Upload folder using huggingface_hub

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 +1,103 @@
+{
+  "architectures": [
+    "AccustomThinker"
+  ],
+  "audio_config": {
+    "activation_dropout": 0.0,
+    "activation_function": "gelu",
+    "attention_dropout": 0.0,
+    "d_model": 1280,
+    "dropout": 0.0,
+    "encoder_attention_heads": 20,
+    "encoder_ffn_dim": 5120,
+    "encoder_layerdrop": 0.0,
+    "encoder_layers": 32,
+    "init_std": 0.02,
+    "initializer_range": 0.02,
+    "max_source_positions": 1500,
+    "model_type": "qwen2_5_omni_audio_encoder",
+    "n_window": 100,
+    "num_hidden_layers": 32,
+    "num_mel_bins": 128,
+    "output_dim": 2048,
+    "scale_embedding": false,
+    "torch_dtype": "float32"
+  },
+  "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": {
+    "attention_dropout": 0.0,
+    "hidden_act": "silu",
+    "hidden_size": 2048,
+    "init_std": 0.02,
+    "initializer_range": 0.02,
+    "intermediate_size": 11008,
+    "max_position_embeddings": 32768,
+    "max_window_layers": 70,
+    "model_type": "qwen2_5_omni_text",
+    "num_attention_heads": 16,
+    "num_hidden_layers": 36,
+    "num_key_value_heads": 2,
+    "rms_norm_eps": 1e-06,
+    "rope_scaling": {
+      "mrope_section": [
+        16,
+        24,
+        24
+      ],
+      "rope_type": "default",
+      "type": "default"
+    },
+    "rope_theta": 1000000.0,
+    "sliding_window": 32768,
+    "torch_dtype": "float32",
+    "use_cache": true,
+    "use_sliding_window": false,
+    "vocab_size": 151936
+  },
+  "torch_dtype": "bfloat16",
+  "transformers_version": "4.52.3",
+  "user_token_id": 872,
+  "video_token_index": 151656,
+  "vision_config": {
+    "depth": 32,
+    "embed_dim": 1280,
+    "fullatt_block_indexes": [
+      7,
+      15,
+      23,
+      31
+    ],
+    "hidden_act": "silu",
+    "hidden_size": 1280,
+    "in_channels": 3,
+    "in_chans": 3,
+    "init_std": 0.02,
+    "initializer_range": 0.02,
+    "intermediate_size": 3420,
+    "model_type": "qwen2_5_omni_vision_encoder",
+    "num_heads": 16,
+    "out_hidden_size": 2048,
+    "patch_size": 14,
+    "spatial_merge_size": 2,
+    "spatial_patch_size": 14,
+    "temporal_patch_size": 2,
+    "tokens_per_second": 25,
+    "torch_dtype": "float32",
+    "window_size": 112
+  },
+  "vision_end_token_id": 151653,
+  "vision_start_token_id": 151652,
+  "vision_token_id": 151654
+}

generation_config.json

@@ -0,0 +1,4 @@
+{
+  "_from_model_config": true,
+  "transformers_version": "4.52.3"
+}

global_step1000/bf16_zero_pp_rank_0_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c705ff181d09603e0134317870903f91b9cdc4b6cca29cc39c817aac63506339
+size 5095659845

global_step1000/bf16_zero_pp_rank_1_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f4a2f6706f2df846522b385308134ab14aa6397eb90d66b15ec108fd9738343c
+size 5095659845

global_step1000/bf16_zero_pp_rank_2_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7ef19352a7047d93bc93485add87af6d2bc305428a892183858d34554c7174f9
+size 5095659845

global_step1000/bf16_zero_pp_rank_3_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:856541a373c1d9d1c11fd47d1d82d53e31bfa098251f6e60b1f06077b231d91a
+size 5095659845

global_step1000/bf16_zero_pp_rank_4_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5f8ea18c9ee7f3d0c170ba8e56d5def108028cc2c57dff5fb9cc5334f41185c5
+size 5095659845

global_step1000/bf16_zero_pp_rank_5_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9649bd1a81daaf22b8df363b900dcbf584316768a352bdf079d70671af5fd9a7
+size 5095659845

global_step1000/bf16_zero_pp_rank_6_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:43cae29c6db462b3a48364fed529adbfdfc61dfd2b9105ad976163d934c75fb5
+size 5095659845

global_step1000/bf16_zero_pp_rank_7_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bad43e8ee6a97a04dc3823a65f0066de10e860b16fd4d5d13aa670c721805cb7
+size 5095659845

global_step1000/zero_pp_rank_0_mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:abbdcb92f3cf5b1b92742748c6f0ad224744a90892577ed42e1ef48cb6b0b470
+size 327584207

global_step1000/zero_pp_rank_1_mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4867cef3522762acebfdea1f7eff16c807737c5d53bc46d09ebda7d71e21784d
+size 327584207

global_step1000/zero_pp_rank_2_mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:19e2636e9db75f2670d1801cdb4334250fbd233c4099801b9a8390a7b15fded2
+size 327584207

global_step1000/zero_pp_rank_3_mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3636749f4cc2bc86c6039405f692fce4e2b26af5951ba9a50dd7125146971b0a
+size 327584207

global_step1000/zero_pp_rank_4_mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ed61f12ce1e432ab91507198bc6667b9c4c0c9672abee7ebc8d3e98a01566724
+size 327584207

global_step1000/zero_pp_rank_5_mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ce744d1b83005ccb4715deff6dcbf249de7d927d1c127cb911d555a1de908d46
+size 327584207

global_step1000/zero_pp_rank_6_mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8b0fd3695f964c9768c3cdb9c1d0748fbe47c84ead554c8beb44bb9959d184d5
+size 327584207

global_step1000/zero_pp_rank_7_mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:391a48aee32c72a1813ff7435e25510858c6d7e9af0a612e7fa17ae3db5506f2
+size 327584207

latest

@@ -0,0 +1 @@
+global_step1000

model-00001-of-00002.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:14d9216c9eabf00decf0b69a3670b99fa32771fe4fa0b96ae98e8e624b1031ba
+size 4994841696

model-00002-of-00002.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:22b05e23b5345f976d408e55fe9fccb2c44a4574ed3f199e980bcf9c2a51046b
+size 4412249056

rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:48710b9a70599805c18daba6558fae4bd505ad20153bc960c075808b568dc199
+size 16389

rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:64057913c08cbc5a331da1b86f8c9959116d1b8f57da5d46b01964018f9b33bc
+size 16389

rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2fdddd79ebffd22cce641a1f9d877f8f71ee0a1f4d5456b324f1502c24a0afc8
+size 16389

rng_state_3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:55f7ffc5f5ef755ee6f41899cbc1c851dd4b531d5a434b37e2a9a173f5681278
+size 16389

rng_state_4.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4eb86677c9c42301e10d06e3d7d7a13feb36bdc968538c0b45a918e0892865d0
+size 16389

rng_state_5.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9eaf133ee591f519dfd4b31f8e16cd3ce86423781ac46677c03ec2d24b095e13
+size 16389

rng_state_6.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6e0113a7a68d6c2907c60343a63f5b2c8e10a1b9e9f0733803097fe1bc2e8e5c
+size 16389

rng_state_7.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cd16c04f5d95db2915d47583857e72337cae993ae34dfd477f05516135f973b2
+size 16389

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|>"
+}

tokenizer_config.json

@@ -0,0 +1,223 @@
+{
+  "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|>",
+  "model_max_length": 32768,
+  "pad_token": "<|endoftext|>",
+  "processor_class": "Qwen2_5OmniProcessor",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null,
+  "use_fast": false,
+  "video_token": "<|VIDEO|>",
+  "vision_bos_token": "<|vision_bos|>",
+  "vision_eos_token": "<|vision_eos|>"
+}

trainer_state.json

@@ -0,0 +1,894 @@
+{
+  "best_global_step": 700,
+  "best_metric": 0.7593940496444702,
+  "best_model_checkpoint": "./thinker_output/07-08_multi_audio/checkpoint-700",
+  "epoch": 2.692489053553385,
+  "eval_steps": 50,
+  "global_step": 1000,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.02694509936005389,
+      "grad_norm": 7.9127708678466355,
+      "learning_rate": 5e-05,
+      "loss": 2.2253,
+      "step": 10
+    },
+    {
+      "epoch": 0.05389019872010778,
+      "grad_norm": 2.129170820552183,
+      "learning_rate": 4.954792043399639e-05,
+      "loss": 1.4156,
+      "step": 20
+    },
+    {
+      "epoch": 0.08083529808016167,
+      "grad_norm": 2.213231596266166,
+      "learning_rate": 4.909584086799277e-05,
+      "loss": 1.2292,
+      "step": 30
+    },
+    {
+      "epoch": 0.10778039744021556,
+      "grad_norm": 1.9446166933638902,
+      "learning_rate": 4.864376130198916e-05,
+      "loss": 1.1526,
+      "step": 40
+    },
+    {
+      "epoch": 0.13472549680026946,
+      "grad_norm": 1.8624996582053734,
+      "learning_rate": 4.8191681735985535e-05,
+      "loss": 1.1116,
+      "step": 50
+    },
+    {
+      "epoch": 0.13472549680026946,
+      "eval_loss": 1.0801820755004883,
+      "eval_runtime": 116.0,
+      "eval_samples_per_second": 43.103,
+      "eval_steps_per_second": 0.681,
+      "step": 50
+    },
+    {
+      "epoch": 0.16167059616032334,
+      "grad_norm": 1.644929460056369,
+      "learning_rate": 4.773960216998192e-05,
+      "loss": 1.0736,
+      "step": 60
+    },
+    {
+      "epoch": 0.18861569552037724,
+      "grad_norm": 1.9154194240377669,
+      "learning_rate": 4.7287522603978304e-05,
+      "loss": 1.0532,
+      "step": 70
+    },
+    {
+      "epoch": 0.21556079488043112,
+      "grad_norm": 1.5827665298198546,
+      "learning_rate": 4.683544303797468e-05,
+      "loss": 1.0312,
+      "step": 80
+    },
+    {
+      "epoch": 0.24250589424048502,
+      "grad_norm": 1.429127675663522,
+      "learning_rate": 4.638336347197107e-05,
+      "loss": 1.0133,
+      "step": 90
+    },
+    {
+      "epoch": 0.2694509936005389,
+      "grad_norm": 1.4187637596535974,
+      "learning_rate": 4.593128390596745e-05,
+      "loss": 0.9944,
+      "step": 100
+    },
+    {
+      "epoch": 0.2694509936005389,
+      "eval_loss": 0.987108588218689,
+      "eval_runtime": 113.3326,
+      "eval_samples_per_second": 44.118,
+      "eval_steps_per_second": 0.697,
+      "step": 100
+    },
+    {
+      "epoch": 0.29639609296059277,
+      "grad_norm": 1.6143534620181745,
+      "learning_rate": 4.547920433996384e-05,
+      "loss": 0.9861,
+      "step": 110
+    },
+    {
+      "epoch": 0.3233411923206467,
+      "grad_norm": 1.470799422498714,
+      "learning_rate": 4.5027124773960215e-05,
+      "loss": 0.9813,
+      "step": 120
+    },
+    {
+      "epoch": 0.3502862916807006,
+      "grad_norm": 1.5329678036861214,
+      "learning_rate": 4.45750452079566e-05,
+      "loss": 0.9725,
+      "step": 130
+    },
+    {
+      "epoch": 0.3772313910407545,
+      "grad_norm": 1.5683570099493096,
+      "learning_rate": 4.4122965641952984e-05,
+      "loss": 0.9604,
+      "step": 140
+    },
+    {
+      "epoch": 0.40417649040080833,
+      "grad_norm": 1.4886000727881106,
+      "learning_rate": 4.367088607594937e-05,
+      "loss": 0.9494,
+      "step": 150
+    },
+    {
+      "epoch": 0.40417649040080833,
+      "eval_loss": 0.9371287226676941,
+      "eval_runtime": 147.0071,
+      "eval_samples_per_second": 34.012,
+      "eval_steps_per_second": 0.537,
+      "step": 150
+    },
+    {
+      "epoch": 0.43112158976086223,
+      "grad_norm": 1.7059179320901345,
+      "learning_rate": 4.3218806509945754e-05,
+      "loss": 0.9409,
+      "step": 160
+    },
+    {
+      "epoch": 0.45806668912091614,
+      "grad_norm": 1.3194635069630314,
+      "learning_rate": 4.276672694394214e-05,
+      "loss": 0.9244,
+      "step": 170
+    },
+    {
+      "epoch": 0.48501178848097004,
+      "grad_norm": 1.3060771848163875,
+      "learning_rate": 4.2314647377938523e-05,
+      "loss": 0.9157,
+      "step": 180
+    },
+    {
+      "epoch": 0.5119568878410239,
+      "grad_norm": 1.4749915311735549,
+      "learning_rate": 4.186256781193491e-05,
+      "loss": 0.9119,
+      "step": 190
+    },
+    {
+      "epoch": 0.5389019872010778,
+      "grad_norm": 1.4869922949493382,
+      "learning_rate": 4.1410488245931286e-05,
+      "loss": 0.913,
+      "step": 200
+    },
+    {
+      "epoch": 0.5389019872010778,
+      "eval_loss": 0.907091498374939,
+      "eval_runtime": 111.8122,
+      "eval_samples_per_second": 44.718,
+      "eval_steps_per_second": 0.707,
+      "step": 200
+    },
+    {
+      "epoch": 0.5658470865611317,
+      "grad_norm": 1.4160894102096444,
+      "learning_rate": 4.095840867992767e-05,
+      "loss": 0.901,
+      "step": 210
+    },
+    {
+      "epoch": 0.5927921859211855,
+      "grad_norm": 1.5792102697843888,
+      "learning_rate": 4.050632911392405e-05,
+      "loss": 0.89,
+      "step": 220
+    },
+    {
+      "epoch": 0.6197372852812395,
+      "grad_norm": 1.2974993115890197,
+      "learning_rate": 4.0054249547920434e-05,
+      "loss": 0.8863,
+      "step": 230
+    },
+    {
+      "epoch": 0.6466823846412934,
+      "grad_norm": 1.3514939321309911,
+      "learning_rate": 3.960216998191682e-05,
+      "loss": 0.8854,
+      "step": 240
+    },
+    {
+      "epoch": 0.6736274840013473,
+      "grad_norm": 1.266918300985221,
+      "learning_rate": 3.9150090415913203e-05,
+      "loss": 0.8845,
+      "step": 250
+    },
+    {
+      "epoch": 0.6736274840013473,
+      "eval_loss": 0.8746693134307861,
+      "eval_runtime": 144.3762,
+      "eval_samples_per_second": 34.632,
+      "eval_steps_per_second": 0.547,
+      "step": 250
+    },
+    {
+      "epoch": 0.7005725833614012,
+      "grad_norm": 1.3519572583606572,
+      "learning_rate": 3.869801084990959e-05,
+      "loss": 0.8721,
+      "step": 260
+    },
+    {
+      "epoch": 0.727517682721455,
+      "grad_norm": 1.2840451675298208,
+      "learning_rate": 3.8245931283905966e-05,
+      "loss": 0.8736,
+      "step": 270
+    },
+    {
+      "epoch": 0.754462782081509,
+      "grad_norm": 1.2178137913182443,
+      "learning_rate": 3.779385171790235e-05,
+      "loss": 0.8664,
+      "step": 280
+    },
+    {
+      "epoch": 0.7814078814415628,
+      "grad_norm": 1.371387951800443,
+      "learning_rate": 3.7341772151898736e-05,
+      "loss": 0.8554,
+      "step": 290
+    },
+    {
+      "epoch": 0.8083529808016167,
+      "grad_norm": 1.344734634165615,
+      "learning_rate": 3.688969258589512e-05,
+      "loss": 0.8638,
+      "step": 300
+    },
+    {
+      "epoch": 0.8083529808016167,
+      "eval_loss": 0.8491566181182861,
+      "eval_runtime": 146.4614,
+      "eval_samples_per_second": 34.139,
+      "eval_steps_per_second": 0.539,
+      "step": 300
+    },
+    {
+      "epoch": 0.8352980801616706,
+      "grad_norm": 1.272788750854563,
+      "learning_rate": 3.6437613019891505e-05,
+      "loss": 0.8541,
+      "step": 310
+    },
+    {
+      "epoch": 0.8622431795217245,
+      "grad_norm": 1.3792910383877415,
+      "learning_rate": 3.598553345388789e-05,
+      "loss": 0.8481,
+      "step": 320
+    },
+    {
+      "epoch": 0.8891882788817784,
+      "grad_norm": 1.203782825434394,
+      "learning_rate": 3.553345388788427e-05,
+      "loss": 0.8376,
+      "step": 330
+    },
+    {
+      "epoch": 0.9161333782418323,
+      "grad_norm": 1.2646961187289552,
+      "learning_rate": 3.508137432188065e-05,
+      "loss": 0.839,
+      "step": 340
+    },
+    {
+      "epoch": 0.9430784776018861,
+      "grad_norm": 1.24393119289486,
+      "learning_rate": 3.462929475587703e-05,
+      "loss": 0.8381,
+      "step": 350
+    },
+    {
+      "epoch": 0.9430784776018861,
+      "eval_loss": 0.8305084109306335,
+      "eval_runtime": 135.5403,
+      "eval_samples_per_second": 36.889,
+      "eval_steps_per_second": 0.583,
+      "step": 350
+    },
+    {
+      "epoch": 0.9700235769619401,
+      "grad_norm": 1.2087402616528558,
+      "learning_rate": 3.4177215189873416e-05,
+      "loss": 0.8264,
+      "step": 360
+    },
+    {
+      "epoch": 0.9969686763219939,
+      "grad_norm": 1.2771110484442496,
+      "learning_rate": 3.37251356238698e-05,
+      "loss": 0.8242,
+      "step": 370
+    },
+    {
+      "epoch": 1.0215560794880432,
+      "grad_norm": 1.332831610585688,
+      "learning_rate": 3.3273056057866185e-05,
+      "loss": 0.6297,
+      "step": 380
+    },
+    {
+      "epoch": 1.048501178848097,
+      "grad_norm": 1.2917409109187712,
+      "learning_rate": 3.282097649186257e-05,
+      "loss": 0.6539,
+      "step": 390
+    },
+    {
+      "epoch": 1.0754462782081509,
+      "grad_norm": 1.3245733128879162,
+      "learning_rate": 3.2368896925858955e-05,
+      "loss": 0.6544,
+      "step": 400
+    },
+    {
+      "epoch": 1.0754462782081509,
+      "eval_loss": 0.8334468007087708,
+      "eval_runtime": 148.2965,
+      "eval_samples_per_second": 33.716,
+      "eval_steps_per_second": 0.533,
+      "step": 400
+    },
+    {
+      "epoch": 1.1023913775682048,
+      "grad_norm": 1.1724140281525666,
+      "learning_rate": 3.191681735985534e-05,
+      "loss": 0.6473,
+      "step": 410
+    },
+    {
+      "epoch": 1.1293364769282586,
+      "grad_norm": 1.2278740780504742,
+      "learning_rate": 3.146473779385172e-05,
+      "loss": 0.6459,
+      "step": 420
+    },
+    {
+      "epoch": 1.1562815762883125,
+      "grad_norm": 1.1783670358458123,
+      "learning_rate": 3.10126582278481e-05,
+      "loss": 0.6573,
+      "step": 430
+    },
+    {
+      "epoch": 1.1832266756483665,
+      "grad_norm": 1.29897205539554,
+      "learning_rate": 3.056057866184449e-05,
+      "loss": 0.6548,
+      "step": 440
+    },
+    {
+      "epoch": 1.2101717750084204,
+      "grad_norm": 1.1568119957121505,
+      "learning_rate": 3.010849909584087e-05,
+      "loss": 0.65,
+      "step": 450
+    },
+    {
+      "epoch": 1.2101717750084204,
+      "eval_loss": 0.8229044675827026,
+      "eval_runtime": 113.6433,
+      "eval_samples_per_second": 43.997,
+      "eval_steps_per_second": 0.695,
+      "step": 450
+    },
+    {
+      "epoch": 1.2371168743684742,
+      "grad_norm": 1.3004916817149637,
+      "learning_rate": 2.9656419529837253e-05,
+      "loss": 0.6474,
+      "step": 460
+    },
+    {
+      "epoch": 1.2640619737285281,
+      "grad_norm": 1.212521588358061,
+      "learning_rate": 2.9204339963833638e-05,
+      "loss": 0.6518,
+      "step": 470
+    },
+    {
+      "epoch": 1.291007073088582,
+      "grad_norm": 1.2553077221366877,
+      "learning_rate": 2.8752260397830023e-05,
+      "loss": 0.6456,
+      "step": 480
+    },
+    {
+      "epoch": 1.3179521724486358,
+      "grad_norm": 1.184218802614123,
+      "learning_rate": 2.83001808318264e-05,
+      "loss": 0.6546,
+      "step": 490
+    },
+    {
+      "epoch": 1.3448972718086898,
+      "grad_norm": 1.136986753079325,
+      "learning_rate": 2.7848101265822786e-05,
+      "loss": 0.6415,
+      "step": 500
+    },
+    {
+      "epoch": 1.3448972718086898,
+      "eval_loss": 0.8110851049423218,
+      "eval_runtime": 137.5711,
+      "eval_samples_per_second": 36.345,
+      "eval_steps_per_second": 0.574,
+      "step": 500
+    },
+    {
+      "epoch": 1.3718423711687437,
+      "grad_norm": 1.2143646330707367,
+      "learning_rate": 2.7396021699819167e-05,
+      "loss": 0.6558,
+      "step": 510
+    },
+    {
+      "epoch": 1.3987874705287977,
+      "grad_norm": 1.265789637913618,
+      "learning_rate": 2.6943942133815552e-05,
+      "loss": 0.6505,
+      "step": 520
+    },
+    {
+      "epoch": 1.4257325698888514,
+      "grad_norm": 1.2191031554639078,
+      "learning_rate": 2.6491862567811937e-05,
+      "loss": 0.6426,
+      "step": 530
+    },
+    {
+      "epoch": 1.4526776692489054,
+      "grad_norm": 1.222440942937553,
+      "learning_rate": 2.603978300180832e-05,
+      "loss": 0.644,
+      "step": 540
+    },
+    {
+      "epoch": 1.4796227686089591,
+      "grad_norm": 1.1495752286958087,
+      "learning_rate": 2.5587703435804706e-05,
+      "loss": 0.6415,
+      "step": 550
+    },
+    {
+      "epoch": 1.4796227686089591,
+      "eval_loss": 0.7958658337593079,
+      "eval_runtime": 142.3592,
+      "eval_samples_per_second": 35.122,
+      "eval_steps_per_second": 0.555,
+      "step": 550
+    },
+    {
+      "epoch": 1.506567867969013,
+      "grad_norm": 1.2049482092443289,
+      "learning_rate": 2.5135623869801084e-05,
+      "loss": 0.6395,
+      "step": 560
+    },
+    {
+      "epoch": 1.533512967329067,
+      "grad_norm": 1.1839061997290048,
+      "learning_rate": 2.468354430379747e-05,
+      "loss": 0.6441,
+      "step": 570
+    },
+    {
+      "epoch": 1.560458066689121,
+      "grad_norm": 1.2057965805365276,
+      "learning_rate": 2.423146473779385e-05,
+      "loss": 0.6335,
+      "step": 580
+    },
+    {
+      "epoch": 1.5874031660491748,
+      "grad_norm": 1.1650282316989717,
+      "learning_rate": 2.3779385171790235e-05,
+      "loss": 0.6339,
+      "step": 590
+    },
+    {
+      "epoch": 1.6143482654092287,
+      "grad_norm": 1.153394411032144,
+      "learning_rate": 2.332730560578662e-05,
+      "loss": 0.6311,
+      "step": 600
+    },
+    {
+      "epoch": 1.6143482654092287,
+      "eval_loss": 0.784841775894165,
+      "eval_runtime": 153.2005,
+      "eval_samples_per_second": 32.637,
+      "eval_steps_per_second": 0.516,
+      "step": 600
+    },
+    {
+      "epoch": 1.6412933647692824,
+      "grad_norm": 1.1963943501637924,
+      "learning_rate": 2.2875226039783005e-05,
+      "loss": 0.6267,
+      "step": 610
+    },
+    {
+      "epoch": 1.6682384641293364,
+      "grad_norm": 1.093066884488607,
+      "learning_rate": 2.2423146473779386e-05,
+      "loss": 0.6289,
+      "step": 620
+    },
+    {
+      "epoch": 1.6951835634893904,
+      "grad_norm": 1.115011570991967,
+      "learning_rate": 2.197106690777577e-05,
+      "loss": 0.6299,
+      "step": 630
+    },
+    {
+      "epoch": 1.7221286628494443,
+      "grad_norm": 1.1700606931618611,
+      "learning_rate": 2.1518987341772153e-05,
+      "loss": 0.6233,
+      "step": 640
+    },
+    {
+      "epoch": 1.7490737622094983,
+      "grad_norm": 1.1565551201360744,
+      "learning_rate": 2.1066907775768534e-05,
+      "loss": 0.624,
+      "step": 650
+    },
+    {
+      "epoch": 1.7490737622094983,
+      "eval_loss": 0.7739421129226685,
+      "eval_runtime": 148.2001,
+      "eval_samples_per_second": 33.738,
+      "eval_steps_per_second": 0.533,
+      "step": 650
+    },
+    {
+      "epoch": 1.776018861569552,
+      "grad_norm": 1.2558394571278007,
+      "learning_rate": 2.061482820976492e-05,
+      "loss": 0.621,
+      "step": 660
+    },
+    {
+      "epoch": 1.802963960929606,
+      "grad_norm": 1.1487412967519839,
+      "learning_rate": 2.0162748643761304e-05,
+      "loss": 0.6201,
+      "step": 670
+    },
+    {
+      "epoch": 1.8299090602896597,
+      "grad_norm": 1.1633700130604714,
+      "learning_rate": 1.971066907775769e-05,
+      "loss": 0.6182,
+      "step": 680
+    },
+    {
+      "epoch": 1.8568541596497137,
+      "grad_norm": 1.0892080498580619,
+      "learning_rate": 1.925858951175407e-05,
+      "loss": 0.6218,
+      "step": 690
+    },
+    {
+      "epoch": 1.8837992590097676,
+      "grad_norm": 1.162193434275119,
+      "learning_rate": 1.8806509945750454e-05,
+      "loss": 0.612,
+      "step": 700
+    },
+    {
+      "epoch": 1.8837992590097676,
+      "eval_loss": 0.7593940496444702,
+      "eval_runtime": 113.5973,
+      "eval_samples_per_second": 44.015,
+      "eval_steps_per_second": 0.695,
+      "step": 700
+    },
+    {
+      "epoch": 1.9107443583698216,
+      "grad_norm": 1.477614030615022,
+      "learning_rate": 1.8354430379746836e-05,
+      "loss": 0.4361,
+      "step": 710
+    },
+    {
+      "epoch": 1.9376894577298753,
+      "grad_norm": 1.211596696990758,
+      "learning_rate": 1.7902350813743217e-05,
+      "loss": 0.4174,
+      "step": 720
+    },
+    {
+      "epoch": 1.9646345570899293,
+      "grad_norm": 1.208841417920371,
+      "learning_rate": 1.7450271247739602e-05,
+      "loss": 0.4162,
+      "step": 730
+    },
+    {
+      "epoch": 1.991579656449983,
+      "grad_norm": 1.2740412194613278,
+      "learning_rate": 1.6998191681735987e-05,
+      "loss": 0.4115,
+      "step": 740
+    },
+    {
+      "epoch": 2.0188615695520378,
+      "grad_norm": 1.1897059830697447,
+      "learning_rate": 1.654611211573237e-05,
+      "loss": 0.4224,
+      "step": 750
+    },
+    {
+      "epoch": 2.0188615695520378,
+      "eval_loss": 0.8485522270202637,
+      "eval_runtime": 115.805,
+      "eval_samples_per_second": 43.176,
+      "eval_steps_per_second": 0.682,
+      "step": 750
+    },
+    {
+      "epoch": 2.0458066689120917,
+      "grad_norm": 1.2120571201563564,
+      "learning_rate": 1.6094032549728753e-05,
+      "loss": 0.4192,
+      "step": 760
+    },
+    {
+      "epoch": 2.0727517682721457,
+      "grad_norm": 1.2006852199531266,
+      "learning_rate": 1.5641952983725134e-05,
+      "loss": 0.4101,
+      "step": 770
+    },
+    {
+      "epoch": 2.099696867632199,
+      "grad_norm": 1.2005979870584054,
+      "learning_rate": 1.5189873417721521e-05,
+      "loss": 0.4164,
+      "step": 780
+    },
+    {
+      "epoch": 2.126641966992253,
+      "grad_norm": 1.148665043003728,
+      "learning_rate": 1.4737793851717904e-05,
+      "loss": 0.4194,
+      "step": 790
+    },
+    {
+      "epoch": 2.153587066352307,
+      "grad_norm": 1.134303382163784,
+      "learning_rate": 1.4285714285714285e-05,
+      "loss": 0.4213,
+      "step": 800
+    },
+    {
+      "epoch": 2.153587066352307,
+      "eval_loss": 0.838707447052002,
+      "eval_runtime": 139.7297,
+      "eval_samples_per_second": 35.783,
+      "eval_steps_per_second": 0.565,
+      "step": 800
+    },
+    {
+      "epoch": 2.180532165712361,
+      "grad_norm": 1.2301409422954543,
+      "learning_rate": 1.383363471971067e-05,
+      "loss": 0.4179,
+      "step": 810
+    },
+    {
+      "epoch": 2.207477265072415,
+      "grad_norm": 1.1312377076286286,
+      "learning_rate": 1.3381555153707053e-05,
+      "loss": 0.4178,
+      "step": 820
+    },
+    {
+      "epoch": 2.234422364432469,
+      "grad_norm": 1.2707527556350058,
+      "learning_rate": 1.2929475587703435e-05,
+      "loss": 0.4125,
+      "step": 830
+    },
+    {
+      "epoch": 2.2613674637925225,
+      "grad_norm": 1.185885154215189,
+      "learning_rate": 1.247739602169982e-05,
+      "loss": 0.414,
+      "step": 840
+    },
+    {
+      "epoch": 2.2883125631525765,
+      "grad_norm": 1.1012295953218187,
+      "learning_rate": 1.2025316455696203e-05,
+      "loss": 0.414,
+      "step": 850
+    },
+    {
+      "epoch": 2.2883125631525765,
+      "eval_loss": 0.831937313079834,
+      "eval_runtime": 144.9314,
+      "eval_samples_per_second": 34.499,
+      "eval_steps_per_second": 0.545,
+      "step": 850
+    },
+    {
+      "epoch": 2.3152576625126304,
+      "grad_norm": 1.1388819607995708,
+      "learning_rate": 1.1573236889692586e-05,
+      "loss": 0.4151,
+      "step": 860
+    },
+    {
+      "epoch": 2.3422027618726844,
+      "grad_norm": 1.2140213674018405,
+      "learning_rate": 1.112115732368897e-05,
+      "loss": 0.4159,
+      "step": 870
+    },
+    {
+      "epoch": 2.3691478612327384,
+      "grad_norm": 1.1631774357185438,
+      "learning_rate": 1.0669077757685354e-05,
+      "loss": 0.4136,
+      "step": 880
+    },
+    {
+      "epoch": 2.3960929605927923,
+      "grad_norm": 1.24334716793248,
+      "learning_rate": 1.0216998191681737e-05,
+      "loss": 0.4076,
+      "step": 890
+    },
+    {
+      "epoch": 2.4230380599528463,
+      "grad_norm": 1.156296740645565,
+      "learning_rate": 9.76491862567812e-06,
+      "loss": 0.4145,
+      "step": 900
+    },
+    {
+      "epoch": 2.4230380599528463,
+      "eval_loss": 0.823464572429657,
+      "eval_runtime": 141.3831,
+      "eval_samples_per_second": 35.365,
+      "eval_steps_per_second": 0.559,
+      "step": 900
+    },
+    {
+      "epoch": 2.4499831593129,
+      "grad_norm": 1.2360058092342656,
+      "learning_rate": 9.312839059674505e-06,
+      "loss": 0.4071,
+      "step": 910
+    },
+    {
+      "epoch": 2.4769282586729537,
+      "grad_norm": 1.1672627428975908,
+      "learning_rate": 8.860759493670886e-06,
+      "loss": 0.4052,
+      "step": 920
+    },
+    {
+      "epoch": 2.5038733580330077,
+      "grad_norm": 1.142198240950353,
+      "learning_rate": 8.408679927667269e-06,
+      "loss": 0.3995,
+      "step": 930
+    },
+    {
+      "epoch": 2.5308184573930617,
+      "grad_norm": 1.1336163759218327,
+      "learning_rate": 7.956600361663654e-06,
+      "loss": 0.4109,
+      "step": 940
+    },
+    {
+      "epoch": 2.5577635567531156,
+      "grad_norm": 1.1520513715336371,
+      "learning_rate": 7.504520795660036e-06,
+      "loss": 0.4021,
+      "step": 950
+    },
+    {
+      "epoch": 2.5577635567531156,
+      "eval_loss": 0.8257409930229187,
+      "eval_runtime": 145.0235,
+      "eval_samples_per_second": 34.477,
+      "eval_steps_per_second": 0.545,
+      "step": 950
+    },
+    {
+      "epoch": 2.5847086561131696,
+      "grad_norm": 1.1137257622476067,
+      "learning_rate": 7.05244122965642e-06,
+      "loss": 0.3979,
+      "step": 960
+    },
+    {
+      "epoch": 2.611653755473223,
+      "grad_norm": 1.107368997807173,
+      "learning_rate": 6.600361663652803e-06,
+      "loss": 0.4011,
+      "step": 970
+    },
+    {
+      "epoch": 2.638598854833277,
+      "grad_norm": 1.1780589934268402,
+      "learning_rate": 6.148282097649186e-06,
+      "loss": 0.4013,
+      "step": 980
+    },
+    {
+      "epoch": 2.665543954193331,
+      "grad_norm": 1.174979594904003,
+      "learning_rate": 5.69620253164557e-06,
+      "loss": 0.3992,
+      "step": 990
+    },
+    {
+      "epoch": 2.692489053553385,
+      "grad_norm": 1.1164061245257322,
+      "learning_rate": 5.244122965641953e-06,
+      "loss": 0.393,
+      "step": 1000
+    },
+    {
+      "epoch": 2.692489053553385,
+      "eval_loss": 0.8172587156295776,
+      "eval_runtime": 147.8118,
+      "eval_samples_per_second": 33.827,
+      "eval_steps_per_second": 0.534,
+      "step": 1000
+    }
+  ],
+  "logging_steps": 10,
+  "max_steps": 1116,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 3,
+  "save_steps": 50,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 588328797732864.0,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}

training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7142dd73a436ab023d78446a9700a5d3d3376496f1ff098532e2130dec697703
+size 8017

zero_to_fp32.py

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