical_p": 1.0,
+      "use_bfloat16": false,
+      "use_cache": true,
+      "use_sliding_window": false,
+      "vocab_size": 152064
+    },
+    "torch_dtype": "float32",
+    "user_token_id": 872,
+    "video_token_index": 151656,
+    "vision_config": {
+      "_attn_implementation_autoset": true,
+      "_name_or_path": "",
+      "add_cross_attention": false,
+      "architectures": null,
+      "bad_words_ids": null,
+      "begin_suppress_tokens": null,
+      "bos_token_id": null,
+      "chunk_size_feed_forward": 0,
+      "cross_attention_hidden_size": null,
+      "decoder_start_token_id": null,
+      "depth": 32,
+      "diversity_penalty": 0.0,
+      "do_sample": false,
+      "early_stopping": false,
+      "embed_dim": 1280,
+      "encoder_no_repeat_ngram_size": 0,
+      "eos_token_id": null,
+      "exponential_decay_length_penalty": null,
+      "finetuning_task": null,
+      "forced_bos_token_id": null,
+      "forced_eos_token_id": null,
+      "fullatt_block_indexes": [
+        7,
+        15,
+        23,
+        31
+      ],
+      "hidden_act": "silu",
+      "hidden_size": 1280,
+      "id2label": {
+        "0": "LABEL_0",
+        "1": "LABEL_1"
+      },
+      "in_channels": 3,
+      "in_chans": 3,
+      "init_std": 0.02,
+      "initializer_range": 0.02,
+      "intermediate_size": 3420,
+      "is_decoder": false,
+      "is_encoder_decoder": false,
+      "label2id": {
+        "LABEL_0": 0,
+        "LABEL_1": 1
+      },
+      "length_penalty": 1.0,
+      "max_length": 20,
+      "min_length": 0,
+      "model_type": "qwen2_5_omni_vision_encoder",
+      "no_repeat_ngram_size": 0,
+      "num_beam_groups": 1,
+      "num_beams": 1,
+      "num_heads": 16,
+      "num_return_sequences": 1,
+      "out_hidden_size": 3584,
+      "output_attentions": false,
+      "output_hidden_states": false,
+      "output_scores": false,
+      "pad_token_id": null,
+      "patch_size": 14,
+      "prefix": null,
+      "problem_type": null,
+      "pruned_heads": {},
+      "remove_invalid_values": false,
+      "repetition_penalty": 1.0,
+      "return_dict": true,
+      "return_dict_in_generate": false,
+      "sep_token_id": null,
+      "spatial_merge_size": 2,
+      "spatial_patch_size": 14,
+      "suppress_tokens": null,
+      "task_specific_params": null,
+      "temperature": 1.0,
+      "temporal_patch_size": 2,
+      "tf_legacy_loss": false,
+      "tie_encoder_decoder": false,
+      "tie_word_embeddings": true,
+      "tokenizer_class": null,
+      "tokens_per_second": 25,
+      "top_k": 50,
+      "top_p": 1.0,
+      "torch_dtype": null,
+      "torchscript": false,
+      "typical_p": 1.0,
+      "use_bfloat16": false,
+      "window_size": 112
+    },
+    "vision_end_token_id": 151653,
+    "vision_start_token_id": 151652,
+    "vision_token_id": 151654
+  },
+  "token2wav_config": {
+    "_attn_implementation_autoset": true,
+    "bigvgan_config": {
+      "_attn_implementation_autoset": true,
+      "_name_or_path": "",
+      "add_cross_attention": false,
+      "architectures": null,
+      "bad_words_ids": null,
+      "begin_suppress_tokens": null,
+      "bos_token_id": null,
+      "chunk_size_feed_forward": 0,
+      "cross_attention_hidden_size": null,
+      "decoder_start_token_id": null,
+      "diversity_penalty": 0.0,
+      "do_sample": false,
+      "early_stopping": false,
+      "encoder_no_repeat_ngram_size": 0,
+      "eos_token_id": null,
+      "exponential_decay_length_penalty": null,
+      "finetuning_task": null,
+      "forced_bos_token_id": null,
+      "forced_eos_token_id": null,
+      "id2label": {
+        "0": "LABEL_0",
+        "1": "LABEL_1"
+      },
+      "is_decoder": false,
+      "is_encoder_decoder": false,
+      "label2id": {
+        "LABEL_0": 0,
+        "LABEL_1": 1
+      },
+      "length_penalty": 1.0,
+      "max_length": 20,
+      "mel_dim": 80,
+      "min_length": 0,
+      "model_type": "qwen2_5_omni_bigvgan",
+      "no_repeat_ngram_size": 0,
+      "num_beam_groups": 1,
+      "num_beams": 1,
+      "num_return_sequences": 1,
+      "output_attentions": false,
+      "output_hidden_states": false,
+      "output_scores": false,
+      "pad_token_id": null,
+      "prefix": null,
+      "problem_type": null,
+      "pruned_heads": {},
+      "remove_invalid_values": false,
+      "repetition_penalty": 1.0,
+      "resblock_dilation_sizes": [
+        [
+          1,
+          3,
+          5
+        ],
+        [
+          1,
+          3,
+          5
+        ],
+        [
+          1,
+          3,
+          5
+        ]
+      ],
+      "resblock_kernel_sizes": [
+        3,
+        7,
+        11
+      ],
+      "return_dict": true,
+      "return_dict_in_generate": false,
+      "sep_token_id": null,
+      "suppress_tokens": null,
+      "task_specific_params": null,
+      "temperature": 1.0,
+      "tf_legacy_loss": false,
+      "tie_encoder_decoder": false,
+      "tie_word_embeddings": true,
+      "tokenizer_class": null,
+      "top_k": 50,
+      "top_p": 1.0,
+      "torch_dtype": null,
+      "torchscript": false,
+      "typical_p": 1.0,
+      "upsample_initial_channel": 1536,
+      "upsample_kernel_sizes": [
+        11,
+        7,
+        4,
+        4,
+        4,
+        4
+      ],
+      "upsample_rates": [
+        5,
+        3,
+        2,
+        2,
+        2,
+        2
+      ],
+      "use_bfloat16": false,
+      "use_bias_at_final": false
+    },
+    "dit_config": {
+      "_attn_implementation_autoset": true,
+      "_name_or_path": "",
+      "add_cross_attention": false,
+      "architectures": null,
+      "bad_words_ids": null,
+      "begin_suppress_tokens": null,
+      "block_size": 24,
+      "bos_token_id": null,
+      "chunk_size_feed_forward": 0,
+      "cross_attention_hidden_size": null,
+      "decoder_start_token_id": null,
+      "depth": 22,
+      "dim": 1024,
+      "diversity_penalty": 0.0,
+      "do_sample": false,
+      "dropout": 0.1,
+      "early_stopping": false,
+      "emb_dim": 512,
+      "enc_attention_channels": 64,
+      "enc_channels": [
+        256,
+        256,
+        256,
+        256,
+        768
+      ],
+      "enc_dilations": [
+        1,
+        2,
+        3,
+        4,
+        1
+      ],
+      "enc_dim": 128,
+      "enc_emb_dim": 192,
+      "enc_global_context": true,
+      "enc_kernel_sizes": [
+        5,
+        3,
+        3,
+        3,
+        1
+      ],
+      "enc_lin_neurons": 192,
+      "enc_res2net_scale": 2,
+      "enc_se_channels": 64,
+      "encoder_no_repeat_ngram_size": 0,
+      "eos_token_id": null,
+      "exponential_decay_length_penalty": null,
+      "ff_mult": 2,
+      "finetuning_task": null,
+      "forced_bos_token_id": null,
+      "forced_eos_token_id": null,
+      "head_dim": 64,
+      "heads": 16,
+      "hidden_size": 1024,
+      "id2label": {
+        "0": "LABEL_0",
+        "1": "LABEL_1"
+      },
+      "is_decoder": false,
+      "is_encoder_decoder": false,
+      "label2id": {
+        "LABEL_0": 0,
+        "LABEL_1": 1
+      },
+      "length_penalty": 1.0,
+      "look_ahead_layers": [
+        10
+      ],
+      "look_backward_layers": [
+        0,
+        20
+      ],
+      "max_length": 20,
+      "max_position_embeddings": 32768,
+      "mel_dim": 80,
+      "min_length": 0,
+      "model_type": "qwen2_5_omni_dit",
+      "no_repeat_ngram_size": 0,
+      "num_attention_heads": 16,
+      "num_beam_groups": 1,
+      "num_beams": 1,
+      "num_embeds": 8193,
+      "num_hidden_layers": 22,
+      "num_return_sequences": 1,
+      "output_attentions": false,
+      "output_hidden_states": false,
+      "output_scores": false,
+      "pad_token_id": null,
+      "prefix": null,
+      "problem_type": null,
+      "pruned_heads": {},
+      "remove_invalid_values": false,
+      "repeats": 2,
+      "repetition_penalty": 1.0,
+      "return_dict": true,
+      "return_dict_in_generate": false,
+      "rope_theta": 10000.0,
+      "sep_token_id": null,
+      "suppress_tokens": null,
+      "task_specific_params": null,
+      "temperature": 1.0,
+      "tf_legacy_loss": false,
+      "tie_encoder_decoder": false,
+      "tie_word_embeddings": true,
+      "tokenizer_class": null,
+      "top_k": 50,
+      "top_p": 1.0,
+      "torch_dtype": "float32",
+      "torchscript": false,
+      "typical_p": 1.0,
+      "use_bfloat16": false
+    },
+    "model_type": "qwen2_5_omni_token2wav",
+    "torch_dtype": "float32"
+  },
+  "torch_dtype": "bfloat16",
+  "transformers_version": "4.52.0.dev0"
+}

generation_config.json

@@ -0,0 +1,4 @@
+{
+  "_from_model_config": true,
+  "transformers_version": "4.52.0.dev0"
+}

model-00001-of-00004.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fa2866bd0f5b18079e4fbf6470659cc975fc73b6025afef81bd863658e130e68
+size 4985055536

model-00002-of-00004.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bf3ecae31a5699f2146af5d9cb4c59947a8395a4bc7f8e2b179b53f7567a5212
+size 4991496832

model-00003-of-00004.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7675269df75c670fa6a63f71a091570028a2d969e4f1602f8d12abb0575c68be
+size 4991496936

model-00004-of-00004.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c975b68f6fbb37fdb31f59e5c4bb2415be9a57f02fa5b8e6ec0acf7fc152b1cb
+size 2895740064

preprocessor_config.json

@@ -0,0 +1,31 @@
+{
+  "chunk_length": 300,
+  "dither": 0.0,
+  "feature_extractor_type": "WhisperFeatureExtractor",
+  "feature_size": 128,
+  "hop_length": 160,
+  "image_mean": [
+    0.48145466,
+    0.4578275,
+    0.40821073
+  ],
+  "image_processor_type": "Qwen2VLImageProcessor",
+  "image_std": [
+    0.26862954,
+    0.26130258,
+    0.27577711
+  ],
+  "max_pixels": 12845056,
+  "merge_size": 2,
+  "min_pixels": 3136,
+  "n_fft": 400,
+  "n_samples": 4800000,
+  "nb_max_frames": 30000,
+  "padding_side": "right",
+  "padding_value": 0.0,
+  "patch_size": 14,
+  "processor_class": "Qwen2_5OmniProcessor",
+  "return_attention_mask": true,
+  "sampling_rate": 16000,
+  "temporal_patch_size": 2
+}

special_tokens_map.json

@@ -0,0 +1,38 @@
+{
+  "additional_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|AUDIO|>",
+    "<|audio_bos|>",
+    "<|audio_eos|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_bos|>",
+    "<|vision_eos|>",
+    "<|vision_pad|>",
+    "<|IMAGE|>",
+    "<|VIDEO|>"
+  ],
+  "audio_bos_token": "<|audio_bos|>",
+  "audio_eos_token": "<|audio_eos|>",
+  "audio_token": "<|AUDIO|>",
+  "eos_token": {
+    "content": "<|im_end|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "image_token": "<|IMAGE|>",
+  "pad_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "video_token": "<|VIDEO|>",
+  "vision_bos_token": "<|vision_bos|>",
+  "vision_eos_token": "<|vision_eos|>"
+}

spk_dict.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6a05609b28f5d42b7b748f0f07592545c8f1f6885b9ae8fff64baf56e86b2a18
+size 259544

tokenizer.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f9711e245647e88538786834977dc8afb51172e879ee661352c587cf01efd6b0
+size 11422037

tokenizer_config.json

@@ -0,0 +1,226 @@
+{
+  "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": "<|AUDIO|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151647": {
+      "content": "<|audio_bos|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151648": {
+      "content": "<|audio_eos|>",
+      "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_bos|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151653": {
+      "content": "<|vision_eos|>",
+      "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|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151656": {
+      "content": "<|VIDEO|>",
+      "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|>",
+    "<|AUDIO|>",
+    "<|audio_bos|>",
+    "<|audio_eos|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_bos|>",
+    "<|vision_eos|>",
+    "<|vision_pad|>",
+    "<|IMAGE|>",
+    "<|VIDEO|>"
+  ],
+  "audio_bos_token": "<|audio_bos|>",
+  "audio_eos_token": "<|audio_eos|>",
+  "audio_token": "<|AUDIO|>",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|im_end|>",
+  "errors": "replace",
+  "extra_special_tokens": {
+    "audio_bos_token": "<|audio_bos|>",
+    "audio_eos_token": "<|audio_eos|>",
+    "audio_token": "<|AUDIO|>",
+    "image_token": "<|IMAGE|>",
+    "video_token": "<|VIDEO|>",
+    "vision_bos_token": "<|vision_bos|>",
+    "vision_eos_token": "<|vision_eos|>"
+  },
+  "image_token": "<|IMAGE|>",
+  "max_length": null,
+  "model_max_length": 32768,
+  "pad_to_multiple_of": null,
+  "pad_token": "<|endoftext|>",
+  "pad_token_type_id": 0,
+  "padding_side": "left",
+  "processor_class": "Qwen2_5OmniProcessor",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null,
+  "video_token": "<|VIDEO|>",
+  "vision_bos_token": "<|vision_bos|>",
+  "vision_eos_token": "<|vision_eos|>"
+}

trainer_state.json

@@ -0,0 +1,2449 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.0,
+  "eval_steps": 500,
+  "global_step": 161,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "advantages": 0.0,
+      "completion_length": 77.421875,
+      "epoch": 0.006211180124223602,
+      "grad_norm": 4.007043838500977,
+      "kl": 0.0,
+      "learning_rate": 9.937888198757763e-07,
+      "loss": -0.0,
+      "reward": 1.46875,
+      "reward_mean": 1.46875,
+      "reward_std": 0.23356688022613525,
+      "rewards/accuracy_reward": 0.46875,
+      "rewards/format_reward": 1.0,
+      "step": 1
+    },
+    {
+      "advantages": 1.30385160446167e-08,
+      "completion_length": 78.921875,
+      "epoch": 0.012422360248447204,
+      "grad_norm": 7.15122652053833,
+      "kl": 0.00041961669921875,
+      "learning_rate": 9.875776397515528e-07,
+      "loss": 0.0,
+      "reward": 1.53125,
+      "reward_mean": 1.53125,
+      "reward_std": 0.2845909595489502,
+      "rewards/accuracy_reward": 0.53125,
+      "rewards/format_reward": 1.0,
+      "step": 2
+    },
+    {
+      "advantages": 0.0,
+      "completion_length": 78.671875,
+      "epoch": 0.018633540372670808,
+      "grad_norm": 2.9171459674835205,
+      "kl": 0.000392913818359375,
+      "learning_rate": 9.813664596273291e-07,
+      "loss": 0.0,
+      "reward": 1.6875,
+      "reward_mean": 1.6875,
+      "reward_std": 0.1552036553621292,
+      "rewards/accuracy_reward": 0.6875,
+      "rewards/format_reward": 1.0,
+      "step": 3
+    },
+    {
+      "advantages": -1.862645149230957e-09,
+      "completion_length": 77.78125,
+      "epoch": 0.024844720496894408,
+      "grad_norm": 5.474589824676514,
+      "kl": 0.000759124755859375,
+      "learning_rate": 9.751552795031055e-07,
+      "loss": 0.0001,
+      "reward": 1.71875,
+      "reward_mean": 1.71875,
+      "reward_std": 0.213067427277565,
+      "rewards/accuracy_reward": 0.71875,
+      "rewards/format_reward": 1.0,
+      "step": 4
+    },
+    {
+      "advantages": -2.7939677238464355e-09,
+      "completion_length": 90.8125,
+      "epoch": 0.031055900621118012,
+      "grad_norm": 5.2480363845825195,
+      "kl": 0.001495361328125,
+      "learning_rate": 9.68944099378882e-07,
+      "loss": 0.0001,
+      "reward": 1.53125,
+      "reward_mean": 1.53125,
+      "reward_std": 0.17570313811302185,
+      "rewards/accuracy_reward": 0.53125,
+      "rewards/format_reward": 1.0,
+      "step": 5
+    },
+    {
+      "advantages": 6.51925802230835e-09,
+      "completion_length": 81.203125,
+      "epoch": 0.037267080745341616,
+      "grad_norm": 8.329508781433105,
+      "kl": 0.006256103515625,
+      "learning_rate": 9.627329192546583e-07,
+      "loss": 0.0006,
+      "reward": 1.375,
+      "reward_mean": 1.375,
+      "reward_std": 0.26409146189689636,
+      "rewards/accuracy_reward": 0.375,
+      "rewards/format_reward": 1.0,
+      "step": 6
+    },
+    {
+      "advantages": -3.725290298461914e-09,
+      "completion_length": 84.265625,
+      "epoch": 0.043478260869565216,
+      "grad_norm": 9.400680541992188,
+      "kl": 0.0106201171875,
+      "learning_rate": 9.565217391304349e-07,
+      "loss": 0.0011,
+      "reward": 1.65625,
+      "reward_mean": 1.65625,
+      "reward_std": 0.2404065877199173,
+      "rewards/accuracy_reward": 0.65625,
+      "rewards/format_reward": 1.0,
+      "step": 7
+    },
+    {
+      "advantages": -2.7939677238464355e-09,
+      "completion_length": 84.109375,
+      "epoch": 0.049689440993788817,
+      "grad_norm": 6.011223316192627,
+      "kl": 0.0057373046875,
+      "learning_rate": 9.503105590062112e-07,
+      "loss": 0.0006,
+      "reward": 1.4375,
+      "reward_mean": 1.4375,
+      "reward_std": 0.22461533546447754,
+      "rewards/accuracy_reward": 0.453125,
+      "rewards/format_reward": 0.984375,
+      "step": 8
+    },
+    {
+      "advantages": -3.725290298461914e-09,
+      "completion_length": 87.046875,
+      "epoch": 0.055900621118012424,
+      "grad_norm": 4.103212356567383,
+      "kl": 0.00244140625,
+      "learning_rate": 9.440993788819875e-07,
+      "loss": 0.0002,
+      "reward": 1.71875,
+      "reward_mean": 1.71875,
+      "reward_std": 0.0578637570142746,
+      "rewards/accuracy_reward": 0.71875,
+      "rewards/format_reward": 1.0,
+      "step": 9
+    },
+    {
+      "advantages": -7.450580596923828e-09,
+      "completion_length": 76.15625,
+      "epoch": 0.062111801242236024,
+      "grad_norm": 7.4132466316223145,
+      "kl": 0.01287841796875,
+      "learning_rate": 9.37888198757764e-07,
+      "loss": 0.0013,
+      "reward": 1.4375,
+      "reward_mean": 1.4375,
+      "reward_std": 0.3335031569004059,
+      "rewards/accuracy_reward": 0.4375,
+      "rewards/format_reward": 1.0,
+      "step": 10
+    },
+    {
+      "advantages": 3.725290298461914e-09,
+      "completion_length": 76.75,
+      "epoch": 0.06832298136645963,
+      "grad_norm": 4.6751017570495605,
+      "kl": 0.0130615234375,
+      "learning_rate": 9.316770186335403e-07,
+      "loss": 0.0013,
+      "reward": 1.28125,
+      "reward_mean": 1.28125,
+      "reward_std": 0.0578637570142746,
+      "rewards/accuracy_reward": 0.28125,
+      "rewards/format_reward": 1.0,
+      "step": 11
+    },
+    {
+      "advantages": -1.862645149230957e-09,
+      "completion_length": 76.390625,
+      "epoch": 0.07453416149068323,
+      "grad_norm": 7.682182788848877,
+      "kl": 0.018310546875,
+      "learning_rate": 9.254658385093167e-07,
+      "loss": 0.0018,
+      "reward": 1.734375,
+      "reward_mean": 1.734375,
+      "reward_std": 0.15992169082164764,
+      "rewards/accuracy_reward": 0.734375,
+      "rewards/format_reward": 1.0,
+      "step": 12
+    },
+    {
+      "advantages": -2.7939677238464355e-09,
+      "completion_length": 85.234375,
+      "epoch": 0.08074534161490683,
+      "grad_norm": 4.814305305480957,
+      "kl": 0.00396728515625,
+      "learning_rate": 9.19254658385093e-07,
+      "loss": 0.0004,
+      "reward": 1.8125,
+      "reward_mean": 1.8125,
+      "reward_std": 0.22461533546447754,
+      "rewards/accuracy_reward": 0.8125,
+      "rewards/format_reward": 1.0,
+      "step": 13
+    },
+    {
+      "advantages": -1.862645149230957e-09,
+      "completion_length": 86.3125,
+      "epoch": 0.08695652173913043,
+      "grad_norm": 189.3062744140625,
+      "kl": 0.033203125,
+      "learning_rate": 9.130434782608695e-07,
+      "loss": 0.0033,
+      "reward": 1.609375,
+      "reward_mean": 1.609375,
+      "reward_std": 0.19044628739356995,
+      "rewards/accuracy_reward": 0.609375,
+      "rewards/format_reward": 1.0,
+      "step": 14
+    },
+    {
+      "advantages": 2.7939677238464355e-09,
+      "completion_length": 69.8125,
+      "epoch": 0.09316770186335403,
+      "grad_norm": 3.5520412921905518,
+      "kl": 0.01129150390625,
+      "learning_rate": 9.06832298136646e-07,
+      "loss": 0.0011,
+      "reward": 1.78125,
+      "reward_mean": 1.78125,
+      "reward_std": 0.10888782143592834,
+      "rewards/accuracy_reward": 0.78125,
+      "rewards/format_reward": 1.0,
+      "step": 15
+    },
+    {
+      "advantages": -9.313225746154785e-10,
+      "completion_length": 76.296875,
+      "epoch": 0.09937888198757763,
+      "grad_norm": 3.526542901992798,
+      "kl": 0.00909423828125,
+      "learning_rate": 9.006211180124223e-07,
+      "loss": 0.0009,
+      "reward": 1.609375,
+      "reward_mean": 1.609375,
+      "reward_std": 0.12255740165710449,
+      "rewards/accuracy_reward": 0.609375,
+      "rewards/format_reward": 1.0,
+      "step": 16
+    },
+    {
+      "advantages": 1.862645149230957e-09,
+      "completion_length": 86.3125,
+      "epoch": 0.10559006211180125,
+      "grad_norm": 5.98048210144043,
+      "kl": 0.0057373046875,
+      "learning_rate": 8.944099378881988e-07,
+      "loss": 0.0006,
+      "reward": 1.71875,
+      "reward_mean": 1.71875,
+      "reward_std": 0.2041158676147461,
+      "rewards/accuracy_reward": 0.71875,
+      "rewards/format_reward": 1.0,
+      "step": 17
+    },
+    {
+      "advantages": 0.0,
+      "completion_length": 82.921875,
+      "epoch": 0.11180124223602485,
+      "grad_norm": 2.0705599784851074,
+      "kl": 0.00592041015625,
+      "learning_rate": 8.881987577639751e-07,
+      "loss": 0.0006,
+      "reward": 1.5625,
+      "reward_mean": 1.5625,
+      "reward_std": 0.06681530922651291,
+      "rewards/accuracy_reward": 0.578125,
+      "rewards/format_reward": 0.984375,
+      "step": 18
+    },
+    {
+      "advantages": -2.7939677238464355e-09,
+      "completion_length": 81.40625,
+      "epoch": 0.11801242236024845,
+      "grad_norm": 9.266715049743652,
+      "kl": 0.0079345703125,
+      "learning_rate": 8.819875776397515e-07,
+      "loss": 0.0008,
+      "reward": 1.546875,
+      "reward_mean": 1.546875,
+      "reward_std": 0.2109457552433014,
+      "rewards/accuracy_reward": 0.5625,
+      "rewards/format_reward": 0.984375,
+      "step": 19
+    },
+    {
+      "advantages": -3.725290298461914e-09,
+      "completion_length": 80.203125,
+      "epoch": 0.12422360248447205,
+      "grad_norm": 10.367863655090332,
+      "kl": 0.0072021484375,
+      "learning_rate": 8.757763975155279e-07,
+      "loss": 0.0007,
+      "reward": 1.40625,
+      "reward_mean": 1.40625,
+      "reward_std": 0.2404065728187561,
+      "rewards/accuracy_reward": 0.40625,
+      "rewards/format_reward": 1.0,
+      "step": 20
+    },
+    {
+      "advantages": 1.862645149230957e-09,
+      "completion_length": 75.734375,
+      "epoch": 0.13043478260869565,
+      "grad_norm": 2.6553478240966797,
+      "kl": 0.00592041015625,
+      "learning_rate": 8.695652173913043e-07,
+      "loss": 0.0006,
+      "reward": 1.578125,
+      "reward_mean": 1.578125,
+      "reward_std": 0.10205793380737305,
+      "rewards/accuracy_reward": 0.578125,
+      "rewards/format_reward": 1.0,
+      "step": 21
+    },
+    {
+      "advantages": 3.725290298461914e-09,
+      "completion_length": 85.8125,
+      "epoch": 0.13664596273291926,
+      "grad_norm": 3.458266496658325,
+      "kl": 0.00604248046875,
+      "learning_rate": 8.633540372670807e-07,
+      "loss": 0.0006,
+      "reward": 1.515625,
+      "reward_mean": 1.515625,
+      "reward_std": 0.15981829166412354,
+      "rewards/accuracy_reward": 0.53125,
+      "rewards/format_reward": 0.984375,
+      "step": 22
+    },
+    {
+      "advantages": -2.7939677238464355e-09,
+      "completion_length": 79.328125,
+      "epoch": 0.14285714285714285,
+      "grad_norm": 3.2002384662628174,
+      "kl": 0.00543212890625,
+      "learning_rate": 8.57142857142857e-07,
+      "loss": 0.0005,
+      "reward": 1.671875,
+      "reward_mean": 1.671875,
+      "reward_std": 0.2109457552433014,
+      "rewards/accuracy_reward": 0.671875,
+      "rewards/format_reward": 1.0,
+      "step": 23
+    },
+    {
+      "advantages": -1.862645149230957e-09,
+      "completion_length": 75.375,
+      "epoch": 0.14906832298136646,
+      "grad_norm": 5.946903705596924,
+      "kl": 0.0087890625,
+      "learning_rate": 8.509316770186336e-07,
+      "loss": 0.0009,
+      "reward": 1.484375,
+      "reward_mean": 1.484375,
+      "reward_std": 0.19044628739356995,
+      "rewards/accuracy_reward": 0.5,
+      "rewards/format_reward": 0.984375,
+      "step": 24
+    },
+    {
+      "advantages": -9.313225746154785e-10,
+      "completion_length": 68.8125,
+      "epoch": 0.15527950310559005,
+      "grad_norm": 4.977855682373047,
+      "kl": 0.008544921875,
+      "learning_rate": 8.447204968944099e-07,
+      "loss": 0.0009,
+      "reward": 1.734375,
+      "reward_mean": 1.734375,
+      "reward_std": 0.12255740165710449,
+      "rewards/accuracy_reward": 0.734375,
+      "rewards/format_reward": 1.0,
+      "step": 25
+    },
+    {
+      "advantages": -8.381903171539307e-09,
+      "completion_length": 83.59375,
+      "epoch": 0.16149068322981366,
+      "grad_norm": 4.409206390380859,
+      "kl": 0.01239013671875,
+      "learning_rate": 8.385093167701863e-07,
+      "loss": 0.0012,
+      "reward": 1.609375,
+      "reward_mean": 1.609375,
+      "reward_std": 0.2198973000049591,
+      "rewards/accuracy_reward": 0.609375,
+      "rewards/format_reward": 1.0,
+      "step": 26
+    },
+    {
+      "advantages": 1.862645149230957e-09,
+      "completion_length": 78.109375,
+      "epoch": 0.16770186335403728,
+      "grad_norm": 3.1185989379882812,
+      "kl": 0.006072998046875,
+      "learning_rate": 8.322981366459628e-07,
+      "loss": 0.0006,
+      "reward": 1.65625,
+      "reward_mean": 1.65625,
+      "reward_std": 0.10888782143592834,
+      "rewards/accuracy_reward": 0.671875,
+      "rewards/format_reward": 0.984375,
+      "step": 27
+    },
+    {
+      "advantages": 4.6566128730773926e-09,
+      "completion_length": 72.0625,
+      "epoch": 0.17391304347826086,
+      "grad_norm": 4.9565935134887695,
+      "kl": 0.010009765625,
+      "learning_rate": 8.260869565217391e-07,
+      "loss": 0.001,
+      "reward": 1.34375,
+      "reward_mean": 1.34375,
+      "reward_std": 0.16675157845020294,
+      "rewards/accuracy_reward": 0.34375,
+      "rewards/format_reward": 1.0,
+      "step": 28
+    },
+    {
+      "advantages": -4.6566128730773926e-09,
+      "completion_length": 84.953125,
+      "epoch": 0.18012422360248448,
+      "grad_norm": 4.35609769821167,
+      "kl": 0.011962890625,
+      "learning_rate": 8.198757763975155e-07,
+      "loss": 0.0012,
+      "reward": 1.46875,
+      "reward_mean": 1.46875,
+      "reward_std": 0.25513991713523865,
+      "rewards/accuracy_reward": 0.484375,
+      "rewards/format_reward": 0.984375,
+      "step": 29
+    },
+    {
+      "advantages": -9.313225746154785e-10,
+      "completion_length": 85.34375,
+      "epoch": 0.18633540372670807,
+      "grad_norm": 5.767938137054443,
+      "kl": 0.009033203125,
+      "learning_rate": 8.136645962732918e-07,
+      "loss": 0.0009,
+      "reward": 1.609375,
+      "reward_mean": 1.609375,
+      "reward_std": 0.1530819982290268,
+      "rewards/accuracy_reward": 0.609375,
+      "rewards/format_reward": 1.0,
+      "step": 30
+    },
+    {
+      "advantages": -5.587935447692871e-09,
+      "completion_length": 83.5625,
+      "epoch": 0.19254658385093168,
+      "grad_norm": 49.12059783935547,
+      "kl": 0.0091552734375,
+      "learning_rate": 8.074534161490683e-07,
+      "loss": 0.0009,
+      "reward": 1.578125,
+      "reward_mean": 1.578125,
+      "reward_std": 0.10205793380737305,
+      "rewards/accuracy_reward": 0.578125,
+      "rewards/format_reward": 1.0,
+      "step": 31
+    },
+    {
+      "advantages": 4.6566128730773926e-09,
+      "completion_length": 77.734375,
+      "epoch": 0.19875776397515527,
+      "grad_norm": 1.4828208684921265,
+      "kl": 0.00970458984375,
+      "learning_rate": 8.012422360248446e-07,
+      "loss": 0.001,
+      "reward": 1.421875,
+      "reward_mean": 1.421875,
+      "reward_std": 0.0646936446428299,
+      "rewards/accuracy_reward": 0.421875,
+      "rewards/format_reward": 1.0,
+      "step": 32
+    },
+    {
+      "advantages": 1.862645149230957e-09,
+      "completion_length": 78.453125,
+      "epoch": 0.20496894409937888,
+      "grad_norm": 7.876468658447266,
+      "kl": 0.020263671875,
+      "learning_rate": 7.95031055900621e-07,
+      "loss": 0.002,
+      "reward": 1.734375,
+      "reward_mean": 1.734375,
+      "reward_std": 0.2109457552433014,
+      "rewards/accuracy_reward": 0.734375,
+      "rewards/format_reward": 1.0,
+      "step": 33
+    },
+    {
+      "advantages": -3.725290298461914e-09,
+      "completion_length": 80.5,
+      "epoch": 0.2111801242236025,
+      "grad_norm": 3.8213541507720947,
+      "kl": 0.01361083984375,
+      "learning_rate": 7.888198757763976e-07,
+      "loss": 0.0014,
+      "reward": 1.46875,
+      "reward_mean": 1.46875,
+      "reward_std": 0.0578637570142746,
+      "rewards/accuracy_reward": 0.46875,
+      "rewards/format_reward": 1.0,
+      "step": 34
+    },
+    {
+      "advantages": -6.51925802230835e-09,
+      "completion_length": 89.5625,
+      "epoch": 0.21739130434782608,
+      "grad_norm": 3.453101634979248,
+      "kl": 0.01470947265625,
+      "learning_rate": 7.826086956521739e-07,
+      "loss": 0.0015,
+      "reward": 1.75,
+      "reward_mean": 1.75,
+      "reward_std": 0.17570312321186066,
+      "rewards/accuracy_reward": 0.75,
+      "rewards/format_reward": 1.0,
+      "step": 35
+    },
+    {
+      "advantages": 0.0,
+      "completion_length": 90.28125,
+      "epoch": 0.2236024844720497,
+      "grad_norm": 2.642101526260376,
+      "kl": 0.0107421875,
+      "learning_rate": 7.763975155279503e-07,
+      "loss": 0.0011,
+      "reward": 1.5625,
+      "reward_mean": 1.5625,
+      "reward_std": 0.06681530922651291,
+      "rewards/accuracy_reward": 0.5625,
+      "rewards/format_reward": 1.0,
+      "step": 36
+    },
+    {
+      "advantages": -3.725290298461914e-09,
+      "completion_length": 80.3125,
+      "epoch": 0.22981366459627328,
+      "grad_norm": 3.5424673557281494,
+      "kl": 0.01239013671875,
+      "learning_rate": 7.701863354037266e-07,
+      "loss": 0.0012,
+      "reward": 1.71875,
+      "reward_mean": 1.71875,
+      "reward_std": 0.0578637570142746,
+      "rewards/accuracy_reward": 0.71875,
+      "rewards/format_reward": 1.0,
+      "step": 37
+    },
+    {
+      "advantages": 0.0,
+      "completion_length": 84.140625,
+      "epoch": 0.2360248447204969,
+      "grad_norm": 0.38800248503685,
+      "kl": 0.01275634765625,
+      "learning_rate": 7.639751552795031e-07,
+      "loss": 0.0013,
+      "reward": 1.375,
+      "reward_mean": 1.375,
+      "reward_std": 0.0,
+      "rewards/accuracy_reward": 0.390625,
+      "rewards/format_reward": 0.984375,
+      "step": 38
+    },
+    {
+      "advantages": 1.862645149230957e-09,
+      "completion_length": 89.109375,
+      "epoch": 0.2422360248447205,
+      "grad_norm": 2.645474433898926,
+      "kl": 0.01397705078125,
+      "learning_rate": 7.577639751552795e-07,
+      "loss": 0.0014,
+      "reward": 1.515625,
+      "reward_mean": 1.515625,
+      "reward_std": 0.04419417306780815,
+      "rewards/accuracy_reward": 0.515625,
+      "rewards/format_reward": 1.0,
+      "step": 39
+    },
+    {
+      "advantages": -3.725290298461914e-09,
+      "completion_length": 72.296875,
+      "epoch": 0.2484472049689441,
+      "grad_norm": 8.574762344360352,
+      "kl": 0.0159912109375,
+      "learning_rate": 7.515527950310558e-07,
+      "loss": 0.0016,
+      "reward": 1.671875,
+      "reward_mean": 1.671875,
+      "reward_std": 0.23144522309303284,
+      "rewards/accuracy_reward": 0.671875,
+      "rewards/format_reward": 1.0,
+      "step": 40
+    },
+    {
+      "advantages": 7.450580596923828e-09,
+      "completion_length": 86.046875,
+      "epoch": 0.2546583850931677,
+      "grad_norm": 36.26329040527344,
+      "kl": 0.0147705078125,
+      "learning_rate": 7.453416149068323e-07,
+      "loss": 0.0015,
+      "reward": 1.65625,
+      "reward_mean": 1.65625,
+      "reward_std": 0.23356688022613525,
+      "rewards/accuracy_reward": 0.65625,
+      "rewards/format_reward": 1.0,
+      "step": 41
+    },
+    {
+      "advantages": -3.725290298461914e-09,
+      "completion_length": 77.0625,
+      "epoch": 0.2608695652173913,
+      "grad_norm": 10.992830276489258,
+      "kl": 0.0113525390625,
+      "learning_rate": 7.391304347826086e-07,
+      "loss": 0.0011,
+      "reward": 1.703125,
+      "reward_mean": 1.703125,
+      "reward_std": 0.24464011192321777,
+      "rewards/accuracy_reward": 0.703125,
+      "rewards/format_reward": 1.0,
+      "step": 42
+    },
+    {
+      "advantages": -1.862645149230957e-09,
+      "completion_length": 86.53125,
+      "epoch": 0.2670807453416149,
+      "grad_norm": 6.251725673675537,
+      "kl": 0.009033203125,
+      "learning_rate": 7.329192546583851e-07,
+      "loss": 0.0009,
+      "reward": 1.609375,
+      "reward_mean": 1.609375,
+      "reward_std": 0.23144522309303284,
+      "rewards/accuracy_reward": 0.609375,
+      "rewards/format_reward": 1.0,
+      "step": 43
+    },
+    {
+      "advantages": -4.6566128730773926e-09,
+      "completion_length": 86.4375,
+      "epoch": 0.2732919254658385,
+      "grad_norm": 3.8048486709594727,
+      "kl": 0.01385498046875,
+      "learning_rate": 7.267080745341615e-07,
+      "loss": 0.0014,
+      "reward": 1.765625,
+      "reward_mean": 1.765625,
+      "reward_std": 0.17358146607875824,
+      "rewards/accuracy_reward": 0.765625,
+      "rewards/format_reward": 1.0,
+      "step": 44
+    },
+    {
+      "advantages": -1.862645149230957e-09,
+      "completion_length": 84.21875,
+      "epoch": 0.2795031055900621,
+      "grad_norm": 2.5062499046325684,
+      "kl": 0.00811767578125,
+      "learning_rate": 7.204968944099379e-07,
+      "loss": 0.0008,
+      "reward": 1.796875,
+      "reward_mean": 1.796875,
+      "reward_std": 0.11100947856903076,
+      "rewards/accuracy_reward": 0.8125,
+      "rewards/format_reward": 0.984375,
+      "step": 45
+    },
+    {
+      "advantages": 3.725290298461914e-09,
+      "completion_length": 77.9375,
+      "epoch": 0.2857142857142857,
+      "grad_norm": 3.8415560722351074,
+      "kl": 0.01165771484375,
+      "learning_rate": 7.142857142857143e-07,
+      "loss": 0.0012,
+      "reward": 1.53125,
+      "reward_mean": 1.53125,
+      "reward_std": 0.1462521106004715,
+      "rewards/accuracy_reward": 0.53125,
+      "rewards/format_reward": 1.0,
+      "step": 46
+    },
+    {
+      "advantages": 1.862645149230957e-09,
+      "completion_length": 82.6875,
+      "epoch": 0.2919254658385093,
+      "grad_norm": 2.903069496154785,
+      "kl": 0.012451171875,
+      "learning_rate": 7.080745341614906e-07,
+      "loss": 0.0012,
+      "reward": 1.578125,
+      "reward_mean": 1.578125,
+      "reward_std": 0.11100947856903076,
+      "rewards/accuracy_reward": 0.59375,
+      "rewards/format_reward": 0.984375,
+      "step": 47
+    },
+    {
+      "advantages": -2.7939677238464355e-09,
+      "completion_length": 75.578125,
+      "epoch": 0.2981366459627329,
+      "grad_norm": 11.884781837463379,
+      "kl": 0.0125732421875,
+      "learning_rate": 7.018633540372671e-07,
+      "loss": 0.0013,
+      "reward": 1.65625,
+      "reward_mean": 1.65625,
+      "reward_std": 0.17570312321186066,
+      "rewards/accuracy_reward": 0.65625,
+      "rewards/format_reward": 1.0,
+      "step": 48
+    },
+    {
+      "advantages": -1.862645149230957e-09,
+      "completion_length": 73.34375,
+      "epoch": 0.30434782608695654,
+      "grad_norm": 2.234876871109009,
+      "kl": 0.0084228515625,
+      "learning_rate": 6.956521739130434e-07,
+      "loss": 0.0008,
+      "reward": 1.484375,
+      "reward_mean": 1.484375,
+      "reward_std": 0.04419417306780815,
+      "rewards/accuracy_reward": 0.484375,
+      "rewards/format_reward": 1.0,
+      "step": 49
+    },
+    {
+      "advantages": 1.862645149230957e-09,
+      "completion_length": 81.8125,
+      "epoch": 0.3105590062111801,
+      "grad_norm": 4.401739597320557,
+      "kl": 0.007232666015625,
+      "learning_rate": 6.894409937888198e-07,
+      "loss": 0.0007,
+      "reward": 1.765625,
+      "reward_mean": 1.765625,
+      "reward_std": 0.17782479524612427,
+      "rewards/accuracy_reward": 0.765625,
+      "rewards/format_reward": 1.0,
+      "step": 50
+    },
+    {
+      "advantages": 0.0,
+      "completion_length": 84.015625,
+      "epoch": 0.3167701863354037,
+      "grad_norm": 0.28293830156326294,
+      "kl": 0.0062255859375,
+      "learning_rate": 6.832298136645962e-07,
+      "loss": 0.0006,
+      "reward": 2.0,
+      "reward_mean": 2.0,
+      "reward_std": 0.0,
+      "rewards/accuracy_reward": 1.0,
+      "rewards/format_reward": 1.0,
+      "step": 51
+    },
+    {
+      "advantages": 0.0,
+      "completion_length": 79.125,
+      "epoch": 0.32298136645962733,
+      "grad_norm": 2.2039246559143066,
+      "kl": 0.0106201171875,
+      "learning_rate": 6.770186335403726e-07,
+      "loss": 0.0011,
+      "reward": 1.75,
+      "reward_mean": 1.75,
+      "reward_std": 0.0883883461356163,
+      "rewards/accuracy_reward": 0.75,
+      "rewards/format_reward": 1.0,
+      "step": 52
+    },
+    {
+      "advantages": 9.313225746154785e-10,
+      "completion_length": 76.0625,
+      "epoch": 0.32919254658385094,
+      "grad_norm": 4.176709175109863,
+      "kl": 0.01123046875,
+      "learning_rate": 6.708074534161491e-07,
+      "loss": 0.0011,
+      "reward": 1.640625,
+      "reward_mean": 1.640625,
+      "reward_std": 0.1530819982290268,
+      "rewards/accuracy_reward": 0.640625,
+      "rewards/format_reward": 1.0,
+      "step": 53
+    },
+    {
+      "advantages": -3.725290298461914e-09,
+      "completion_length": 81.125,
+      "epoch": 0.33540372670807456,
+      "grad_norm": 30.12848663330078,
+      "kl": 0.099609375,
+      "learning_rate": 6.645962732919254e-07,
+      "loss": 0.01,
+      "reward": 1.71875,
+      "reward_mean": 1.71875,
+      "reward_std": 0.1462520956993103,
+      "rewards/accuracy_reward": 0.71875,
+      "rewards/format_reward": 1.0,
+      "step": 54
+    },
+    {
+      "advantages": -1.862645149230957e-09,
+      "completion_length": 80.609375,
+      "epoch": 0.3416149068322981,
+      "grad_norm": 12.808406829833984,
+      "kl": 0.01416015625,
+      "learning_rate": 6.583850931677019e-07,
+      "loss": 0.0014,
+      "reward": 1.6875,
+      "reward_mean": 1.6875,
+      "reward_std": 0.2238783985376358,
+      "rewards/accuracy_reward": 0.703125,
+      "rewards/format_reward": 0.984375,
+      "step": 55
+    },
+    {
+      "advantages": -5.587935447692871e-09,
+      "completion_length": 76.15625,
+      "epoch": 0.34782608695652173,
+      "grad_norm": 5.750046253204346,
+      "kl": 0.01019287109375,
+      "learning_rate": 6.521739130434782e-07,
+      "loss": 0.001,
+      "reward": 1.5,
+      "reward_mean": 1.5,
+      "reward_std": 0.2041158676147461,
+      "rewards/accuracy_reward": 0.515625,
+      "rewards/format_reward": 0.984375,
+      "step": 56
+    },
+    {
+      "advantages": -3.725290298461914e-09,
+      "completion_length": 76.765625,
+      "epoch": 0.35403726708074534,
+      "grad_norm": 4.7853102684021,
+      "kl": 0.010986328125,
+      "learning_rate": 6.459627329192546e-07,
+      "loss": 0.0011,
+      "reward": 1.328125,
+      "reward_mean": 1.328125,
+      "reward_std": 0.19044628739356995,
+      "rewards/accuracy_reward": 0.34375,
+      "rewards/format_reward": 0.984375,
+      "step": 57
+    },
+    {
+      "advantages": 4.6566128730773926e-09,
+      "completion_length": 88.796875,
+      "epoch": 0.36024844720496896,
+      "grad_norm": 1.7344197034835815,
+      "kl": 0.00982666015625,
+      "learning_rate": 6.39751552795031e-07,
+      "loss": 0.001,
+      "reward": 1.671875,
+      "reward_mean": 1.671875,
+      "reward_std": 0.0646936446428299,
+      "rewards/accuracy_reward": 0.671875,
+      "rewards/format_reward": 1.0,
+      "step": 58
+    },
+    {
+      "advantages": -9.313225746154785e-09,
+      "completion_length": 84.28125,
+      "epoch": 0.36645962732919257,
+      "grad_norm": 3.1260499954223633,
+      "kl": 0.01416015625,
+      "learning_rate": 6.335403726708074e-07,
+      "loss": 0.0014,
+      "reward": 1.6875,
+      "reward_mean": 1.6875,
+      "reward_std": 0.1828794628381729,
+      "rewards/accuracy_reward": 0.703125,
+      "rewards/format_reward": 0.984375,
+      "step": 59
+    },
+    {
+      "advantages": -3.725290298461914e-09,
+      "completion_length": 77.21875,
+      "epoch": 0.37267080745341613,
+      "grad_norm": 1.7963190078735352,
+      "kl": 0.0089111328125,
+      "learning_rate": 6.273291925465838e-07,
+      "loss": 0.0009,
+      "reward": 1.84375,
+      "reward_mean": 1.84375,
+      "reward_std": 0.0578637570142746,
+      "rewards/accuracy_reward": 0.84375,
+      "rewards/format_reward": 1.0,
+      "step": 60
+    },
+    {
+      "advantages": 0.0,
+      "completion_length": 82.25,
+      "epoch": 0.37888198757763975,
+      "grad_norm": 2.5049538612365723,
+      "kl": 0.00787353515625,
+      "learning_rate": 6.211180124223601e-07,
+      "loss": 0.0008,
+      "reward": 1.625,
+      "reward_mean": 1.625,
+      "reward_std": 0.0883883461356163,
+      "rewards/accuracy_reward": 0.625,
+      "rewards/format_reward": 1.0,
+      "step": 61
+    },
+    {
+      "advantages": -1.210719347000122e-08,
+      "completion_length": 80.375,
+      "epoch": 0.38509316770186336,
+      "grad_norm": 5.739541530609131,
+      "kl": 0.01312255859375,
+      "learning_rate": 6.149068322981367e-07,
+      "loss": 0.0013,
+      "reward": 1.75,
+      "reward_mean": 1.75,
+      "reward_std": 0.2177756428718567,
+      "rewards/accuracy_reward": 0.75,
+      "rewards/format_reward": 1.0,
+      "step": 62
+    },
+    {
+      "advantages": 1.862645149230957e-09,
+      "completion_length": 84.296875,
+      "epoch": 0.391304347826087,
+      "grad_norm": 4.335031032562256,
+      "kl": 0.01116943359375,
+      "learning_rate": 6.08695652173913e-07,
+      "loss": 0.0011,
+      "reward": 1.90625,
+      "reward_mean": 1.90625,
+      "reward_std": 0.2041158676147461,
+      "rewards/accuracy_reward": 0.90625,
+      "rewards/format_reward": 1.0,
+      "step": 63
+    },
+    {
+      "advantages": 1.862645149230957e-09,
+      "completion_length": 86.828125,
+      "epoch": 0.39751552795031053,
+      "grad_norm": 4.443232536315918,
+      "kl": 0.01336669921875,
+      "learning_rate": 6.024844720496894e-07,
+      "loss": 0.0013,
+      "reward": 1.703125,
+      "reward_mean": 1.703125,
+      "reward_std": 0.19939783215522766,
+      "rewards/accuracy_reward": 0.703125,
+      "rewards/format_reward": 1.0,
+      "step": 64
+    },
+    {
+      "advantages": 5.587935447692871e-09,
+      "completion_length": 75.71875,
+      "epoch": 0.40372670807453415,
+      "grad_norm": 7.092515468597412,
+      "kl": 0.01251220703125,
+      "learning_rate": 5.962732919254659e-07,
+      "loss": 0.0013,
+      "reward": 1.65625,
+      "reward_mean": 1.65625,
+      "reward_std": 0.23827511072158813,
+      "rewards/accuracy_reward": 0.65625,
+      "rewards/format_reward": 1.0,
+      "step": 65
+    },
+    {
+      "advantages": 4.6566128730773926e-09,
+      "completion_length": 82.140625,
+      "epoch": 0.40993788819875776,
+      "grad_norm": 4.468729496002197,
+      "kl": 0.0211181640625,
+      "learning_rate": 5.900621118012422e-07,
+      "loss": 0.0021,
+      "reward": 1.796875,
+      "reward_mean": 1.796875,
+      "reward_std": 0.0646936446428299,
+      "rewards/accuracy_reward": 0.796875,
+      "rewards/format_reward": 1.0,
+      "step": 66
+    },
+    {
+      "advantages": 4.6566128730773926e-09,
+      "completion_length": 74.890625,
+      "epoch": 0.4161490683229814,
+      "grad_norm": 9.289567947387695,
+      "kl": 0.01611328125,
+      "learning_rate": 5.838509316770186e-07,
+      "loss": 0.0016,
+      "reward": 1.421875,
+      "reward_mean": 1.421875,
+      "reward_std": 0.1983242630958557,
+      "rewards/accuracy_reward": 0.421875,
+      "rewards/format_reward": 1.0,
+      "step": 67
+    },
+    {
+      "advantages": 0.0,
+      "completion_length": 77.625,
+      "epoch": 0.422360248447205,
+      "grad_norm": 0.4326918125152588,
+      "kl": 0.0140380859375,
+      "learning_rate": 5.77639751552795e-07,
+      "loss": 0.0014,
+      "reward": 1.875,
+      "reward_mean": 1.875,
+      "reward_std": 0.0,
+      "rewards/accuracy_reward": 0.875,
+      "rewards/format_reward": 1.0,
+      "step": 68
+    },
+    {
+      "advantages": -9.313225746154785e-10,
+      "completion_length": 81.71875,
+      "epoch": 0.42857142857142855,
+      "grad_norm": 5.539842128753662,
+      "kl": 0.04296875,
+      "learning_rate": 5.714285714285714e-07,
+      "loss": 0.0043,
+      "reward": 1.4375,
+      "reward_mean": 1.4375,
+      "reward_std": 0.34352827072143555,
+      "rewards/accuracy_reward": 0.453125,
+      "rewards/format_reward": 0.984375,
+      "step": 69
+    },
+    {
+      "advantages": 0.0,
+      "completion_length": 90.703125,
+      "epoch": 0.43478260869565216,
+      "grad_norm": 0.46686819195747375,
+      "kl": 0.0074462890625,
+      "learning_rate": 5.652173913043477e-07,
+      "loss": 0.0007,
+      "reward": 1.875,
+      "reward_mean": 1.875,
+      "reward_std": 0.0,
+      "rewards/accuracy_reward": 0.875,
+      "rewards/format_reward": 1.0,
+      "step": 70
+    },
+    {
+      "advantages": 1.862645149230957e-09,
+      "completion_length": 83.578125,
+      "epoch": 0.4409937888198758,
+      "grad_norm": 8.54028606414795,
+      "kl": 0.00897216796875,
+      "learning_rate": 5.590062111801241e-07,
+      "loss": 0.0009,
+      "reward": 1.765625,
+      "reward_mean": 1.765625,
+      "reward_std": 0.04419417306780815,
+      "rewards/accuracy_reward": 0.765625,
+      "rewards/format_reward": 1.0,
+      "step": 71
+    },
+    {
+      "advantages": 7.450580596923828e-09,
+      "completion_length": 84.25,
+      "epoch": 0.4472049689440994,
+      "grad_norm": 12.895256996154785,
+      "kl": 0.00579833984375,
+      "learning_rate": 5.527950310559007e-07,
+      "loss": 0.0006,
+      "reward": 1.453125,
+      "reward_mean": 1.453125,
+      "reward_std": 0.12255740165710449,
+      "rewards/accuracy_reward": 0.453125,
+      "rewards/format_reward": 1.0,
+      "step": 72
+    },
+    {
+      "advantages": -9.313225746154785e-09,
+      "completion_length": 77.15625,
+      "epoch": 0.453416149068323,
+      "grad_norm": 5.548634052276611,
+      "kl": 0.0123291015625,
+      "learning_rate": 5.46583850931677e-07,
+      "loss": 0.0012,
+      "reward": 1.796875,
+      "reward_mean": 1.796875,
+      "reward_std": 0.31983357667922974,
+      "rewards/accuracy_reward": 0.796875,
+      "rewards/format_reward": 1.0,
+      "step": 73
+    },
+    {
+      "advantages": -1.0244548320770264e-08,
+      "completion_length": 84.75,
+      "epoch": 0.45962732919254656,
+      "grad_norm": 3.4154112339019775,
+      "kl": 0.018798828125,
+      "learning_rate": 5.403726708074534e-07,
+      "loss": 0.0019,
+      "reward": 1.78125,
+      "reward_mean": 1.78125,
+      "reward_std": 0.19727616012096405,
+      "rewards/accuracy_reward": 0.78125,
+      "rewards/format_reward": 1.0,
+      "step": 74
+    },
+    {
+      "advantages": -4.6566128730773926e-09,
+      "completion_length": 83.015625,
+      "epoch": 0.4658385093167702,
+      "grad_norm": 3.4328691959381104,
+      "kl": 0.01275634765625,
+      "learning_rate": 5.341614906832298e-07,
+      "loss": 0.0013,
+      "reward": 1.53125,
+      "reward_mean": 1.53125,
+      "reward_std": 0.23356688022613525,
+      "rewards/accuracy_reward": 0.53125,
+      "rewards/format_reward": 1.0,
+      "step": 75
+    },
+    {
+      "advantages": -1.862645149230957e-09,
+      "completion_length": 78.609375,
+      "epoch": 0.4720496894409938,
+      "grad_norm": 3.627190113067627,
+      "kl": 0.0142822265625,
+      "learning_rate": 5.279503105590062e-07,
+      "loss": 0.0014,
+      "reward": 1.9375,
+      "reward_mean": 1.9375,
+      "reward_std": 0.1462521106004715,
+      "rewards/accuracy_reward": 0.9375,
+      "rewards/format_reward": 1.0,
+      "step": 76
+    },
+    {
+      "advantages": -3.725290298461914e-09,
+      "completion_length": 80.46875,
+      "epoch": 0.4782608695652174,
+      "grad_norm": 10.168981552124023,
+      "kl": 0.01251220703125,
+      "learning_rate": 5.217391304347825e-07,
+      "loss": 0.0013,
+      "reward": 1.515625,
+      "reward_mean": 1.515625,
+      "reward_std": 0.2109457552433014,
+      "rewards/accuracy_reward": 0.515625,
+      "rewards/format_reward": 1.0,
+      "step": 77
+    },
+    {
+      "advantages": 0.0,
+      "completion_length": 83.421875,
+      "epoch": 0.484472049689441,
+      "grad_norm": 20.923242568969727,
+      "kl": 0.0113525390625,
+      "learning_rate": 5.15527950310559e-07,
+      "loss": 0.0011,
+      "reward": 1.828125,
+      "reward_mean": 1.828125,
+      "reward_std": 0.19044628739356995,
+      "rewards/accuracy_reward": 0.828125,
+      "rewards/format_reward": 1.0,
+      "step": 78
+    },
+    {
+      "advantages": 2.7939677238464355e-09,
+      "completion_length": 75.109375,
+      "epoch": 0.4906832298136646,
+      "grad_norm": 3.643770933151245,
+      "kl": 0.00811767578125,
+      "learning_rate": 5.093167701863354e-07,
+      "loss": 0.0008,
+      "reward": 1.78125,
+      "reward_mean": 1.78125,
+      "reward_std": 0.10888782143592834,
+      "rewards/accuracy_reward": 0.78125,
+      "rewards/format_reward": 1.0,
+      "step": 79
+    },
+    {
+      "advantages": 1.862645149230957e-09,
+      "completion_length": 81.65625,
+      "epoch": 0.4968944099378882,
+      "grad_norm": 4.883938312530518,
+      "kl": 0.015625,
+      "learning_rate": 5.031055900621117e-07,
+      "loss": 0.0016,
+      "reward": 1.25,
+      "reward_mean": 1.25,
+      "reward_std": 0.2130674123764038,
+      "rewards/accuracy_reward": 0.25,
+      "rewards/format_reward": 1.0,
+      "step": 80
+    },
+    {
+      "advantages": -3.725290298461914e-09,
+      "completion_length": 82.515625,
+      "epoch": 0.5031055900621118,
+      "grad_norm": 1.3860398530960083,
+      "kl": 0.00799560546875,
+      "learning_rate": 4.968944099378881e-07,
+      "loss": 0.0008,
+      "reward": 1.71875,
+      "reward_mean": 1.71875,
+      "reward_std": 0.0578637570142746,
+      "rewards/accuracy_reward": 0.71875,
+      "rewards/format_reward": 1.0,
+      "step": 81
+    },
+    {
+      "advantages": 9.313225746154785e-10,
+      "completion_length": 77.6875,
+      "epoch": 0.5093167701863354,
+      "grad_norm": 3.7328872680664062,
+      "kl": 0.0181884765625,
+      "learning_rate": 4.906832298136646e-07,
+      "loss": 0.0018,
+      "reward": 1.75,
+      "reward_mean": 1.75,
+      "reward_std": 0.16675157845020294,
+      "rewards/accuracy_reward": 0.75,
+      "rewards/format_reward": 1.0,
+      "step": 82
+    },
+    {
+      "advantages": 3.725290298461914e-09,
+      "completion_length": 76.4375,
+      "epoch": 0.515527950310559,
+      "grad_norm": 3.6228644847869873,
+      "kl": 0.01446533203125,
+      "learning_rate": 4.84472049689441e-07,
+      "loss": 0.0014,
+      "reward": 1.46875,
+      "reward_mean": 1.46875,
+      "reward_std": 0.1246790662407875,
+      "rewards/accuracy_reward": 0.46875,
+      "rewards/format_reward": 1.0,
+      "step": 83
+    },
+    {
+      "advantages": -1.862645149230957e-09,
+      "completion_length": 79.140625,
+      "epoch": 0.5217391304347826,
+      "grad_norm": 5.579171180725098,
+      "kl": 0.0159912109375,
+      "learning_rate": 4.782608695652174e-07,
+      "loss": 0.0016,
+      "reward": 1.65625,
+      "reward_mean": 1.65625,
+      "reward_std": 0.23356688022613525,
+      "rewards/accuracy_reward": 0.671875,
+      "rewards/format_reward": 0.984375,
+      "step": 84
+    },
+    {
+      "advantages": -5.587935447692871e-09,
+      "completion_length": 80.0,
+      "epoch": 0.5279503105590062,
+      "grad_norm": 9.611387252807617,
+      "kl": 0.01080322265625,
+      "learning_rate": 4.7204968944099376e-07,
+      "loss": 0.0011,
+      "reward": 1.828125,
+      "reward_mean": 1.828125,
+      "reward_std": 0.13258251547813416,
+      "rewards/accuracy_reward": 0.828125,
+      "rewards/format_reward": 1.0,
+      "step": 85
+    },
+    {
+      "advantages": -1.862645149230957e-09,
+      "completion_length": 85.5,
+      "epoch": 0.5341614906832298,
+      "grad_norm": 4.1448540687561035,
+      "kl": 0.01007080078125,
+      "learning_rate": 4.6583850931677014e-07,
+      "loss": 0.001,
+      "reward": 1.859375,
+      "reward_mean": 1.859375,
+      "reward_std": 0.17358146607875824,
+      "rewards/accuracy_reward": 0.859375,
+      "rewards/format_reward": 1.0,
+      "step": 86
+    },
+    {
+      "advantages": 1.862645149230957e-09,
+      "completion_length": 75.5,
+      "epoch": 0.5403726708074534,
+      "grad_norm": 5.654483795166016,
+      "kl": 0.01123046875,
+      "learning_rate": 4.596273291925465e-07,
+      "loss": 0.0011,
+      "reward": 1.796875,
+      "reward_mean": 1.796875,
+      "reward_std": 0.1530819982290268,
+      "rewards/accuracy_reward": 0.796875,
+      "rewards/format_reward": 1.0,
+      "step": 87
+    },
+    {
+      "advantages": -3.725290298461914e-09,
+      "completion_length": 72.671875,
+      "epoch": 0.546583850931677,
+      "grad_norm": 2.2370052337646484,
+      "kl": 0.0137939453125,
+      "learning_rate": 4.53416149068323e-07,
+      "loss": 0.0014,
+      "reward": 1.46875,
+      "reward_mean": 1.46875,
+      "reward_std": 0.0883883461356163,
+      "rewards/accuracy_reward": 0.46875,
+      "rewards/format_reward": 1.0,
+      "step": 88
+    },
+    {
+      "advantages": 0.0,
+      "completion_length": 81.84375,
+      "epoch": 0.5527950310559007,
+      "grad_norm": 1.389394760131836,
+      "kl": 0.00836181640625,
+      "learning_rate": 4.472049689440994e-07,
+      "loss": 0.0008,
+      "reward": 1.75,
+      "reward_mean": 1.75,
+      "reward_std": 0.06681530922651291,
+      "rewards/accuracy_reward": 0.765625,
+      "rewards/format_reward": 0.984375,
+      "step": 89
+    },
+    {
+      "advantages": 0.0,
+      "completion_length": 74.59375,
+      "epoch": 0.5590062111801242,
+      "grad_norm": 2.353760242462158,
+      "kl": 0.00811767578125,
+      "learning_rate": 4.4099378881987576e-07,
+      "loss": 0.0008,
+      "reward": 1.6875,
+      "reward_mean": 1.6875,
+      "reward_std": 0.06681530922651291,
+      "rewards/accuracy_reward": 0.6875,
+      "rewards/format_reward": 1.0,
+      "step": 90
+    },
+    {
+      "advantages": 0.0,
+      "completion_length": 85.28125,
+      "epoch": 0.5652173913043478,
+      "grad_norm": 1.5767848491668701,
+      "kl": 0.009765625,
+      "learning_rate": 4.3478260869565214e-07,
+      "loss": 0.001,
+      "reward": 1.75,
+      "reward_mean": 1.75,
+      "reward_std": 0.0,
+      "rewards/accuracy_reward": 0.75,
+      "rewards/format_reward": 1.0,
+      "step": 91
+    },
+    {
+      "advantages": -8.381903171539307e-09,
+      "completion_length": 81.859375,
+      "epoch": 0.5714285714285714,
+      "grad_norm": 3.835320234298706,
+      "kl": 0.0181884765625,
+      "learning_rate": 4.285714285714285e-07,
+      "loss": 0.0018,
+      "reward": 1.671875,
+      "reward_mean": 1.671875,
+      "reward_std": 0.1530819982290268,
+      "rewards/accuracy_reward": 0.6875,
+      "rewards/format_reward": 0.984375,
+      "step": 92
+    },
+    {
+      "advantages": 3.725290298461914e-09,
+      "completion_length": 83.671875,
+      "epoch": 0.577639751552795,
+      "grad_norm": 9.30271053314209,
+      "kl": 0.017822265625,
+      "learning_rate": 4.2236024844720495e-07,
+      "loss": 0.0018,
+      "reward": 1.796875,
+      "reward_mean": 1.796875,
+      "reward_std": 0.23144522309303284,
+      "rewards/accuracy_reward": 0.796875,
+      "rewards/format_reward": 1.0,
+      "step": 93
+    },
+    {
+      "advantages": -4.6566128730773926e-09,
+      "completion_length": 77.53125,
+      "epoch": 0.5838509316770186,
+      "grad_norm": 6.170975685119629,
+      "kl": 0.009521484375,
+      "learning_rate": 4.161490683229814e-07,
+      "loss": 0.001,
+      "reward": 1.65625,
+      "reward_mean": 1.65625,
+      "reward_std": 0.16675157845020294,
+      "rewards/accuracy_reward": 0.65625,
+      "rewards/format_reward": 1.0,
+      "step": 94
+    },
+    {
+      "advantages": -1.862645149230957e-09,
+      "completion_length": 85.640625,
+      "epoch": 0.5900621118012422,
+      "grad_norm": 4.217593669891357,
+      "kl": 0.01409912109375,
+      "learning_rate": 4.0993788819875776e-07,
+      "loss": 0.0014,
+      "reward": 1.734375,
+      "reward_mean": 1.734375,
+      "reward_std": 0.15992169082164764,
+      "rewards/accuracy_reward": 0.75,
+      "rewards/format_reward": 0.984375,
+      "step": 95
+    },
+    {
+      "advantages": 0.0,
+      "completion_length": 77.296875,
+      "epoch": 0.5962732919254659,
+      "grad_norm": 6.138365268707275,
+      "kl": 0.0106201171875,
+      "learning_rate": 4.0372670807453413e-07,
+      "loss": 0.0011,
+      "reward": 1.375,
+      "reward_mean": 1.375,
+      "reward_std": 0.06681530922651291,
+      "rewards/accuracy_reward": 0.390625,
+      "rewards/format_reward": 0.984375,
+      "step": 96
+    },
+    {
+      "advantages": 1.862645149230957e-09,
+      "completion_length": 76.484375,
+      "epoch": 0.6024844720496895,
+      "grad_norm": 1.2896429300308228,
+      "kl": 0.00970458984375,
+      "learning_rate": 3.975155279503105e-07,
+      "loss": 0.001,
+      "reward": 1.71875,
+      "reward_mean": 1.71875,
+      "reward_std": 0.0883883461356163,
+      "rewards/accuracy_reward": 0.734375,
+      "rewards/format_reward": 0.984375,
+      "step": 97
+    },
+    {
+      "advantages": -3.725290298461914e-09,
+      "completion_length": 81.71875,
+      "epoch": 0.6086956521739131,
+      "grad_norm": 6.941093444824219,
+      "kl": 0.01165771484375,
+      "learning_rate": 3.9130434782608694e-07,
+      "loss": 0.0012,
+      "reward": 1.71875,
+      "reward_mean": 1.71875,
+      "reward_std": 0.0578637570142746,
+      "rewards/accuracy_reward": 0.71875,
+      "rewards/format_reward": 1.0,
+      "step": 98
+    },
+    {
+      "advantages": -3.725290298461914e-09,
+      "completion_length": 81.390625,
+      "epoch": 0.6149068322981367,
+      "grad_norm": 3.163457155227661,
+      "kl": 0.00787353515625,
+      "learning_rate": 3.850931677018633e-07,
+      "loss": 0.0008,
+      "reward": 1.96875,
+      "reward_mean": 1.96875,
+      "reward_std": 0.0883883461356163,
+      "rewards/accuracy_reward": 0.96875,
+      "rewards/format_reward": 1.0,
+      "step": 99
+    },
+    {
+      "advantages": 4.6566128730773926e-09,
+      "completion_length": 79.1875,
+      "epoch": 0.6211180124223602,
+      "grad_norm": 4.2669830322265625,
+      "kl": 0.0108642578125,
+      "learning_rate": 3.7888198757763975e-07,
+      "loss": 0.0011,
+      "reward": 1.671875,
+      "reward_mean": 1.671875,
+      "reward_std": 0.0646936446428299,
+      "rewards/accuracy_reward": 0.671875,
+      "rewards/format_reward": 1.0,
+      "step": 100
+    },
+    {
+      "advantages": -1.862645149230957e-09,
+      "completion_length": 77.578125,
+      "epoch": 0.6273291925465838,
+      "grad_norm": 6.153615474700928,
+      "kl": 0.0101318359375,
+      "learning_rate": 3.7267080745341613e-07,
+      "loss": 0.001,
+      "reward": 1.359375,
+      "reward_mean": 1.359375,
+      "reward_std": 0.04419417306780815,
+      "rewards/accuracy_reward": 0.359375,
+      "rewards/format_reward": 1.0,
+      "step": 101
+    },
+    {
+      "advantages": -1.862645149230957e-09,
+      "completion_length": 81.53125,
+      "epoch": 0.6335403726708074,
+      "grad_norm": 4.077609539031982,
+      "kl": 0.0181884765625,
+      "learning_rate": 3.6645962732919256e-07,
+      "loss": 0.0018,
+      "reward": 1.84375,
+      "reward_mean": 1.84375,
+      "reward_std": 0.2177756428718567,
+      "rewards/accuracy_reward": 0.84375,
+      "rewards/format_reward": 1.0,
+      "step": 102
+    },
+    {
+      "advantages": -3.725290298461914e-09,
+      "completion_length": 80.375,
+      "epoch": 0.639751552795031,
+      "grad_norm": 3.084027051925659,
+      "kl": 0.01007080078125,
+      "learning_rate": 3.6024844720496894e-07,
+      "loss": 0.001,
+      "reward": 1.53125,
+      "reward_mean": 1.53125,
+      "reward_std": 0.1462521106004715,
+      "rewards/accuracy_reward": 0.53125,
+      "rewards/format_reward": 1.0,
+      "step": 103
+    },
+    {
+      "advantages": -3.725290298461914e-09,
+      "completion_length": 83.953125,
+      "epoch": 0.6459627329192547,
+      "grad_norm": 2.0512335300445557,
+      "kl": 0.007476806640625,
+      "learning_rate": 3.540372670807453e-07,
+      "loss": 0.0007,
+      "reward": 1.453125,
+      "reward_mean": 1.453125,
+      "reward_std": 0.0646936446428299,
+      "rewards/accuracy_reward": 0.453125,
+      "rewards/format_reward": 1.0,
+      "step": 104
+    },
+    {
+      "advantages": 0.0,
+      "completion_length": 82.5625,
+      "epoch": 0.6521739130434783,
+      "grad_norm": 0.5302374362945557,
+      "kl": 0.00982666015625,
+      "learning_rate": 3.478260869565217e-07,
+      "loss": 0.001,
+      "reward": 1.75,
+      "reward_mean": 1.75,
+      "reward_std": 0.0,
+      "rewards/accuracy_reward": 0.75,
+      "rewards/format_reward": 1.0,
+      "step": 105
+    },
+    {
+      "advantages": -4.6566128730773926e-09,
+      "completion_length": 75.953125,
+      "epoch": 0.6583850931677019,
+      "grad_norm": 6.2678751945495605,
+      "kl": 0.0111083984375,
+      "learning_rate": 3.416149068322981e-07,
+      "loss": 0.0011,
+      "reward": 1.890625,
+      "reward_mean": 1.890625,
+      "reward_std": 0.1315089464187622,
+      "rewards/accuracy_reward": 0.890625,
+      "rewards/format_reward": 1.0,
+      "step": 106
+    },
+    {
+      "advantages": -3.725290298461914e-09,
+      "completion_length": 78.328125,
+      "epoch": 0.6645962732919255,
+      "grad_norm": 1.7859537601470947,
+      "kl": 0.00946044921875,
+      "learning_rate": 3.3540372670807456e-07,
+      "loss": 0.0009,
+      "reward": 1.71875,
+      "reward_mean": 1.71875,
+      "reward_std": 0.0883883461356163,
+      "rewards/accuracy_reward": 0.71875,
+      "rewards/format_reward": 1.0,
+      "step": 107
+    },
+    {
+      "advantages": -3.725290298461914e-09,
+      "completion_length": 91.8125,
+      "epoch": 0.6708074534161491,
+      "grad_norm": 2.7167623043060303,
+      "kl": 0.0081787109375,
+      "learning_rate": 3.2919254658385094e-07,
+      "loss": 0.0008,
+      "reward": 1.65625,
+      "reward_mean": 1.65625,
+      "reward_std": 0.0883883461356163,
+      "rewards/accuracy_reward": 0.671875,
+      "rewards/format_reward": 0.984375,
+      "step": 108
+    },
+    {
+      "advantages": 0.0,
+      "completion_length": 76.984375,
+      "epoch": 0.6770186335403726,
+      "grad_norm": 5.3628058433532715,
+      "kl": 0.009033203125,
+      "learning_rate": 3.229813664596273e-07,
+      "loss": 0.0009,
+      "reward": 1.515625,
+      "reward_mean": 1.515625,
+      "reward_std": 0.19044628739356995,
+      "rewards/accuracy_reward": 0.515625,
+      "rewards/format_reward": 1.0,
+      "step": 109
+    },
+    {
+      "advantages": -1.862645149230957e-09,
+      "completion_length": 73.5,
+      "epoch": 0.6832298136645962,
+      "grad_norm": 3.2727582454681396,
+      "kl": 0.0108642578125,
+      "learning_rate": 3.167701863354037e-07,
+      "loss": 0.0011,
+      "reward": 1.609375,
+      "reward_mean": 1.609375,
+      "reward_std": 0.04419417306780815,
+      "rewards/accuracy_reward": 0.609375,
+      "rewards/format_reward": 1.0,
+      "step": 110
+    },
+    {
+      "advantages": 0.0,
+      "completion_length": 75.75,
+      "epoch": 0.6894409937888198,
+      "grad_norm": 11.552366256713867,
+      "kl": 0.0145263671875,
+      "learning_rate": 3.105590062111801e-07,
+      "loss": 0.0015,
+      "reward": 1.75,
+      "reward_mean": 1.75,
+      "reward_std": 0.1157275140285492,
+      "rewards/accuracy_reward": 0.75,
+      "rewards/format_reward": 1.0,
+      "step": 111
+    },
+    {
+      "advantages": 3.725290298461914e-09,
+      "completion_length": 85.734375,
+      "epoch": 0.6956521739130435,
+      "grad_norm": 6.025736331939697,
+      "kl": 0.01556396484375,
+      "learning_rate": 3.043478260869565e-07,
+      "loss": 0.0016,
+      "reward": 1.59375,
+      "reward_mean": 1.59375,
+      "reward_std": 0.1552036553621292,
+      "rewards/accuracy_reward": 0.59375,
+      "rewards/format_reward": 1.0,
+      "step": 112
+    },
+    {
+      "advantages": -3.725290298461914e-09,
+      "completion_length": 82.734375,
+      "epoch": 0.7018633540372671,
+      "grad_norm": 15.336418151855469,
+      "kl": 0.057373046875,
+      "learning_rate": 2.9813664596273294e-07,
+      "loss": 0.0057,
+      "reward": 1.84375,
+      "reward_mean": 1.84375,
+      "reward_std": 0.0883883461356163,
+      "rewards/accuracy_reward": 0.84375,
+      "rewards/format_reward": 1.0,
+      "step": 113
+    },
+    {
+      "advantages": -3.725290298461914e-09,
+      "completion_length": 78.953125,
+      "epoch": 0.7080745341614907,
+      "grad_norm": 65.76184844970703,
+      "kl": 0.01385498046875,
+      "learning_rate": 2.919254658385093e-07,
+      "loss": 0.0014,
+      "reward": 1.90625,
+      "reward_mean": 1.90625,
+      "reward_std": 0.1552036553621292,
+      "rewards/accuracy_reward": 0.90625,
+      "rewards/format_reward": 1.0,
+      "step": 114
+    },
+    {
+      "advantages": -1.862645149230957e-09,
+      "completion_length": 79.765625,
+      "epoch": 0.7142857142857143,
+      "grad_norm": 3.660456657409668,
+      "kl": 0.0194091796875,
+      "learning_rate": 2.857142857142857e-07,
+      "loss": 0.0019,
+      "reward": 1.59375,
+      "reward_mean": 1.59375,
+      "reward_std": 0.10888782143592834,
+      "rewards/accuracy_reward": 0.59375,
+      "rewards/format_reward": 1.0,
+      "step": 115
+    },
+    {
+      "advantages": -5.587935447692871e-09,
+      "completion_length": 76.34375,
+      "epoch": 0.7204968944099379,
+      "grad_norm": 4.989613056182861,
+      "kl": 0.00787353515625,
+      "learning_rate": 2.7950310559006207e-07,
+      "loss": 0.0008,
+      "reward": 1.828125,
+      "reward_mean": 1.828125,
+      "reward_std": 0.13258251547813416,
+      "rewards/accuracy_reward": 0.84375,
+      "rewards/format_reward": 0.984375,
+      "step": 116
+    },
+    {
+      "advantages": -9.313225746154785e-10,
+      "completion_length": 77.859375,
+      "epoch": 0.7267080745341615,
+      "grad_norm": 2.4932050704956055,
+      "kl": 0.0081787109375,
+      "learning_rate": 2.732919254658385e-07,
+      "loss": 0.0008,
+      "reward": 1.859375,
+      "reward_mean": 1.859375,
+      "reward_std": 0.12255740165710449,
+      "rewards/accuracy_reward": 0.859375,
+      "rewards/format_reward": 1.0,
+      "step": 117
+    },
+    {
+      "advantages": 1.862645149230957e-09,
+      "completion_length": 84.5,
+      "epoch": 0.7329192546583851,
+      "grad_norm": 5.0420732498168945,
+      "kl": 0.01226806640625,
+      "learning_rate": 2.670807453416149e-07,
+      "loss": 0.0012,
+      "reward": 1.640625,
+      "reward_mean": 1.640625,
+      "reward_std": 0.23144522309303284,
+      "rewards/accuracy_reward": 0.640625,
+      "rewards/format_reward": 1.0,
+      "step": 118
+    },
+    {
+      "advantages": -1.862645149230957e-09,
+      "completion_length": 79.703125,
+      "epoch": 0.7391304347826086,
+      "grad_norm": 3.599855899810791,
+      "kl": 0.0086669921875,
+      "learning_rate": 2.6086956521739126e-07,
+      "loss": 0.0009,
+      "reward": 1.484375,
+      "reward_mean": 1.484375,
+      "reward_std": 0.13258251547813416,
+      "rewards/accuracy_reward": 0.484375,
+      "rewards/format_reward": 1.0,
+      "step": 119
+    },
+    {
+      "advantages": -5.587935447692871e-09,
+      "completion_length": 79.1875,
+      "epoch": 0.7453416149068323,
+      "grad_norm": 3.320706605911255,
+      "kl": 0.0133056640625,
+      "learning_rate": 2.546583850931677e-07,
+      "loss": 0.0013,
+      "reward": 1.828125,
+      "reward_mean": 1.828125,
+      "reward_std": 0.10205793380737305,
+      "rewards/accuracy_reward": 0.828125,
+      "rewards/format_reward": 1.0,
+      "step": 120
+    },
+    {
+      "advantages": -8.381903171539307e-09,
+      "completion_length": 95.203125,
+      "epoch": 0.7515527950310559,
+      "grad_norm": 2.8366851806640625,
+      "kl": 0.0064697265625,
+      "learning_rate": 2.4844720496894407e-07,
+      "loss": 0.0006,
+      "reward": 1.671875,
+      "reward_mean": 1.671875,
+      "reward_std": 0.1530819833278656,
+      "rewards/accuracy_reward": 0.671875,
+      "rewards/format_reward": 1.0,
+      "step": 121
+    },
+    {
+      "advantages": 1.862645149230957e-09,
+      "completion_length": 78.765625,
+      "epoch": 0.7577639751552795,
+      "grad_norm": 3.376732587814331,
+      "kl": 0.00860595703125,
+      "learning_rate": 2.422360248447205e-07,
+      "loss": 0.0009,
+      "reward": 1.640625,
+      "reward_mean": 1.640625,
+      "reward_std": 0.10205793380737305,
+      "rewards/accuracy_reward": 0.640625,
+      "rewards/format_reward": 1.0,
+      "step": 122
+    },
+    {
+      "advantages": 1.862645149230957e-09,
+      "completion_length": 79.453125,
+      "epoch": 0.7639751552795031,
+      "grad_norm": 3.5682129859924316,
+      "kl": 0.018798828125,
+      "learning_rate": 2.3602484472049688e-07,
+      "loss": 0.0019,
+      "reward": 1.671875,
+      "reward_mean": 1.671875,
+      "reward_std": 0.1804211586713791,
+      "rewards/accuracy_reward": 0.671875,
+      "rewards/format_reward": 1.0,
+      "step": 123
+    },
+    {
+      "advantages": 4.6566128730773926e-09,
+      "completion_length": 74.96875,
+      "epoch": 0.7701863354037267,
+      "grad_norm": 2.6698434352874756,
+      "kl": 0.006256103515625,
+      "learning_rate": 2.2981366459627326e-07,
+      "loss": 0.0006,
+      "reward": 1.546875,
+      "reward_mean": 1.546875,
+      "reward_std": 0.0646936446428299,
+      "rewards/accuracy_reward": 0.546875,
+      "rewards/format_reward": 1.0,
+      "step": 124
+    },
+    {
+      "advantages": 0.0,
+      "completion_length": 78.421875,
+      "epoch": 0.7763975155279503,
+      "grad_norm": 3.1063811779022217,
+      "kl": 0.01214599609375,
+      "learning_rate": 2.236024844720497e-07,
+      "loss": 0.0012,
+      "reward": 1.765625,
+      "reward_mean": 1.765625,
+      "reward_std": 0.12255740165710449,
+      "rewards/accuracy_reward": 0.765625,
+      "rewards/format_reward": 1.0,
+      "step": 125
+    },
+    {
+      "advantages": 1.862645149230957e-09,
+      "completion_length": 86.703125,
+      "epoch": 0.782608695652174,
+      "grad_norm": 2.7392446994781494,
+      "kl": 0.00634765625,
+      "learning_rate": 2.1739130434782607e-07,
+      "loss": 0.0006,
+      "reward": 1.640625,
+      "reward_mean": 1.640625,
+      "reward_std": 0.04419417306780815,
+      "rewards/accuracy_reward": 0.640625,
+      "rewards/format_reward": 1.0,
+      "step": 126
+    },
+    {
+      "advantages": 7.450580596923828e-09,
+      "completion_length": 83.546875,
+      "epoch": 0.7888198757763976,
+      "grad_norm": 9.345684051513672,
+      "kl": 0.008056640625,
+      "learning_rate": 2.1118012422360247e-07,
+      "loss": 0.0008,
+      "reward": 1.296875,
+      "reward_mean": 1.296875,
+      "reward_std": 0.19044628739356995,
+      "rewards/accuracy_reward": 0.296875,
+      "rewards/format_reward": 1.0,
+      "step": 127
+    },
+    {
+      "advantages": 0.0,
+      "completion_length": 85.84375,
+      "epoch": 0.7950310559006211,
+      "grad_norm": 0.22835175693035126,
+      "kl": 0.0084228515625,
+      "learning_rate": 2.0496894409937888e-07,
+      "loss": 0.0008,
+      "reward": 1.75,
+      "reward_mean": 1.75,
+      "reward_std": 0.0,
+      "rewards/accuracy_reward": 0.75,
+      "rewards/format_reward": 1.0,
+      "step": 128
+    },
+    {
+      "advantages": 0.0,
+      "completion_length": 81.828125,
+      "epoch": 0.8012422360248447,
+      "grad_norm": 2.44989275932312,
+      "kl": 0.007171630859375,
+      "learning_rate": 1.9875776397515526e-07,
+      "loss": 0.0007,
+      "reward": 1.5,
+      "reward_mean": 1.5,
+      "reward_std": 0.0883883461356163,
+      "rewards/accuracy_reward": 0.5,
+      "rewards/format_reward": 1.0,
+      "step": 129
+    },
+    {
+      "advantages": 9.313225746154785e-10,
+      "completion_length": 83.296875,
+      "epoch": 0.8074534161490683,
+      "grad_norm": 26.60379409790039,
+      "kl": 0.01031494140625,
+      "learning_rate": 1.9254658385093166e-07,
+      "loss": 0.001,
+      "reward": 1.640625,
+      "reward_mean": 1.640625,
+      "reward_std": 0.1530819982290268,
+      "rewards/accuracy_reward": 0.640625,
+      "rewards/format_reward": 1.0,
+      "step": 130
+    },
+    {
+      "advantages": -3.725290298461914e-09,
+      "completion_length": 75.296875,
+      "epoch": 0.8136645962732919,
+      "grad_norm": 2.649775981903076,
+      "kl": 0.00787353515625,
+      "learning_rate": 1.8633540372670807e-07,
+      "loss": 0.0008,
+      "reward": 1.71875,
+      "reward_mean": 1.71875,
+      "reward_std": 0.0883883461356163,
+      "rewards/accuracy_reward": 0.71875,
+      "rewards/format_reward": 1.0,
+      "step": 131
+    },
+    {
+      "advantages": -7.450580596923828e-09,
+      "completion_length": 72.921875,
+      "epoch": 0.8198757763975155,
+      "grad_norm": 6.021523952484131,
+      "kl": 0.017333984375,
+      "learning_rate": 1.8012422360248447e-07,
+      "loss": 0.0017,
+      "reward": 1.546875,
+      "reward_mean": 1.546875,
+      "reward_std": 0.17358146607875824,
+      "rewards/accuracy_reward": 0.546875,
+      "rewards/format_reward": 1.0,
+      "step": 132
+    },
+    {
+      "advantages": 1.862645149230957e-09,
+      "completion_length": 80.203125,
+      "epoch": 0.8260869565217391,
+      "grad_norm": 5.850553035736084,
+      "kl": 0.0118408203125,
+      "learning_rate": 1.7391304347826085e-07,
+      "loss": 0.0012,
+      "reward": 1.671875,
+      "reward_mean": 1.671875,
+      "reward_std": 0.25726157426834106,
+      "rewards/accuracy_reward": 0.671875,
+      "rewards/format_reward": 1.0,
+      "step": 133
+    },
+    {
+      "advantages": -5.587935447692871e-09,
+      "completion_length": 90.234375,
+      "epoch": 0.8322981366459627,
+      "grad_norm": 9.700899124145508,
+      "kl": 0.01422119140625,
+      "learning_rate": 1.6770186335403728e-07,
+      "loss": 0.0014,
+      "reward": 1.78125,
+      "reward_mean": 1.78125,
+      "reward_std": 0.2651650309562683,
+      "rewards/accuracy_reward": 0.828125,
+      "rewards/format_reward": 0.953125,
+      "step": 134
+    },
+    {
+      "advantages": 3.725290298461914e-09,
+      "completion_length": 81.90625,
+      "epoch": 0.8385093167701864,
+      "grad_norm": 2.9975473880767822,
+      "kl": 0.00909423828125,
+      "learning_rate": 1.6149068322981366e-07,
+      "loss": 0.0009,
+      "reward": 1.6875,
+      "reward_mean": 1.6875,
+      "reward_std": 0.1552036553621292,
+      "rewards/accuracy_reward": 0.6875,
+      "rewards/format_reward": 1.0,
+      "step": 135
+    },
+    {
+      "advantages": -3.725290298461914e-09,
+      "completion_length": 79.3125,
+      "epoch": 0.84472049689441,
+      "grad_norm": 4.324582099914551,
+      "kl": 0.01165771484375,
+      "learning_rate": 1.5527950310559004e-07,
+      "loss": 0.0012,
+      "reward": 1.84375,
+      "reward_mean": 1.84375,
+      "reward_std": 0.2177756428718567,
+      "rewards/accuracy_reward": 0.859375,
+      "rewards/format_reward": 0.984375,
+      "step": 136
+    },
+    {
+      "advantages": 7.450580596923828e-09,
+      "completion_length": 80.265625,
+      "epoch": 0.8509316770186336,
+      "grad_norm": 3.8911736011505127,
+      "kl": 0.009521484375,
+      "learning_rate": 1.4906832298136647e-07,
+      "loss": 0.001,
+      "reward": 1.5625,
+      "reward_mean": 1.5625,
+      "reward_std": 0.1552036553621292,
+      "rewards/accuracy_reward": 0.5625,
+      "rewards/format_reward": 1.0,
+      "step": 137
+    },
+    {
+      "advantages": 0.0,
+      "completion_length": 78.40625,
+      "epoch": 0.8571428571428571,
+      "grad_norm": 2.864941120147705,
+      "kl": 0.01129150390625,
+      "learning_rate": 1.4285714285714285e-07,
+      "loss": 0.0011,
+      "reward": 1.765625,
+      "reward_mean": 1.765625,
+      "reward_std": 0.10205793380737305,
+      "rewards/accuracy_reward": 0.765625,
+      "rewards/format_reward": 1.0,
+      "step": 138
+    },
+    {
+      "advantages": 3.725290298461914e-09,
+      "completion_length": 80.21875,
+      "epoch": 0.8633540372670807,
+      "grad_norm": 5.788990497589111,
+      "kl": 0.00799560546875,
+      "learning_rate": 1.3664596273291925e-07,
+      "loss": 0.0008,
+      "reward": 1.5625,
+      "reward_mean": 1.5625,
+      "reward_std": 0.1552036553621292,
+      "rewards/accuracy_reward": 0.5625,
+      "rewards/format_reward": 1.0,
+      "step": 139
+    },
+    {
+      "advantages": -1.862645149230957e-09,
+      "completion_length": 92.90625,
+      "epoch": 0.8695652173913043,
+      "grad_norm": 4.130926609039307,
+      "kl": 0.0111083984375,
+      "learning_rate": 1.3043478260869563e-07,
+      "loss": 0.0011,
+      "reward": 1.859375,
+      "reward_mean": 1.859375,
+      "reward_std": 0.2198973000049591,
+      "rewards/accuracy_reward": 0.859375,
+      "rewards/format_reward": 1.0,
+      "step": 140
+    },
+    {
+      "advantages": -3.725290298461914e-09,
+      "completion_length": 79.109375,
+      "epoch": 0.8757763975155279,
+      "grad_norm": 3.025212287902832,
+      "kl": 0.01324462890625,
+      "learning_rate": 1.2422360248447204e-07,
+      "loss": 0.0013,
+      "reward": 1.46875,
+      "reward_mean": 1.46875,
+      "reward_std": 0.1767766922712326,
+      "rewards/accuracy_reward": 0.484375,
+      "rewards/format_reward": 0.984375,
+      "step": 141
+    },
+    {
+      "advantages": 3.725290298461914e-09,
+      "completion_length": 78.25,
+      "epoch": 0.8819875776397516,
+      "grad_norm": 6.828762531280518,
+      "kl": 0.01177978515625,
+      "learning_rate": 1.1801242236024844e-07,
+      "loss": 0.0012,
+      "reward": 1.40625,
+      "reward_mean": 1.40625,
+      "reward_std": 0.1462520956993103,
+      "rewards/accuracy_reward": 0.40625,
+      "rewards/format_reward": 1.0,
+      "step": 142
+    },
+    {
+      "advantages": 0.0,
+      "completion_length": 73.625,
+      "epoch": 0.8881987577639752,
+      "grad_norm": 3.4486515522003174,
+      "kl": 0.00762939453125,
+      "learning_rate": 1.1180124223602484e-07,
+      "loss": 0.0008,
+      "reward": 1.65625,
+      "reward_mean": 1.65625,
+      "reward_std": 0.1552036553621292,
+      "rewards/accuracy_reward": 0.65625,
+      "rewards/format_reward": 1.0,
+      "step": 143
+    },
+    {
+      "advantages": -3.725290298461914e-09,
+      "completion_length": 80.359375,
+      "epoch": 0.8944099378881988,
+      "grad_norm": 8.272978782653809,
+      "kl": 0.00628662109375,
+      "learning_rate": 1.0559006211180124e-07,
+      "loss": 0.0006,
+      "reward": 1.71875,
+      "reward_mean": 1.71875,
+      "reward_std": 0.0883883461356163,
+      "rewards/accuracy_reward": 0.734375,
+      "rewards/format_reward": 0.984375,
+      "step": 144
+    },
+    {
+      "advantages": -7.450580596923828e-09,
+      "completion_length": 81.1875,
+      "epoch": 0.9006211180124224,
+      "grad_norm": 4.848587512969971,
+      "kl": 0.017578125,
+      "learning_rate": 9.937888198757763e-08,
+      "loss": 0.0018,
+      "reward": 1.5625,
+      "reward_mean": 1.5625,
+      "reward_std": 0.2041158676147461,
+      "rewards/accuracy_reward": 0.5625,
+      "rewards/format_reward": 1.0,
+      "step": 145
+    },
+    {
+      "advantages": 0.0,
+      "completion_length": 79.984375,
+      "epoch": 0.906832298136646,
+      "grad_norm": 0.3604845702648163,
+      "kl": 0.00958251953125,
+      "learning_rate": 9.316770186335403e-08,
+      "loss": 0.001,
+      "reward": 1.75,
+      "reward_mean": 1.75,
+      "reward_std": 0.0,
+      "rewards/accuracy_reward": 0.75,
+      "rewards/format_reward": 1.0,
+      "step": 146
+    },
+    {
+      "advantages": 5.587935447692871e-09,
+      "completion_length": 76.5625,
+      "epoch": 0.9130434782608695,
+      "grad_norm": 10.680438995361328,
+      "kl": 0.01116943359375,
+      "learning_rate": 8.695652173913042e-08,
+      "loss": 0.0011,
+      "reward": 1.671875,
+      "reward_mean": 1.671875,
+      "reward_std": 0.19939783215522766,
+      "rewards/accuracy_reward": 0.671875,
+      "rewards/format_reward": 1.0,
+      "step": 147
+    },
+    {
+      "advantages": -3.725290298461914e-09,
+      "completion_length": 76.984375,
+      "epoch": 0.9192546583850931,
+      "grad_norm": 2.091907024383545,
+      "kl": 0.010498046875,
+      "learning_rate": 8.074534161490683e-08,
+      "loss": 0.0011,
+      "reward": 1.640625,
+      "reward_mean": 1.640625,
+      "reward_std": 0.08010874688625336,
+      "rewards/accuracy_reward": 0.65625,
+      "rewards/format_reward": 0.984375,
+      "step": 148
+    },
+    {
+      "advantages": 0.0,
+      "completion_length": 76.203125,
+      "epoch": 0.9254658385093167,
+      "grad_norm": 0.20045147836208344,
+      "kl": 0.0078125,
+      "learning_rate": 7.453416149068323e-08,
+      "loss": 0.0008,
+      "reward": 1.75,
+      "reward_mean": 1.75,
+      "reward_std": 0.0,
+      "rewards/accuracy_reward": 0.75,
+      "rewards/format_reward": 1.0,
+      "step": 149
+    },
+    {
+      "advantages": 0.0,
+      "completion_length": 84.140625,
+      "epoch": 0.9316770186335404,
+      "grad_norm": 3.21720814704895,
+      "kl": 0.008544921875,
+      "learning_rate": 6.832298136645963e-08,
+      "loss": 0.0009,
+      "reward": 1.6875,
+      "reward_mean": 1.6875,
+      "reward_std": 0.06681530922651291,
+      "rewards/accuracy_reward": 0.6875,
+      "rewards/format_reward": 1.0,
+      "step": 150
+    },
+    {
+      "advantages": -1.862645149230957e-09,
+      "completion_length": 82.734375,
+      "epoch": 0.937888198757764,
+      "grad_norm": 7.955801963806152,
+      "kl": 0.01263427734375,
+      "learning_rate": 6.211180124223602e-08,
+      "loss": 0.0013,
+      "reward": 1.734375,
+      "reward_mean": 1.734375,
+      "reward_std": 0.10205793380737305,
+      "rewards/accuracy_reward": 0.734375,
+      "rewards/format_reward": 1.0,
+      "step": 151
+    },
+    {
+      "advantages": 3.725290298461914e-09,
+      "completion_length": 72.96875,
+      "epoch": 0.9440993788819876,
+      "grad_norm": 3.563530921936035,
+      "kl": 0.0093994140625,
+      "learning_rate": 5.590062111801242e-08,
+      "loss": 0.0009,
+      "reward": 1.90625,
+      "reward_mean": 1.90625,
+      "reward_std": 0.0578637570142746,
+      "rewards/accuracy_reward": 0.90625,
+      "rewards/format_reward": 1.0,
+      "step": 152
+    },
+    {
+      "advantages": -9.313225746154785e-09,
+      "completion_length": 83.125,
+      "epoch": 0.9503105590062112,
+      "grad_norm": 9.811988830566406,
+      "kl": 0.0184326171875,
+      "learning_rate": 4.9689440993788814e-08,
+      "loss": 0.0019,
+      "reward": 1.796875,
+      "reward_mean": 1.796875,
+      "reward_std": 0.15992169082164764,
+      "rewards/accuracy_reward": 0.796875,
+      "rewards/format_reward": 1.0,
+      "step": 153
+    },
+    {
+      "advantages": -1.0244548320770264e-08,
+      "completion_length": 83.515625,
+      "epoch": 0.9565217391304348,
+      "grad_norm": 3.8269639015197754,
+      "kl": 0.0133056640625,
+      "learning_rate": 4.347826086956521e-08,
+      "loss": 0.0013,
+      "reward": 1.78125,
+      "reward_mean": 1.78125,
+      "reward_std": 0.16675157845020294,
+      "rewards/accuracy_reward": 0.78125,
+      "rewards/format_reward": 1.0,
+      "step": 154
+    },
+    {
+      "advantages": -7.450580596923828e-09,
+      "completion_length": 85.671875,
+      "epoch": 0.9627329192546584,
+      "grad_norm": 3.470165252685547,
+      "kl": 0.01300048828125,
+      "learning_rate": 3.726708074534162e-08,
+      "loss": 0.0013,
+      "reward": 1.5625,
+      "reward_mean": 1.5625,
+      "reward_std": 0.1462520956993103,
+      "rewards/accuracy_reward": 0.5625,
+      "rewards/format_reward": 1.0,
+      "step": 155
+    },
+    {
+      "advantages": -9.313225746154785e-10,
+      "completion_length": 84.203125,
+      "epoch": 0.968944099378882,
+      "grad_norm": 2.550407648086548,
+      "kl": 0.00946044921875,
+      "learning_rate": 3.105590062111801e-08,
+      "loss": 0.0009,
+      "reward": 1.625,
+      "reward_mean": 1.625,
+      "reward_std": 0.16675157845020294,
+      "rewards/accuracy_reward": 0.640625,
+      "rewards/format_reward": 0.984375,
+      "step": 156
+    },
+    {
+      "advantages": 3.725290298461914e-09,
+      "completion_length": 76.296875,
+      "epoch": 0.9751552795031055,
+      "grad_norm": 3.396425247192383,
+      "kl": 0.00714111328125,
+      "learning_rate": 2.4844720496894407e-08,
+      "loss": 0.0007,
+      "reward": 1.546875,
+      "reward_mean": 1.546875,
+      "reward_std": 0.15992169082164764,
+      "rewards/accuracy_reward": 0.546875,
+      "rewards/format_reward": 1.0,
+      "step": 157
+    },
+    {
+      "advantages": -2.7939677238464355e-09,
+      "completion_length": 73.859375,
+      "epoch": 0.9813664596273292,
+      "grad_norm": 3.776041030883789,
+      "kl": 0.00921630859375,
+      "learning_rate": 1.863354037267081e-08,
+      "loss": 0.0009,
+      "reward": 1.59375,
+      "reward_mean": 1.59375,
+      "reward_std": 0.10888782143592834,
+      "rewards/accuracy_reward": 0.59375,
+      "rewards/format_reward": 1.0,
+      "step": 158
+    },
+    {
+      "advantages": 1.862645149230957e-09,
+      "completion_length": 77.953125,
+      "epoch": 0.9875776397515528,
+      "grad_norm": 3.304471254348755,
+      "kl": 0.01263427734375,
+      "learning_rate": 1.2422360248447204e-08,
+      "loss": 0.0013,
+      "reward": 1.796875,
+      "reward_mean": 1.796875,
+      "reward_std": 0.11100947856903076,
+      "rewards/accuracy_reward": 0.796875,
+      "rewards/format_reward": 1.0,
+      "step": 159
+    },
+    {
+      "advantages": 0.0,
+      "completion_length": 79.25,
+      "epoch": 0.9937888198757764,
+      "grad_norm": 5.823967456817627,
+      "kl": 0.00897216796875,
+      "learning_rate": 6.211180124223602e-09,
+      "loss": 0.0009,
+      "reward": 1.5,
+      "reward_mean": 1.5,
+      "reward_std": 0.0883883461356163,
+      "rewards/accuracy_reward": 0.5,
+      "rewards/format_reward": 1.0,
+      "step": 160
+    },
+    {
+      "advantages": -0.5890890955924988,
+      "completion_length": 89.33333587646484,
+      "epoch": 1.0,
+      "grad_norm": 2.0931286811828613,
+      "kl": 0.00677490234375,
+      "learning_rate": 0.0,
+      "loss": 0.001,
+      "reward": 1.6666667461395264,
+      "reward_mean": 1.875,
+      "reward_std": 0.3535533845424652,
+      "rewards/accuracy_reward": 0.6666666865348816,
+      "rewards/format_reward": 1.0,
+      "step": 161
+    }
+  ],
+  "logging_steps": 1.0,
+  "max_steps": 161,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 1,
+  "save_steps": 10,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": true
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 0.0,
+  "train_batch_size": 1,
+  "trial_name": null,
+  "trial_params": null
+}

training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3ff22b61060846c443ccf16d8054c34f49711075b591a288d7706ce0f75243ee
+size 8056

zero_to_fp32.py

@@ -0,0 +1,674 @@
+#!/usr/bin/env python
+
+# Copyright (c) Microsoft Corporation.
+# SPDX-License-Identifier: Apache-2.0
+
+# DeepSpeed Team
+
+# This script extracts fp32 consolidated weights from a zero 1, 2 and 3 DeepSpeed checkpoints. It gets
+# copied into the top level checkpoint dir, so the user can easily do the conversion at any point in
+# the future. Once extracted, the weights don't require DeepSpeed and can be used in any
+# application.
+#
+# example:
+#   python zero_to_fp32.py . output_dir/
+#   or
+#   python zero_to_fp32.py . output_dir/ --safe_serialization
+
+import argparse
+import torch
+import glob
+import math
+import os
+import re
+import json
+from tqdm import tqdm
+from collections import OrderedDict
+from dataclasses import dataclass
+
+# while this script doesn't use deepspeed to recover data, since the checkpoints are pickled with
+# DeepSpeed data structures it has to be available in the current python environment.
+from deepspeed.utils import logger
+from deepspeed.checkpoint.constants import (DS_VERSION, OPTIMIZER_STATE_DICT, SINGLE_PARTITION_OF_FP32_GROUPS,
+                                            FP32_FLAT_GROUPS, ZERO_STAGE, PARTITION_COUNT, PARAM_SHAPES, BUFFER_NAMES,
+                                            FROZEN_PARAM_SHAPES, FROZEN_PARAM_FRAGMENTS)
+
+
+@dataclass
+class zero_model_state:
+    buffers: dict()
+    param_shapes: dict()
+    shared_params: list
+    ds_version: int
+    frozen_param_shapes: dict()
+    frozen_param_fragments: dict()
+
+
+debug = 0
+
+# load to cpu
+device = torch.device('cpu')
+
+
+def atoi(text):
+    return int(text) if text.isdigit() else text
+
+
+def natural_keys(text):
+    '''
+    alist.sort(key=natural_keys) sorts in human order
+    http://nedbatchelder.com/blog/200712/human_sorting.html
+    (See Toothy's implementation in the comments)
+    '''
+    return [atoi(c) for c in re.split(r'(\d+)', text)]
+
+
+def get_model_state_file(checkpoint_dir, zero_stage):
+    if not os.path.isdir(checkpoint_dir):
+        raise FileNotFoundError(f"Directory '{checkpoint_dir}' doesn't exist")
+
+    # there should be only one file
+    if zero_stage <= 2:
+        file = os.path.join(checkpoint_dir, "mp_rank_00_model_states.pt")
+    elif zero_stage == 3:
+        file = os.path.join(checkpoint_dir, "zero_pp_rank_0_mp_rank_00_model_states.pt")
+
+    if not os.path.exists(file):
+        raise FileNotFoundError(f"can't find model states file at '{file}'")
+
+    return file
+
+
+def get_checkpoint_files(checkpoint_dir, glob_pattern):
+    # XXX: need to test that this simple glob rule works for multi-node setup too
+    ckpt_files = sorted(glob.glob(os.path.join(checkpoint_dir, glob_pattern)), key=natural_keys)
+
+    if len(ckpt_files) == 0:
+        raise FileNotFoundError(f"can't find {glob_pattern} files in directory '{checkpoint_dir}'")
+
+    return ckpt_files
+
+
+def get_optim_files(checkpoint_dir):
+    return get_checkpoint_files(checkpoint_dir, "*_optim_states.pt")
+
+
+def get_model_state_files(checkpoint_dir):
+    return get_checkpoint_files(checkpoint_dir, "*_model_states.pt")
+
+
+def parse_model_states(files):
+    zero_model_states = []
+    for file in files:
+        state_dict = torch.load(file, map_location=device)
+
+        if BUFFER_NAMES not in state_dict:
+            raise ValueError(f"{file} is not a model state checkpoint")
+        buffer_names = state_dict[BUFFER_NAMES]
+        if debug:
+            print("Found buffers:", buffer_names)
+
+        # recover just the buffers while restoring them to fp32 if they were saved in fp16
+        buffers = {k: v.float() for k, v in state_dict["module"].items() if k in buffer_names}
+        param_shapes = state_dict[PARAM_SHAPES]
+
+        # collect parameters that are included in param_shapes
+        param_names = []
+        for s in param_shapes:
+            for name in s.keys():
+                param_names.append(name)
+
+        # update with frozen parameters
+        frozen_param_shapes = state_dict.get(FROZEN_PARAM_SHAPES, None)
+        if frozen_param_shapes is not None:
+            if debug:
+                print(f"Found frozen_param_shapes: {frozen_param_shapes}")
+            param_names += list(frozen_param_shapes.keys())
+
+        # handle shared params
+        shared_params = [[k, v] for k, v in state_dict["shared_params"].items()]
+
+        ds_version = state_dict.get(DS_VERSION, None)
+
+        frozen_param_fragments = state_dict.get(FROZEN_PARAM_FRAGMENTS, None)
+
+        z_model_state = zero_model_state(buffers=buffers,
+                                         param_shapes=param_shapes,
+                                         shared_params=shared_params,
+                                         ds_version=ds_version,
+                                         frozen_param_shapes=frozen_param_shapes,
+                                         frozen_param_fragments=frozen_param_fragments)
+        zero_model_states.append(z_model_state)
+
+    return zero_model_states
+
+
+def parse_optim_states(files, ds_checkpoint_dir):
+    total_files = len(files)
+    state_dicts = []
+    for f in files:
+        state_dict = torch.load(f, map_location=device)
+        # immediately discard the potentially huge 2 optimizer states as we only care for fp32 master weights
+        # and also handle the case where it was already removed by another helper script
+        state_dict["optimizer_state_dict"].pop("optimizer_state_dict", None)
+        state_dicts.append(state_dict)
+
+    if not ZERO_STAGE in state_dicts[0][OPTIMIZER_STATE_DICT]:
+        raise ValueError(f"{files[0]} is not a zero checkpoint")
+    zero_stage = state_dicts[0][OPTIMIZER_STATE_DICT][ZERO_STAGE]
+    world_size = state_dicts[0][OPTIMIZER_STATE_DICT][PARTITION_COUNT]
+
+    # For ZeRO-2 each param group can have different partition_count as data parallelism for expert
+    # parameters can be different from data parallelism for non-expert parameters. So we can just
+    # use the max of the partition_count to get the dp world_size.
+
+    if type(world_size) is list:
+        world_size = max(world_size)
+
+    if world_size != total_files:
+        raise ValueError(
+            f"Expected {world_size} of '*_optim_states.pt' under '{ds_checkpoint_dir}' but found {total_files} files. "
+            "Possibly due to an overwrite of an old checkpoint, or a checkpoint didn't get saved by one or more processes."
+        )
+
+    # the groups are named differently in each stage
+    if zero_stage <= 2:
+        fp32_groups_key = SINGLE_PARTITION_OF_FP32_GROUPS
+    elif zero_stage == 3:
+        fp32_groups_key = FP32_FLAT_GROUPS
+    else:
+        raise ValueError(f"unknown zero stage {zero_stage}")
+
+    if zero_stage <= 2:
+        fp32_flat_groups = [state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key] for i in range(len(state_dicts))]
+    elif zero_stage == 3:
+        # if there is more than one param group, there will be multiple flattened tensors - one
+        # flattened tensor per group - for simplicity merge them into a single tensor
+        #
+        # XXX: could make the script more memory efficient for when there are multiple groups - it
+        # will require matching the sub-lists of param_shapes for each param group flattened tensor
+
+        fp32_flat_groups = [
+            torch.cat(state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key], 0) for i in range(len(state_dicts))
+        ]
+
+    return zero_stage, world_size, fp32_flat_groups
+
+
+def _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters):
+    """
+    Returns fp32 state_dict reconstructed from ds checkpoint
+
+    Args:
+        - ``ds_checkpoint_dir``: path to the deepspeed checkpoint folder (where the optimizer files are)
+
+    """
+    print(f"Processing zero checkpoint '{ds_checkpoint_dir}'")
+
+    optim_files = get_optim_files(ds_checkpoint_dir)
+    zero_stage, world_size, fp32_flat_groups = parse_optim_states(optim_files, ds_checkpoint_dir)
+    print(f"Detected checkpoint of type zero stage {zero_stage}, world_size: {world_size}")
+
+    model_files = get_model_state_files(ds_checkpoint_dir)
+
+    zero_model_states = parse_model_states(model_files)
+    print(f'Parsing checkpoint created by deepspeed=={zero_model_states[0].ds_version}')
+
+    if zero_stage <= 2:
+        return _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                                          exclude_frozen_parameters)
+    elif zero_stage == 3:
+        return _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                                          exclude_frozen_parameters)
+
+
+def _zero2_merge_frozen_params(state_dict, zero_model_states):
+    if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
+        return
+
+    frozen_param_shapes = zero_model_states[0].frozen_param_shapes
+    frozen_param_fragments = zero_model_states[0].frozen_param_fragments
+
+    if debug:
+        num_elem = sum(s.numel() for s in frozen_param_shapes.values())
+        print(f'rank 0: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
+
+        wanted_params = len(frozen_param_shapes)
+        wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
+        avail_numel = sum([p.numel() for p in frozen_param_fragments.values()])
+        print(f'Frozen params: Have {avail_numel} numels to process.')
+        print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
+
+    total_params = 0
+    total_numel = 0
+    for name, shape in frozen_param_shapes.items():
+        total_params += 1
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+
+        state_dict[name] = frozen_param_fragments[name]
+
+        if debug:
+            print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
+
+    print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _has_callable(obj, fn):
+    attr = getattr(obj, fn, None)
+    return callable(attr)
+
+
+def _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
+    param_shapes = zero_model_states[0].param_shapes
+
+    # Reconstruction protocol:
+    #
+    # XXX: document this
+
+    if debug:
+        for i in range(world_size):
+            for j in range(len(fp32_flat_groups[0])):
+                print(f"{FP32_FLAT_GROUPS}[{i}][{j}].shape={fp32_flat_groups[i][j].shape}")
+
+    # XXX: memory usage doubles here (zero2)
+    num_param_groups = len(fp32_flat_groups[0])
+    merged_single_partition_of_fp32_groups = []
+    for i in range(num_param_groups):
+        merged_partitions = [sd[i] for sd in fp32_flat_groups]
+        full_single_fp32_vector = torch.cat(merged_partitions, 0)
+        merged_single_partition_of_fp32_groups.append(full_single_fp32_vector)
+    avail_numel = sum(
+        [full_single_fp32_vector.numel() for full_single_fp32_vector in merged_single_partition_of_fp32_groups])
+
+    if debug:
+        wanted_params = sum([len(shapes) for shapes in param_shapes])
+        wanted_numel = sum([sum(shape.numel() for shape in shapes.values()) for shapes in param_shapes])
+        # not asserting if there is a mismatch due to possible padding
+        print(f"Have {avail_numel} numels to process.")
+        print(f"Need {wanted_numel} numels in {wanted_params} params.")
+
+    # params
+    # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
+    # out-of-core computing solution
+    total_numel = 0
+    total_params = 0
+    for shapes, full_single_fp32_vector in zip(param_shapes, merged_single_partition_of_fp32_groups):
+        offset = 0
+        avail_numel = full_single_fp32_vector.numel()
+        for name, shape in shapes.items():
+
+            unpartitioned_numel = shape.numel() if _has_callable(shape, 'numel') else math.prod(shape)
+            total_numel += unpartitioned_numel
+            total_params += 1
+
+            if debug:
+                print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
+            state_dict[name] = full_single_fp32_vector.narrow(0, offset, unpartitioned_numel).view(shape)
+            offset += unpartitioned_numel
+
+        # Z2 started to align to 2*world_size to improve nccl performance. Therefore both offset and
+        # avail_numel can differ by anywhere between 0..2*world_size. Due to two unrelated complex
+        # paddings performed in the code it's almost impossible to predict the exact numbers w/o the
+        # live optimizer object, so we are checking that the numbers are within the right range
+        align_to = 2 * world_size
+
+        def zero2_align(x):
+            return align_to * math.ceil(x / align_to)
+
+        if debug:
+            print(f"original offset={offset}, avail_numel={avail_numel}")
+
+        offset = zero2_align(offset)
+        avail_numel = zero2_align(avail_numel)
+
+        if debug:
+            print(f"aligned  offset={offset}, avail_numel={avail_numel}")
+
+        # Sanity check
+        if offset != avail_numel:
+            raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
+
+    print(f"Reconstructed fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                               exclude_frozen_parameters):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    if not exclude_frozen_parameters:
+        _zero2_merge_frozen_params(state_dict, zero_model_states)
+
+    _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
+
+    # recover shared parameters
+    for pair in zero_model_states[0].shared_params:
+        if pair[1] in state_dict:
+            state_dict[pair[0]] = state_dict[pair[1]]
+
+    return state_dict
+
+
+def zero3_partitioned_param_info(unpartitioned_numel, world_size):
+    remainder = unpartitioned_numel % world_size
+    padding_numel = (world_size - remainder) if remainder else 0
+    partitioned_numel = math.ceil(unpartitioned_numel / world_size)
+    return partitioned_numel, padding_numel
+
+
+def _zero3_merge_frozen_params(state_dict, world_size, zero_model_states):
+    if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
+        return
+
+    if debug:
+        for i in range(world_size):
+            num_elem = sum(s.numel() for s in zero_model_states[i].frozen_param_fragments.values())
+            print(f'rank {i}: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
+
+        frozen_param_shapes = zero_model_states[0].frozen_param_shapes
+        wanted_params = len(frozen_param_shapes)
+        wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
+        avail_numel = sum([p.numel() for p in zero_model_states[0].frozen_param_fragments.values()]) * world_size
+        print(f'Frozen params: Have {avail_numel} numels to process.')
+        print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
+
+    total_params = 0
+    total_numel = 0
+    for name, shape in zero_model_states[0].frozen_param_shapes.items():
+        total_params += 1
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+
+        param_frags = tuple(model_state.frozen_param_fragments[name] for model_state in zero_model_states)
+        state_dict[name] = torch.cat(param_frags, 0).narrow(0, 0, unpartitioned_numel).view(shape)
+
+        partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
+
+        if debug:
+            print(
+                f"Frozen params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
+            )
+
+    print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
+    param_shapes = zero_model_states[0].param_shapes
+    avail_numel = fp32_flat_groups[0].numel() * world_size
+    # Reconstruction protocol: For zero3 we need to zip the partitions together at boundary of each
+    # param, re-consolidating each param, while dealing with padding if any
+
+    # merge list of dicts, preserving order
+    param_shapes = {k: v for d in param_shapes for k, v in d.items()}
+
+    if debug:
+        for i in range(world_size):
+            print(f"{FP32_FLAT_GROUPS}[{i}].shape={fp32_flat_groups[i].shape}")
+
+        wanted_params = len(param_shapes)
+        wanted_numel = sum(shape.numel() for shape in param_shapes.values())
+        # not asserting if there is a mismatch due to possible padding
+        avail_numel = fp32_flat_groups[0].numel() * world_size
+        print(f"Trainable params: Have {avail_numel} numels to process.")
+        print(f"Trainable params: Need {wanted_numel} numels in {wanted_params} params.")
+
+    # params
+    # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
+    # out-of-core computing solution
+    offset = 0
+    total_numel = 0
+    total_params = 0
+    for name, shape in tqdm(param_shapes.items(), desc='Gathering Sharded Weights'):
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+        total_params += 1
+        partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
+
+        if debug:
+            print(
+                f"Trainable params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
+            )
+
+        # XXX: memory usage doubles here
+        state_dict[name] = torch.cat(
+            tuple(fp32_flat_groups[i].narrow(0, offset, partitioned_numel) for i in range(world_size)),
+            0).narrow(0, 0, unpartitioned_numel).view(shape)
+        offset += partitioned_numel
+
+    offset *= world_size
+
+    # Sanity check
+    if offset != avail_numel:
+        raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
+
+    print(f"Reconstructed Trainable fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                               exclude_frozen_parameters):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    if not exclude_frozen_parameters:
+        _zero3_merge_frozen_params(state_dict, world_size, zero_model_states)
+
+    _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
+
+    # recover shared parameters
+    for pair in zero_model_states[0].shared_params:
+        if pair[1] in state_dict:
+            state_dict[pair[0]] = state_dict[pair[1]]
+
+    return state_dict
+
+
+def get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag=None, exclude_frozen_parameters=False):
+    """
+    Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated state_dict that can be loaded with
+    ``load_state_dict()`` and used for training without DeepSpeed or shared with others, for example
+    via a model hub.
+
+    Args:
+        - ``checkpoint_dir``: path to the desired checkpoint folder
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in 'latest' file. e.g., ``global_step14``
+        - ``exclude_frozen_parameters``: exclude frozen parameters
+
+    Returns:
+        - pytorch ``state_dict``
+
+    Note: this approach may not work if your application doesn't have sufficient free CPU memory and
+    you may need to use the offline approach using the ``zero_to_fp32.py`` script that is saved with
+    the checkpoint.
+
+    A typical usage might be ::
+
+        from deepspeed.utils.zero_to_fp32 import get_fp32_state_dict_from_zero_checkpoint
+        # do the training and checkpoint saving
+        state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir) # already on cpu
+        model = model.cpu() # move to cpu
+        model.load_state_dict(state_dict)
+        # submit to model hub or save the model to share with others
+
+    In this example the ``model`` will no longer be usable in the deepspeed context of the same
+    application. i.e. you will need to re-initialize the deepspeed engine, since
+    ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
+
+    If you want it all done for you, use ``load_state_dict_from_zero_checkpoint`` instead.
+
+    """
+    if tag is None:
+        latest_path = os.path.join(checkpoint_dir, 'latest')
+        if os.path.isfile(latest_path):
+            with open(latest_path, 'r') as fd:
+                tag = fd.read().strip()
+        else:
+            raise ValueError(f"Unable to find 'latest' file at {latest_path}")
+
+    ds_checkpoint_dir = os.path.join(checkpoint_dir, tag)
+
+    if not os.path.isdir(ds_checkpoint_dir):
+        raise FileNotFoundError(f"Directory '{ds_checkpoint_dir}' doesn't exist")
+
+    return _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters)
+
+
+def convert_zero_checkpoint_to_fp32_state_dict(checkpoint_dir,
+                                               output_dir,
+                                               max_shard_size="5GB",
+                                               safe_serialization=False,
+                                               tag=None,
+                                               exclude_frozen_parameters=False):
+    """
+    Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict`` file that can be
+    loaded with ``torch.load(file)`` + ``load_state_dict()`` and used for training without DeepSpeed.
+
+    Args:
+        - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
+        - ``output_dir``: directory to the pytorch fp32 state_dict output files
+        - ``max_shard_size``: the maximum size for a checkpoint before being sharded, default value is 5GB
+        - ``safe_serialization``:  whether to save the model using `safetensors` or the traditional PyTorch way (that uses `pickle`).
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
+        - ``exclude_frozen_parameters``: exclude frozen parameters
+    """
+    # Dependency pre-check
+    if safe_serialization:
+        try:
+            from safetensors.torch import save_file
+        except ImportError:
+            print('If you want to use `safe_serialization`, please `pip install safetensors`')
+            raise
+    if max_shard_size is not None:
+        try:
+            from huggingface_hub import split_torch_state_dict_into_shards
+        except ImportError:
+            print('If you want to use `max_shard_size`, please `pip install huggingface_hub`')
+            raise
+
+    # Convert zero checkpoint to state_dict
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag, exclude_frozen_parameters)
+
+    # Shard the model if it is too big.
+    weights_name = "model.safetensors" if safe_serialization else "pytorch_model.bin"
+    if max_shard_size is not None:
+        filename_pattern = weights_name.replace(".bin", "{suffix}.bin").replace(".safetensors", "{suffix}.safetensors")
+        state_dict_split = split_torch_state_dict_into_shards(state_dict,
+                                                              filename_pattern=filename_pattern,
+                                                              max_shard_size=max_shard_size)
+    else:
+        from collections import namedtuple
+        StateDictSplit = namedtuple("StateDictSplit", ["is_sharded", "filename_to_tensors"])
+        state_dict_split = StateDictSplit(is_sharded=False,
+                                          filename_to_tensors={weights_name: list(state_dict.keys())})
+
+    # Save the model
+    filename_to_tensors = state_dict_split.filename_to_tensors.items()
+    for shard_file, tensors in tqdm(filename_to_tensors, desc="Saving checkpoint shards"):
+        shard = {tensor: state_dict[tensor].contiguous() for tensor in tensors}
+        output_path = os.path.join(output_dir, shard_file)
+        if safe_serialization:
+            save_file(shard, output_path, metadata={"format": "pt"})
+        else:
+            torch.save(shard, output_path)
+
+    # Save index if sharded
+    if state_dict_split.is_sharded:
+        index = {
+            "metadata": state_dict_split.metadata,
+            "weight_map": state_dict_split.tensor_to_filename,
+        }
+        save_index_file = "model.safetensors.index.json" if safe_serialization else "pytorch_model.bin.index.json"
+        save_index_file = os.path.join(output_dir, save_index_file)
+        with open(save_index_file, "w", encoding="utf-8") as f:
+            content = json.dumps(index, indent=2, sort_keys=True) + "\n"
+            f.write(content)
+
+
+def load_state_dict_from_zero_checkpoint(model, checkpoint_dir, tag=None):
+    """
+    1. Put the provided model to cpu
+    2. Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict``
+    3. Load it into the provided model
+
+    Args:
+        - ``model``: the model object to update
+        - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
+
+    Returns:
+        - ``model`: modified model
+
+    Make sure you have plenty of CPU memory available before you call this function. If you don't
+    have enough use the ``zero_to_fp32.py`` utility to do the conversion. You will find it
+    conveniently placed for you in the checkpoint folder.
+
+    A typical usage might be ::
+
+        from deepspeed.utils.zero_to_fp32 import load_state_dict_from_zero_checkpoint
+        model = load_state_dict_from_zero_checkpoint(trainer.model, checkpoint_dir)
+        # submit to model hub or save the model to share with others
+
+    Note, that once this was run, the ``model`` will no longer be usable in the deepspeed context
+    of the same application. i.e. you will need to re-initialize the deepspeed engine, since
+    ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
+
+    """
+    logger.info(f"Extracting fp32 weights")
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag)
+
+    logger.info(f"Overwriting model with fp32 weights")
+    model = model.cpu()
+    model.load_state_dict(state_dict, strict=False)
+
+    return model
+
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument("checkpoint_dir",
+                        type=str,
+                        help="path to the desired checkpoint folder, e.g., path/checkpoint-12")
+    parser.add_argument("output_dir",
+                        type=str,
+                        help="directory to the pytorch fp32 state_dict output files"
+                        "(e.g. path/checkpoint-12-output/)")
+    parser.add_argument(
+        "--max_shard_size",
+        type=str,
+        default="5GB",
+        help="The maximum size for a checkpoint before being sharded. Checkpoints shard will then be each of size"
+        "lower than this size. If expressed as a string, needs to be digits followed by a unit (like `5MB`"
+        "We default it to 5GB in order for models to be able to run easily on free-tier google colab instances"
+        "without CPU OOM issues.")
+    parser.add_argument(
+        "--safe_serialization",
+        default=False,
+        action='store_true',
+        help="Whether to save the model using `safetensors` or the traditional PyTorch way (that uses `pickle`).")
+    parser.add_argument("-t",
+                        "--tag",
+                        type=str,
+                        default=None,
+                        help="checkpoint tag used as a unique identifier for checkpoint. e.g., global_step1")
+    parser.add_argument("--exclude_frozen_parameters", action='store_true', help="exclude frozen parameters")
+    parser.add_argument("-d", "--debug", action='store_true', help="enable debug")
+    args = parser.parse_args()
+
+    debug = args.debug
+
+    convert_zero_checkpoint_to_fp32_state_dict(args.checkpoint_dir,
+                                               args.output_dir,
+                                               max_shard_size=args.max_shard_size,
+                                               safe_serialization=args.safe_serialization,
+                                               tag=args.tag,
+                                               exclude_frozen_parameters=args.exclude_frozen_parameters)