Upload folder using huggingface_hub

.gitattributes

@@ -38,3 +38,4 @@ limo/full/checkpoint-1030/tokenizer.json filter=lfs diff=lfs merge=lfs -text
 limo/full/checkpoint-1545/tokenizer.json filter=lfs diff=lfs merge=lfs -text
 s1k-1.1/full/checkpoint-75/tokenizer.json filter=lfs diff=lfs merge=lfs -text
 s1k-1.1/full/checkpoint-150/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+s1k-1.1/full/checkpoint-225/tokenizer.json filter=lfs diff=lfs merge=lfs -text

s1k-1.1/full/checkpoint-225/config.json

@@ -0,0 +1,30 @@
+{
+  "_name_or_path": "/mnt/public/data/lh/models/DeepSeek-R1-Distill-Qwen-1.5B",
+  "architectures": [
+    "Qwen2ForCausalLM"
+  ],
+  "attention_dropout": 0.0,
+  "bos_token_id": 151643,
+  "eos_token_id": 151643,
+  "hidden_act": "silu",
+  "hidden_size": 1536,
+  "initializer_range": 0.02,
+  "intermediate_size": 8960,
+  "max_position_embeddings": 131072,
+  "max_window_layers": 21,
+  "model_type": "qwen2",
+  "num_attention_heads": 12,
+  "num_hidden_layers": 28,
+  "num_key_value_heads": 2,
+  "rms_norm_eps": 1e-06,
+  "rope_scaling": null,
+  "rope_theta": 10000,
+  "sliding_window": null,
+  "tie_word_embeddings": false,
+  "torch_dtype": "bfloat16",
+  "transformers_version": "4.46.1",
+  "use_cache": false,
+  "use_mrope": false,
+  "use_sliding_window": false,
+  "vocab_size": 151936
+}

s1k-1.1/full/checkpoint-225/generation_config.json

@@ -0,0 +1,9 @@
+{
+  "_from_model_config": true,
+  "bos_token_id": 151646,
+  "do_sample": true,
+  "eos_token_id": 151643,
+  "temperature": 0.6,
+  "top_p": 0.95,
+  "transformers_version": "4.46.1"
+}

s1k-1.1/full/checkpoint-225/global_step225/bf16_zero_pp_rank_0_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:722b59518703b26027322f1ece90274a9b8ea3438595be8e28d52704dbd46c10
+size 2665636656

s1k-1.1/full/checkpoint-225/global_step225/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:a01146c9649b4ae679de89928c4eefffac5f86b8daef49266f91e7026b032aec
+size 2665636656

s1k-1.1/full/checkpoint-225/global_step225/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:abc81616062d9a803f9a2700f7a3bd44ef64bc3a5d893035f8b151d7f2255ca4
+size 2665636656

s1k-1.1/full/checkpoint-225/global_step225/bf16_zero_pp_rank_3_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:95baef54d5b319df9beb10358037673f5f81cf15e8f6f6d5475451640fafe93a
+size 2665636656

s1k-1.1/full/checkpoint-225/global_step225/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:d99d386ae91463be483a7fb4e330f906e08b6a39274eb898495576252eec9a9d
+size 2665636656

s1k-1.1/full/checkpoint-225/global_step225/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:ea2f1f2b4b14240db891543cb722f6c8c7afd7e5f9b68e07e35a48a1a7b4e0b0
+size 2665636656

s1k-1.1/full/checkpoint-225/global_step225/bf16_zero_pp_rank_6_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:22624d97d1265852e023bf09303f2f621bac7f64114ba78eb8e3d80f80009f32
+size 2665636656

s1k-1.1/full/checkpoint-225/global_step225/bf16_zero_pp_rank_7_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7ae4f1021be6e3e455d73fe27ec4b3f60f96888c4840c6acf802fba14c80cbe6
+size 2665636656

s1k-1.1/full/checkpoint-225/global_step225/zero_pp_rank_0_mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fb15bcd1c9b1afd2e0f745959fa8aa8e71dda0275d04d5965a6eb5cdeaf953e2
+size 168405

s1k-1.1/full/checkpoint-225/global_step225/zero_pp_rank_1_mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:830b6b75d6461408a503fefde50d0ec9473a212daacb818c0669ec7ebcdf59cb
+size 168405

s1k-1.1/full/checkpoint-225/global_step225/zero_pp_rank_2_mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0414b924ea4d866f19475c27a987bd3032ef66eb60b331897cc1e2af1922b4e2
+size 168405

s1k-1.1/full/checkpoint-225/global_step225/zero_pp_rank_3_mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2a8656fe6b4564422b4acc3b3e19d1828a24fd2f5e8b2a00ed66f137288348fa
+size 168405

s1k-1.1/full/checkpoint-225/global_step225/zero_pp_rank_4_mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f0de8a03d2d0abc6601602492a055da3a26029bee99f40bef9af668651ee15ec
+size 168405

s1k-1.1/full/checkpoint-225/global_step225/zero_pp_rank_5_mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:aa930234ec57f829c16bcc9aaf35b69687e62b39dada817b4f9c7c52e1ea23be
+size 168405

s1k-1.1/full/checkpoint-225/global_step225/zero_pp_rank_6_mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4d43c20dfe2ba7b959d176ab94fe5185add47aa507964b2a01375352527ba476
+size 168405

s1k-1.1/full/checkpoint-225/global_step225/zero_pp_rank_7_mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e2746163875458e7ca84c20471d1853cf54af86c83caaca591368dc1b46fa2e3
+size 168405

s1k-1.1/full/checkpoint-225/latest

@@ -0,0 +1 @@
+global_step225

s1k-1.1/full/checkpoint-225/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c7d77f9ab393e40b24fa4207d69449526fea35af8e9eff95aee7e7b50dd6eecf
+size 3554214752

s1k-1.1/full/checkpoint-225/rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:07d994b317c4df888a1a1aabc0c532e81f1fa34c18c8313cb2feadca3bb37194
+size 15984

s1k-1.1/full/checkpoint-225/rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4f5b05860618aa49c7f5d8c366d6ee73cf8b3b0d0adc17d9313b72621630d0aa
+size 15984

s1k-1.1/full/checkpoint-225/rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7262faf861e984775b4fd85bc76a11b0b8b04037690e8a08a58cf9ff5328a042
+size 15984

s1k-1.1/full/checkpoint-225/rng_state_3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9479cad91150e2e266d17eb95fe678579a770f6df6b53496cf72067b186b094d
+size 15984

s1k-1.1/full/checkpoint-225/rng_state_4.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:435cb6cf559e0ce3fe0d4582cac16ea40b48b7a64589952402a4c399cafbfc00
+size 15984

s1k-1.1/full/checkpoint-225/rng_state_5.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f51001b0d8dc5792180c3a9705ccbfa66b61d46d7639afb6f7abf409629ed74f
+size 15984

s1k-1.1/full/checkpoint-225/rng_state_6.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e1e87084f11088fdce293e1fbbb05e35f5c7385b00e2f9ba195bf61cb36f757d
+size 15984

s1k-1.1/full/checkpoint-225/rng_state_7.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3d32e9bdd65145ae509e6c6ef4f6ea9d842f94a34c34a0d7d2ab6c248d3f2121
+size 15984

s1k-1.1/full/checkpoint-225/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:736fd5de2fc57872d8f115e475efd2d8ab0cd7c18acc1cd716b760a78e0c760a
+size 1064

s1k-1.1/full/checkpoint-225/special_tokens_map.json

@@ -0,0 +1,32 @@
+{
+  "additional_special_tokens": [
+    {
+      "content": "<|im_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false
+    }
+  ],
+  "bos_token": {
+    "content": "<｜begin▁of▁sentence｜>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "eos_token": {
+    "content": "<｜end▁of▁sentence｜>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": {
+    "content": "<｜end▁of▁sentence｜>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

s1k-1.1/full/checkpoint-225/tokenizer.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:02643f00207dfc5ed248992486bde04314c21dca556bf65ce520690962b8db63
+size 11422965

s1k-1.1/full/checkpoint-225/tokenizer_config.json

@@ -0,0 +1,207 @@
+{
+  "add_bos_token": true,
+  "add_eos_token": false,
+  "add_prefix_space": null,
+  "added_tokens_decoder": {
+    "151643": {
+      "content": "<｜end▁of▁sentence｜>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151644": {
+      "content": "<｜User｜>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151645": {
+      "content": "<｜Assistant｜>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151646": {
+      "content": "<｜begin▁of▁sentence｜>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151647": {
+      "content": "<|EOT|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151648": {
+      "content": "<think>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151649": {
+      "content": "</think>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151650": {
+      "content": "<|quad_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151651": {
+      "content": "<|quad_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151652": {
+      "content": "<|vision_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151653": {
+      "content": "<|vision_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151654": {
+      "content": "<|vision_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151655": {
+      "content": "<|image_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151656": {
+      "content": "<|video_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151657": {
+      "content": "<tool_call>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151658": {
+      "content": "</tool_call>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151659": {
+      "content": "<|fim_prefix|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151660": {
+      "content": "<|fim_middle|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151661": {
+      "content": "<|fim_suffix|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151662": {
+      "content": "<|fim_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151663": {
+      "content": "<|repo_name|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151664": {
+      "content": "<|file_sep|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151665": {
+      "content": "<|im_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    }
+  },
+  "additional_special_tokens": [
+    "<|im_end|>"
+  ],
+  "bos_token": "<｜begin▁of▁sentence｜>",
+  "chat_template": "{% if not add_generation_prompt is defined %}{% set add_generation_prompt = false %}{% endif %}{% set ns = namespace(is_first=false, is_tool=false, is_output_first=true, system_prompt='') %}{%- for message in messages %}{%- if message['role'] == 'system' %}{% set ns.system_prompt = message['content'] %}{%- endif %}{%- endfor %}{{bos_token}}{{ns.system_prompt}}{%- for message in messages %}{%- if message['role'] == 'user' %}{%- set ns.is_tool = false -%}{{'<｜User｜>' + message['content']}}{%- endif %}{%- if message['role'] == 'assistant' and message['content'] is none %}{%- set ns.is_tool = false -%}{%- for tool in message['tool_calls']%}{%- if not ns.is_first %}{{'<｜Assistant｜><｜tool▁calls▁begin｜><｜tool▁call▁begin｜>' + tool['type'] + '<｜tool▁sep｜>' + tool['function']['name'] + '\\n' + '```json' + '\\n' + tool['function']['arguments'] + '\\n' + '```' + '<｜tool▁call▁end｜>'}}{%- set ns.is_first = true -%}{%- else %}{{'\\n' + '<｜tool▁call▁begin｜>' + tool['type'] + '<｜tool▁sep｜>' + tool['function']['name'] + '\\n' + '```json' + '\\n' + tool['function']['arguments'] + '\\n' + '```' + '<｜tool▁call▁end｜>'}}{{'<｜tool▁calls▁end｜><｜end▁of▁sentence｜>'}}{%- endif %}{%- endfor %}{%- endif %}{%- if message['role'] == 'assistant' and message['content'] is not none %}{%- if ns.is_tool %}{{'<｜tool▁outputs▁end｜>' + message['content'] + '<｜end▁of▁sentence｜>'}}{%- set ns.is_tool = false -%}{%- else %}{% set content = message['content'] %}{% if '</think>' in content %}{% set content = content.split('</think>')[-1] %}{% endif %}{{'<｜Assistant｜>' + content + '<｜end▁of▁sentence｜>'}}{%- endif %}{%- endif %}{%- if message['role'] == 'tool' %}{%- set ns.is_tool = true -%}{%- if ns.is_output_first %}{{'<｜tool▁outputs▁begin｜><｜tool▁output▁begin｜>' + message['content'] + '<｜tool▁output▁end｜>'}}{%- set ns.is_output_first = false %}{%- else %}{{'\\n<｜tool▁output▁begin｜>' + message['content'] + '<｜tool▁output▁end｜>'}}{%- endif %}{%- endif %}{%- endfor -%}{% if ns.is_tool %}{{'<｜tool▁outputs▁end｜>'}}{% endif %}{% if add_generation_prompt and not ns.is_tool %}{{'<｜Assistant｜><think>\\n'}}{% endif %}",
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<｜end▁of▁sentence｜>",
+  "legacy": true,
+  "model_max_length": 20000,
+  "pad_token": "<｜end▁of▁sentence｜>",
+  "padding_side": "right",
+  "sp_model_kwargs": {},
+  "split_special_tokens": false,
+  "tokenizer_class": "LlamaTokenizer",
+  "unk_token": null,
+  "use_default_system_prompt": false
+}

s1k-1.1/full/checkpoint-225/trainer_state.json

@@ -0,0 +1,1608 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 14.4,
+  "eval_steps": 500,
+  "global_step": 225,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.064,
+      "grad_norm": 1.5593382120132446,
+      "learning_rate": 9.999512620046523e-06,
+      "loss": 0.9965,
+      "step": 1
+    },
+    {
+      "epoch": 0.128,
+      "grad_norm": 1.223217487335205,
+      "learning_rate": 9.998050575201772e-06,
+      "loss": 0.9668,
+      "step": 2
+    },
+    {
+      "epoch": 0.192,
+      "grad_norm": 1.0882679224014282,
+      "learning_rate": 9.995614150494293e-06,
+      "loss": 0.9414,
+      "step": 3
+    },
+    {
+      "epoch": 0.256,
+      "grad_norm": 1.0166572332382202,
+      "learning_rate": 9.992203820909906e-06,
+      "loss": 1.0033,
+      "step": 4
+    },
+    {
+      "epoch": 0.32,
+      "grad_norm": 0.5519472360610962,
+      "learning_rate": 9.987820251299121e-06,
+      "loss": 0.9555,
+      "step": 5
+    },
+    {
+      "epoch": 0.384,
+      "grad_norm": 0.39769354462623596,
+      "learning_rate": 9.982464296247523e-06,
+      "loss": 0.9841,
+      "step": 6
+    },
+    {
+      "epoch": 0.448,
+      "grad_norm": 0.3815975487232208,
+      "learning_rate": 9.976136999909156e-06,
+      "loss": 0.9409,
+      "step": 7
+    },
+    {
+      "epoch": 0.512,
+      "grad_norm": 0.792307436466217,
+      "learning_rate": 9.968839595802982e-06,
+      "loss": 0.9572,
+      "step": 8
+    },
+    {
+      "epoch": 0.576,
+      "grad_norm": 0.8832745552062988,
+      "learning_rate": 9.960573506572391e-06,
+      "loss": 0.9821,
+      "step": 9
+    },
+    {
+      "epoch": 0.64,
+      "grad_norm": 0.8655721545219421,
+      "learning_rate": 9.951340343707852e-06,
+      "loss": 0.9556,
+      "step": 10
+    },
+    {
+      "epoch": 0.704,
+      "grad_norm": 0.7657322287559509,
+      "learning_rate": 9.941141907232766e-06,
+      "loss": 0.8964,
+      "step": 11
+    },
+    {
+      "epoch": 0.768,
+      "grad_norm": 0.6843923330307007,
+      "learning_rate": 9.929980185352525e-06,
+      "loss": 0.9246,
+      "step": 12
+    },
+    {
+      "epoch": 0.832,
+      "grad_norm": 0.5020281076431274,
+      "learning_rate": 9.91785735406693e-06,
+      "loss": 0.8841,
+      "step": 13
+    },
+    {
+      "epoch": 0.896,
+      "grad_norm": 0.3772108852863312,
+      "learning_rate": 9.904775776745959e-06,
+      "loss": 0.8916,
+      "step": 14
+    },
+    {
+      "epoch": 0.96,
+      "grad_norm": 0.3307054042816162,
+      "learning_rate": 9.890738003669029e-06,
+      "loss": 0.9113,
+      "step": 15
+    },
+    {
+      "epoch": 1.024,
+      "grad_norm": 7.4732513427734375,
+      "learning_rate": 9.875746771527817e-06,
+      "loss": 1.1901,
+      "step": 16
+    },
+    {
+      "epoch": 1.088,
+      "grad_norm": 0.5649796724319458,
+      "learning_rate": 9.859805002892733e-06,
+      "loss": 0.9004,
+      "step": 17
+    },
+    {
+      "epoch": 1.152,
+      "grad_norm": 0.48279643058776855,
+      "learning_rate": 9.842915805643156e-06,
+      "loss": 0.9656,
+      "step": 18
+    },
+    {
+      "epoch": 1.216,
+      "grad_norm": 0.386542409658432,
+      "learning_rate": 9.825082472361558e-06,
+      "loss": 0.8151,
+      "step": 19
+    },
+    {
+      "epoch": 1.28,
+      "grad_norm": 0.45578762888908386,
+      "learning_rate": 9.806308479691595e-06,
+      "loss": 1.01,
+      "step": 20
+    },
+    {
+      "epoch": 1.3439999999999999,
+      "grad_norm": 0.42114177346229553,
+      "learning_rate": 9.786597487660336e-06,
+      "loss": 0.8718,
+      "step": 21
+    },
+    {
+      "epoch": 1.408,
+      "grad_norm": 0.3763532340526581,
+      "learning_rate": 9.765953338964736e-06,
+      "loss": 0.9343,
+      "step": 22
+    },
+    {
+      "epoch": 1.472,
+      "grad_norm": 0.29271769523620605,
+      "learning_rate": 9.744380058222483e-06,
+      "loss": 0.8225,
+      "step": 23
+    },
+    {
+      "epoch": 1.536,
+      "grad_norm": 0.2854185402393341,
+      "learning_rate": 9.721881851187406e-06,
+      "loss": 0.9583,
+      "step": 24
+    },
+    {
+      "epoch": 1.6,
+      "grad_norm": 0.19865305721759796,
+      "learning_rate": 9.698463103929542e-06,
+      "loss": 0.8836,
+      "step": 25
+    },
+    {
+      "epoch": 1.6640000000000001,
+      "grad_norm": 0.19694137573242188,
+      "learning_rate": 9.674128381980073e-06,
+      "loss": 0.9426,
+      "step": 26
+    },
+    {
+      "epoch": 1.728,
+      "grad_norm": 0.18151699006557465,
+      "learning_rate": 9.648882429441258e-06,
+      "loss": 0.9363,
+      "step": 27
+    },
+    {
+      "epoch": 1.792,
+      "grad_norm": 0.19295278191566467,
+      "learning_rate": 9.622730168061568e-06,
+      "loss": 0.8627,
+      "step": 28
+    },
+    {
+      "epoch": 1.8559999999999999,
+      "grad_norm": 0.25808241963386536,
+      "learning_rate": 9.595676696276173e-06,
+      "loss": 1.0116,
+      "step": 29
+    },
+    {
+      "epoch": 1.92,
+      "grad_norm": 0.22762736678123474,
+      "learning_rate": 9.567727288213005e-06,
+      "loss": 0.8429,
+      "step": 30
+    },
+    {
+      "epoch": 1.984,
+      "grad_norm": 0.29724806547164917,
+      "learning_rate": 9.538887392664544e-06,
+      "loss": 1.0651,
+      "step": 31
+    },
+    {
+      "epoch": 2.048,
+      "grad_norm": 0.3138919472694397,
+      "learning_rate": 9.50916263202557e-06,
+      "loss": 1.1384,
+      "step": 32
+    },
+    {
+      "epoch": 2.112,
+      "grad_norm": 0.19795119762420654,
+      "learning_rate": 9.478558801197065e-06,
+      "loss": 0.865,
+      "step": 33
+    },
+    {
+      "epoch": 2.176,
+      "grad_norm": 0.21303512156009674,
+      "learning_rate": 9.44708186645649e-06,
+      "loss": 0.9073,
+      "step": 34
+    },
+    {
+      "epoch": 2.24,
+      "grad_norm": 0.1803194284439087,
+      "learning_rate": 9.414737964294636e-06,
+      "loss": 0.887,
+      "step": 35
+    },
+    {
+      "epoch": 2.304,
+      "grad_norm": 0.15727847814559937,
+      "learning_rate": 9.381533400219319e-06,
+      "loss": 0.8712,
+      "step": 36
+    },
+    {
+      "epoch": 2.368,
+      "grad_norm": 0.15251287817955017,
+      "learning_rate": 9.347474647526095e-06,
+      "loss": 0.8583,
+      "step": 37
+    },
+    {
+      "epoch": 2.432,
+      "grad_norm": 0.1633308231830597,
+      "learning_rate": 9.312568346036288e-06,
+      "loss": 0.9924,
+      "step": 38
+    },
+    {
+      "epoch": 2.496,
+      "grad_norm": 0.18498282134532928,
+      "learning_rate": 9.276821300802535e-06,
+      "loss": 0.9926,
+      "step": 39
+    },
+    {
+      "epoch": 2.56,
+      "grad_norm": 0.16592994332313538,
+      "learning_rate": 9.24024048078213e-06,
+      "loss": 0.841,
+      "step": 40
+    },
+    {
+      "epoch": 2.624,
+      "grad_norm": 0.18246057629585266,
+      "learning_rate": 9.202833017478421e-06,
+      "loss": 0.8179,
+      "step": 41
+    },
+    {
+      "epoch": 2.6879999999999997,
+      "grad_norm": 0.1779918223619461,
+      "learning_rate": 9.164606203550498e-06,
+      "loss": 0.9315,
+      "step": 42
+    },
+    {
+      "epoch": 2.752,
+      "grad_norm": 0.15593133866786957,
+      "learning_rate": 9.125567491391476e-06,
+      "loss": 0.804,
+      "step": 43
+    },
+    {
+      "epoch": 2.816,
+      "grad_norm": 0.1553468406200409,
+      "learning_rate": 9.085724491675642e-06,
+      "loss": 0.9364,
+      "step": 44
+    },
+    {
+      "epoch": 2.88,
+      "grad_norm": 0.17488640546798706,
+      "learning_rate": 9.045084971874738e-06,
+      "loss": 0.876,
+      "step": 45
+    },
+    {
+      "epoch": 2.944,
+      "grad_norm": 0.16460111737251282,
+      "learning_rate": 9.003656854743667e-06,
+      "loss": 0.9124,
+      "step": 46
+    },
+    {
+      "epoch": 3.008,
+      "grad_norm": 0.19362136721611023,
+      "learning_rate": 8.961448216775955e-06,
+      "loss": 1.1932,
+      "step": 47
+    },
+    {
+      "epoch": 3.072,
+      "grad_norm": 0.15131668746471405,
+      "learning_rate": 8.9184672866292e-06,
+      "loss": 0.8872,
+      "step": 48
+    },
+    {
+      "epoch": 3.136,
+      "grad_norm": 0.14356467127799988,
+      "learning_rate": 8.874722443520898e-06,
+      "loss": 0.9082,
+      "step": 49
+    },
+    {
+      "epoch": 3.2,
+      "grad_norm": 0.12759928405284882,
+      "learning_rate": 8.83022221559489e-06,
+      "loss": 0.8845,
+      "step": 50
+    },
+    {
+      "epoch": 3.2640000000000002,
+      "grad_norm": 0.15475597977638245,
+      "learning_rate": 8.784975278258783e-06,
+      "loss": 0.9539,
+      "step": 51
+    },
+    {
+      "epoch": 3.328,
+      "grad_norm": 0.13172124326229095,
+      "learning_rate": 8.73899045249266e-06,
+      "loss": 0.8267,
+      "step": 52
+    },
+    {
+      "epoch": 3.392,
+      "grad_norm": 0.1589364856481552,
+      "learning_rate": 8.692276703129421e-06,
+      "loss": 0.8551,
+      "step": 53
+    },
+    {
+      "epoch": 3.456,
+      "grad_norm": 0.1318974792957306,
+      "learning_rate": 8.644843137107058e-06,
+      "loss": 0.8704,
+      "step": 54
+    },
+    {
+      "epoch": 3.52,
+      "grad_norm": 0.1344061940908432,
+      "learning_rate": 8.596699001693257e-06,
+      "loss": 0.8843,
+      "step": 55
+    },
+    {
+      "epoch": 3.584,
+      "grad_norm": 0.14365153014659882,
+      "learning_rate": 8.547853682682605e-06,
+      "loss": 0.8233,
+      "step": 56
+    },
+    {
+      "epoch": 3.648,
+      "grad_norm": 0.12842649221420288,
+      "learning_rate": 8.498316702566828e-06,
+      "loss": 0.8805,
+      "step": 57
+    },
+    {
+      "epoch": 3.7119999999999997,
+      "grad_norm": 0.12912048399448395,
+      "learning_rate": 8.44809771867835e-06,
+      "loss": 0.8576,
+      "step": 58
+    },
+    {
+      "epoch": 3.776,
+      "grad_norm": 0.12904053926467896,
+      "learning_rate": 8.397206521307584e-06,
+      "loss": 0.8536,
+      "step": 59
+    },
+    {
+      "epoch": 3.84,
+      "grad_norm": 0.14415085315704346,
+      "learning_rate": 8.345653031794292e-06,
+      "loss": 0.9647,
+      "step": 60
+    },
+    {
+      "epoch": 3.904,
+      "grad_norm": 0.12525947391986847,
+      "learning_rate": 8.293447300593402e-06,
+      "loss": 0.8804,
+      "step": 61
+    },
+    {
+      "epoch": 3.968,
+      "grad_norm": 0.13248352706432343,
+      "learning_rate": 8.240599505315656e-06,
+      "loss": 0.9382,
+      "step": 62
+    },
+    {
+      "epoch": 4.032,
+      "grad_norm": 0.17219072580337524,
+      "learning_rate": 8.18711994874345e-06,
+      "loss": 1.1208,
+      "step": 63
+    },
+    {
+      "epoch": 4.096,
+      "grad_norm": 0.12371776252985,
+      "learning_rate": 8.133019056822303e-06,
+      "loss": 0.8632,
+      "step": 64
+    },
+    {
+      "epoch": 4.16,
+      "grad_norm": 0.12638618052005768,
+      "learning_rate": 8.078307376628292e-06,
+      "loss": 0.9224,
+      "step": 65
+    },
+    {
+      "epoch": 4.224,
+      "grad_norm": 0.12182217836380005,
+      "learning_rate": 8.022995574311876e-06,
+      "loss": 0.8605,
+      "step": 66
+    },
+    {
+      "epoch": 4.288,
+      "grad_norm": 0.13366594910621643,
+      "learning_rate": 7.967094433018508e-06,
+      "loss": 0.8479,
+      "step": 67
+    },
+    {
+      "epoch": 4.352,
+      "grad_norm": 0.13507261872291565,
+      "learning_rate": 7.910614850786448e-06,
+      "loss": 0.8565,
+      "step": 68
+    },
+    {
+      "epoch": 4.416,
+      "grad_norm": 0.1269528865814209,
+      "learning_rate": 7.85356783842216e-06,
+      "loss": 0.8956,
+      "step": 69
+    },
+    {
+      "epoch": 4.48,
+      "grad_norm": 0.11027321964502335,
+      "learning_rate": 7.795964517353734e-06,
+      "loss": 0.7397,
+      "step": 70
+    },
+    {
+      "epoch": 4.5440000000000005,
+      "grad_norm": 0.1362013965845108,
+      "learning_rate": 7.737816117462752e-06,
+      "loss": 0.922,
+      "step": 71
+    },
+    {
+      "epoch": 4.608,
+      "grad_norm": 0.12494367361068726,
+      "learning_rate": 7.679133974894984e-06,
+      "loss": 0.7995,
+      "step": 72
+    },
+    {
+      "epoch": 4.672,
+      "grad_norm": 0.12939389050006866,
+      "learning_rate": 7.619929529850397e-06,
+      "loss": 0.8423,
+      "step": 73
+    },
+    {
+      "epoch": 4.736,
+      "grad_norm": 0.1359853446483612,
+      "learning_rate": 7.560214324352858e-06,
+      "loss": 0.9657,
+      "step": 74
+    },
+    {
+      "epoch": 4.8,
+      "grad_norm": 0.12302935123443604,
+      "learning_rate": 7.500000000000001e-06,
+      "loss": 0.8209,
+      "step": 75
+    },
+    {
+      "epoch": 4.864,
+      "grad_norm": 0.11996591091156006,
+      "learning_rate": 7.4392982956936644e-06,
+      "loss": 0.7744,
+      "step": 76
+    },
+    {
+      "epoch": 4.928,
+      "grad_norm": 0.13304197788238525,
+      "learning_rate": 7.378121045351378e-06,
+      "loss": 0.9597,
+      "step": 77
+    },
+    {
+      "epoch": 4.992,
+      "grad_norm": 0.1741103231906891,
+      "learning_rate": 7.31648017559931e-06,
+      "loss": 1.1339,
+      "step": 78
+    },
+    {
+      "epoch": 5.056,
+      "grad_norm": 0.1462891697883606,
+      "learning_rate": 7.254387703447154e-06,
+      "loss": 0.9125,
+      "step": 79
+    },
+    {
+      "epoch": 5.12,
+      "grad_norm": 0.11839975416660309,
+      "learning_rate": 7.191855733945388e-06,
+      "loss": 0.9398,
+      "step": 80
+    },
+    {
+      "epoch": 5.184,
+      "grad_norm": 0.1265653818845749,
+      "learning_rate": 7.128896457825364e-06,
+      "loss": 0.8048,
+      "step": 81
+    },
+    {
+      "epoch": 5.248,
+      "grad_norm": 0.12180934846401215,
+      "learning_rate": 7.06552214912271e-06,
+      "loss": 0.7876,
+      "step": 82
+    },
+    {
+      "epoch": 5.312,
+      "grad_norm": 0.13135825097560883,
+      "learning_rate": 7.0017451627844765e-06,
+      "loss": 0.8986,
+      "step": 83
+    },
+    {
+      "epoch": 5.376,
+      "grad_norm": 0.12408099323511124,
+      "learning_rate": 6.9375779322605154e-06,
+      "loss": 0.8158,
+      "step": 84
+    },
+    {
+      "epoch": 5.44,
+      "grad_norm": 0.12082885205745697,
+      "learning_rate": 6.873032967079562e-06,
+      "loss": 0.9035,
+      "step": 85
+    },
+    {
+      "epoch": 5.504,
+      "grad_norm": 0.12515701353549957,
+      "learning_rate": 6.808122850410461e-06,
+      "loss": 0.8674,
+      "step": 86
+    },
+    {
+      "epoch": 5.568,
+      "grad_norm": 0.12426020950078964,
+      "learning_rate": 6.7428602366090764e-06,
+      "loss": 0.7893,
+      "step": 87
+    },
+    {
+      "epoch": 5.632,
+      "grad_norm": 0.12863430380821228,
+      "learning_rate": 6.677257848751276e-06,
+      "loss": 0.8792,
+      "step": 88
+    },
+    {
+      "epoch": 5.696,
+      "grad_norm": 0.1337299644947052,
+      "learning_rate": 6.611328476152557e-06,
+      "loss": 0.8779,
+      "step": 89
+    },
+    {
+      "epoch": 5.76,
+      "grad_norm": 0.13116392493247986,
+      "learning_rate": 6.545084971874738e-06,
+      "loss": 0.8662,
+      "step": 90
+    },
+    {
+      "epoch": 5.824,
+      "grad_norm": 0.12727566063404083,
+      "learning_rate": 6.4785402502202345e-06,
+      "loss": 0.8901,
+      "step": 91
+    },
+    {
+      "epoch": 5.888,
+      "grad_norm": 0.12160185724496841,
+      "learning_rate": 6.411707284214384e-06,
+      "loss": 0.8608,
+      "step": 92
+    },
+    {
+      "epoch": 5.952,
+      "grad_norm": 0.1284012645483017,
+      "learning_rate": 6.344599103076329e-06,
+      "loss": 0.8304,
+      "step": 93
+    },
+    {
+      "epoch": 6.016,
+      "grad_norm": 0.19524367153644562,
+      "learning_rate": 6.277228789678953e-06,
+      "loss": 1.2065,
+      "step": 94
+    },
+    {
+      "epoch": 6.08,
+      "grad_norm": 0.11789774894714355,
+      "learning_rate": 6.209609477998339e-06,
+      "loss": 0.77,
+      "step": 95
+    },
+    {
+      "epoch": 6.144,
+      "grad_norm": 0.12179970741271973,
+      "learning_rate": 6.141754350553279e-06,
+      "loss": 0.9022,
+      "step": 96
+    },
+    {
+      "epoch": 6.208,
+      "grad_norm": 0.12981604039669037,
+      "learning_rate": 6.073676635835317e-06,
+      "loss": 0.8849,
+      "step": 97
+    },
+    {
+      "epoch": 6.272,
+      "grad_norm": 0.13244786858558655,
+      "learning_rate": 6.005389605729824e-06,
+      "loss": 0.789,
+      "step": 98
+    },
+    {
+      "epoch": 6.336,
+      "grad_norm": 0.12296921759843826,
+      "learning_rate": 5.936906572928625e-06,
+      "loss": 0.8968,
+      "step": 99
+    },
+    {
+      "epoch": 6.4,
+      "grad_norm": 0.1255040466785431,
+      "learning_rate": 5.8682408883346535e-06,
+      "loss": 0.8117,
+      "step": 100
+    },
+    {
+      "epoch": 6.464,
+      "grad_norm": 0.13155046105384827,
+      "learning_rate": 5.799405938459175e-06,
+      "loss": 0.92,
+      "step": 101
+    },
+    {
+      "epoch": 6.5280000000000005,
+      "grad_norm": 0.12069229036569595,
+      "learning_rate": 5.730415142812059e-06,
+      "loss": 0.8371,
+      "step": 102
+    },
+    {
+      "epoch": 6.592,
+      "grad_norm": 0.12130100280046463,
+      "learning_rate": 5.661281951285613e-06,
+      "loss": 0.8401,
+      "step": 103
+    },
+    {
+      "epoch": 6.656,
+      "grad_norm": 0.12238356471061707,
+      "learning_rate": 5.592019841532507e-06,
+      "loss": 0.838,
+      "step": 104
+    },
+    {
+      "epoch": 6.72,
+      "grad_norm": 0.13303394615650177,
+      "learning_rate": 5.522642316338268e-06,
+      "loss": 0.9041,
+      "step": 105
+    },
+    {
+      "epoch": 6.784,
+      "grad_norm": 0.1444772183895111,
+      "learning_rate": 5.453162900988902e-06,
+      "loss": 0.9033,
+      "step": 106
+    },
+    {
+      "epoch": 6.848,
+      "grad_norm": 0.12250606715679169,
+      "learning_rate": 5.383595140634093e-06,
+      "loss": 0.7294,
+      "step": 107
+    },
+    {
+      "epoch": 6.912,
+      "grad_norm": 0.12612612545490265,
+      "learning_rate": 5.3139525976465675e-06,
+      "loss": 0.8332,
+      "step": 108
+    },
+    {
+      "epoch": 6.976,
+      "grad_norm": 0.14953787624835968,
+      "learning_rate": 5.244248848978067e-06,
+      "loss": 0.9709,
+      "step": 109
+    },
+    {
+      "epoch": 7.04,
+      "grad_norm": 0.14709977805614471,
+      "learning_rate": 5.174497483512506e-06,
+      "loss": 0.9743,
+      "step": 110
+    },
+    {
+      "epoch": 7.104,
+      "grad_norm": 0.12281537801027298,
+      "learning_rate": 5.1047120994167855e-06,
+      "loss": 0.8208,
+      "step": 111
+    },
+    {
+      "epoch": 7.168,
+      "grad_norm": 0.13773250579833984,
+      "learning_rate": 5.034906301489808e-06,
+      "loss": 0.8659,
+      "step": 112
+    },
+    {
+      "epoch": 7.232,
+      "grad_norm": 0.12388751655817032,
+      "learning_rate": 4.965093698510192e-06,
+      "loss": 0.8874,
+      "step": 113
+    },
+    {
+      "epoch": 7.296,
+      "grad_norm": 0.1328098624944687,
+      "learning_rate": 4.895287900583216e-06,
+      "loss": 0.8538,
+      "step": 114
+    },
+    {
+      "epoch": 7.36,
+      "grad_norm": 0.1243242472410202,
+      "learning_rate": 4.825502516487497e-06,
+      "loss": 0.7639,
+      "step": 115
+    },
+    {
+      "epoch": 7.424,
+      "grad_norm": 0.12820030748844147,
+      "learning_rate": 4.755751151021934e-06,
+      "loss": 0.891,
+      "step": 116
+    },
+    {
+      "epoch": 7.4879999999999995,
+      "grad_norm": 0.12143447995185852,
+      "learning_rate": 4.686047402353433e-06,
+      "loss": 0.8208,
+      "step": 117
+    },
+    {
+      "epoch": 7.552,
+      "grad_norm": 0.12033109366893768,
+      "learning_rate": 4.6164048593659076e-06,
+      "loss": 0.7866,
+      "step": 118
+    },
+    {
+      "epoch": 7.616,
+      "grad_norm": 0.11808149516582489,
+      "learning_rate": 4.546837099011101e-06,
+      "loss": 0.8194,
+      "step": 119
+    },
+    {
+      "epoch": 7.68,
+      "grad_norm": 0.13013483583927155,
+      "learning_rate": 4.477357683661734e-06,
+      "loss": 0.9189,
+      "step": 120
+    },
+    {
+      "epoch": 7.744,
+      "grad_norm": 0.12187657505273819,
+      "learning_rate": 4.4079801584674955e-06,
+      "loss": 0.8126,
+      "step": 121
+    },
+    {
+      "epoch": 7.808,
+      "grad_norm": 0.12875136733055115,
+      "learning_rate": 4.3387180487143875e-06,
+      "loss": 0.92,
+      "step": 122
+    },
+    {
+      "epoch": 7.872,
+      "grad_norm": 0.12279701232910156,
+      "learning_rate": 4.269584857187942e-06,
+      "loss": 0.7819,
+      "step": 123
+    },
+    {
+      "epoch": 7.936,
+      "grad_norm": 0.12862621247768402,
+      "learning_rate": 4.200594061540827e-06,
+      "loss": 0.8319,
+      "step": 124
+    },
+    {
+      "epoch": 8.0,
+      "grad_norm": 0.19300128519535065,
+      "learning_rate": 4.131759111665349e-06,
+      "loss": 1.1825,
+      "step": 125
+    },
+    {
+      "epoch": 8.064,
+      "grad_norm": 0.11853662878274918,
+      "learning_rate": 4.063093427071376e-06,
+      "loss": 0.8276,
+      "step": 126
+    },
+    {
+      "epoch": 8.128,
+      "grad_norm": 0.12296923995018005,
+      "learning_rate": 3.994610394270178e-06,
+      "loss": 0.8654,
+      "step": 127
+    },
+    {
+      "epoch": 8.192,
+      "grad_norm": 0.12605859339237213,
+      "learning_rate": 3.926323364164684e-06,
+      "loss": 0.8355,
+      "step": 128
+    },
+    {
+      "epoch": 8.256,
+      "grad_norm": 0.11515746265649796,
+      "learning_rate": 3.8582456494467214e-06,
+      "loss": 0.8135,
+      "step": 129
+    },
+    {
+      "epoch": 8.32,
+      "grad_norm": 0.12183663994073868,
+      "learning_rate": 3.790390522001662e-06,
+      "loss": 0.8603,
+      "step": 130
+    },
+    {
+      "epoch": 8.384,
+      "grad_norm": 0.12571197748184204,
+      "learning_rate": 3.7227712103210485e-06,
+      "loss": 0.8231,
+      "step": 131
+    },
+    {
+      "epoch": 8.448,
+      "grad_norm": 0.12782686948776245,
+      "learning_rate": 3.655400896923672e-06,
+      "loss": 0.8856,
+      "step": 132
+    },
+    {
+      "epoch": 8.512,
+      "grad_norm": 0.12933164834976196,
+      "learning_rate": 3.5882927157856175e-06,
+      "loss": 0.8418,
+      "step": 133
+    },
+    {
+      "epoch": 8.576,
+      "grad_norm": 0.1267186850309372,
+      "learning_rate": 3.521459749779769e-06,
+      "loss": 0.7858,
+      "step": 134
+    },
+    {
+      "epoch": 8.64,
+      "grad_norm": 0.13652652502059937,
+      "learning_rate": 3.4549150281252635e-06,
+      "loss": 0.7915,
+      "step": 135
+    },
+    {
+      "epoch": 8.704,
+      "grad_norm": 0.13402396440505981,
+      "learning_rate": 3.3886715238474454e-06,
+      "loss": 0.8035,
+      "step": 136
+    },
+    {
+      "epoch": 8.768,
+      "grad_norm": 0.12407074123620987,
+      "learning_rate": 3.322742151248726e-06,
+      "loss": 0.8221,
+      "step": 137
+    },
+    {
+      "epoch": 8.832,
+      "grad_norm": 0.12156077474355698,
+      "learning_rate": 3.2571397633909252e-06,
+      "loss": 0.8522,
+      "step": 138
+    },
+    {
+      "epoch": 8.896,
+      "grad_norm": 0.12711839377880096,
+      "learning_rate": 3.1918771495895395e-06,
+      "loss": 0.8802,
+      "step": 139
+    },
+    {
+      "epoch": 8.96,
+      "grad_norm": 0.1299351155757904,
+      "learning_rate": 3.12696703292044e-06,
+      "loss": 0.8643,
+      "step": 140
+    },
+    {
+      "epoch": 9.024,
+      "grad_norm": 0.17408832907676697,
+      "learning_rate": 3.0624220677394854e-06,
+      "loss": 1.0836,
+      "step": 141
+    },
+    {
+      "epoch": 9.088,
+      "grad_norm": 0.1219247579574585,
+      "learning_rate": 2.9982548372155264e-06,
+      "loss": 0.9013,
+      "step": 142
+    },
+    {
+      "epoch": 9.152,
+      "grad_norm": 0.12802451848983765,
+      "learning_rate": 2.934477850877292e-06,
+      "loss": 0.8827,
+      "step": 143
+    },
+    {
+      "epoch": 9.216,
+      "grad_norm": 0.11756030470132828,
+      "learning_rate": 2.871103542174637e-06,
+      "loss": 0.7977,
+      "step": 144
+    },
+    {
+      "epoch": 9.28,
+      "grad_norm": 0.1230553537607193,
+      "learning_rate": 2.8081442660546126e-06,
+      "loss": 0.8603,
+      "step": 145
+    },
+    {
+      "epoch": 9.344,
+      "grad_norm": 0.12843072414398193,
+      "learning_rate": 2.7456122965528475e-06,
+      "loss": 0.7889,
+      "step": 146
+    },
+    {
+      "epoch": 9.408,
+      "grad_norm": 0.13244116306304932,
+      "learning_rate": 2.683519824400693e-06,
+      "loss": 0.8419,
+      "step": 147
+    },
+    {
+      "epoch": 9.472,
+      "grad_norm": 0.1266927719116211,
+      "learning_rate": 2.6218789546486235e-06,
+      "loss": 0.8097,
+      "step": 148
+    },
+    {
+      "epoch": 9.536,
+      "grad_norm": 0.11919289082288742,
+      "learning_rate": 2.560701704306336e-06,
+      "loss": 0.7759,
+      "step": 149
+    },
+    {
+      "epoch": 9.6,
+      "grad_norm": 0.12574751675128937,
+      "learning_rate": 2.5000000000000015e-06,
+      "loss": 0.8389,
+      "step": 150
+    },
+    {
+      "epoch": 9.664,
+      "grad_norm": 0.14269892871379852,
+      "learning_rate": 2.4397856756471435e-06,
+      "loss": 0.8767,
+      "step": 151
+    },
+    {
+      "epoch": 9.728,
+      "grad_norm": 0.12222199141979218,
+      "learning_rate": 2.380070470149605e-06,
+      "loss": 0.7306,
+      "step": 152
+    },
+    {
+      "epoch": 9.792,
+      "grad_norm": 0.11675389856100082,
+      "learning_rate": 2.320866025105016e-06,
+      "loss": 0.739,
+      "step": 153
+    },
+    {
+      "epoch": 9.856,
+      "grad_norm": 0.13254772126674652,
+      "learning_rate": 2.2621838825372496e-06,
+      "loss": 0.9458,
+      "step": 154
+    },
+    {
+      "epoch": 9.92,
+      "grad_norm": 0.11959590762853622,
+      "learning_rate": 2.204035482646267e-06,
+      "loss": 0.8222,
+      "step": 155
+    },
+    {
+      "epoch": 9.984,
+      "grad_norm": 0.15119311213493347,
+      "learning_rate": 2.146432161577842e-06,
+      "loss": 1.0054,
+      "step": 156
+    },
+    {
+      "epoch": 10.048,
+      "grad_norm": 0.14419986307621002,
+      "learning_rate": 2.0893851492135536e-06,
+      "loss": 0.9804,
+      "step": 157
+    },
+    {
+      "epoch": 10.112,
+      "grad_norm": 0.11810199171304703,
+      "learning_rate": 2.0329055669814936e-06,
+      "loss": 0.7982,
+      "step": 158
+    },
+    {
+      "epoch": 10.176,
+      "grad_norm": 0.1280771642923355,
+      "learning_rate": 1.977004425688126e-06,
+      "loss": 0.808,
+      "step": 159
+    },
+    {
+      "epoch": 10.24,
+      "grad_norm": 0.12855829298496246,
+      "learning_rate": 1.9216926233717087e-06,
+      "loss": 0.8768,
+      "step": 160
+    },
+    {
+      "epoch": 10.304,
+      "grad_norm": 0.12173178791999817,
+      "learning_rate": 1.8669809431776991e-06,
+      "loss": 0.8733,
+      "step": 161
+    },
+    {
+      "epoch": 10.368,
+      "grad_norm": 0.11684292554855347,
+      "learning_rate": 1.8128800512565514e-06,
+      "loss": 0.7227,
+      "step": 162
+    },
+    {
+      "epoch": 10.432,
+      "grad_norm": 0.1230064183473587,
+      "learning_rate": 1.7594004946843458e-06,
+      "loss": 0.8293,
+      "step": 163
+    },
+    {
+      "epoch": 10.496,
+      "grad_norm": 0.11779855191707611,
+      "learning_rate": 1.7065526994065973e-06,
+      "loss": 0.7961,
+      "step": 164
+    },
+    {
+      "epoch": 10.56,
+      "grad_norm": 0.12627413868904114,
+      "learning_rate": 1.6543469682057105e-06,
+      "loss": 0.8582,
+      "step": 165
+    },
+    {
+      "epoch": 10.624,
+      "grad_norm": 0.12739939987659454,
+      "learning_rate": 1.6027934786924187e-06,
+      "loss": 0.8743,
+      "step": 166
+    },
+    {
+      "epoch": 10.688,
+      "grad_norm": 0.12277089059352875,
+      "learning_rate": 1.551902281321651e-06,
+      "loss": 0.8122,
+      "step": 167
+    },
+    {
+      "epoch": 10.752,
+      "grad_norm": 0.12769779562950134,
+      "learning_rate": 1.5016832974331725e-06,
+      "loss": 0.8025,
+      "step": 168
+    },
+    {
+      "epoch": 10.816,
+      "grad_norm": 0.12428705394268036,
+      "learning_rate": 1.4521463173173966e-06,
+      "loss": 0.8071,
+      "step": 169
+    },
+    {
+      "epoch": 10.88,
+      "grad_norm": 0.12309806048870087,
+      "learning_rate": 1.4033009983067454e-06,
+      "loss": 0.8643,
+      "step": 170
+    },
+    {
+      "epoch": 10.943999999999999,
+      "grad_norm": 0.1236419677734375,
+      "learning_rate": 1.3551568628929434e-06,
+      "loss": 0.8408,
+      "step": 171
+    },
+    {
+      "epoch": 11.008,
+      "grad_norm": 0.17760780453681946,
+      "learning_rate": 1.3077232968705805e-06,
+      "loss": 1.157,
+      "step": 172
+    },
+    {
+      "epoch": 11.072,
+      "grad_norm": 0.11716406047344208,
+      "learning_rate": 1.2610095475073415e-06,
+      "loss": 0.7641,
+      "step": 173
+    },
+    {
+      "epoch": 11.136,
+      "grad_norm": 0.12319278717041016,
+      "learning_rate": 1.2150247217412186e-06,
+      "loss": 0.8901,
+      "step": 174
+    },
+    {
+      "epoch": 11.2,
+      "grad_norm": 0.1214849203824997,
+      "learning_rate": 1.1697777844051105e-06,
+      "loss": 0.8119,
+      "step": 175
+    },
+    {
+      "epoch": 11.264,
+      "grad_norm": 0.12246395647525787,
+      "learning_rate": 1.1252775564791023e-06,
+      "loss": 0.8347,
+      "step": 176
+    },
+    {
+      "epoch": 11.328,
+      "grad_norm": 0.1226000189781189,
+      "learning_rate": 1.0815327133708015e-06,
+      "loss": 0.8874,
+      "step": 177
+    },
+    {
+      "epoch": 11.392,
+      "grad_norm": 0.12461017072200775,
+      "learning_rate": 1.0385517832240472e-06,
+      "loss": 0.7858,
+      "step": 178
+    },
+    {
+      "epoch": 11.456,
+      "grad_norm": 0.1249910444021225,
+      "learning_rate": 9.963431452563331e-07,
+      "loss": 0.8025,
+      "step": 179
+    },
+    {
+      "epoch": 11.52,
+      "grad_norm": 0.11750364303588867,
+      "learning_rate": 9.549150281252633e-07,
+      "loss": 0.8357,
+      "step": 180
+    },
+    {
+      "epoch": 11.584,
+      "grad_norm": 0.12066759914159775,
+      "learning_rate": 9.142755083243577e-07,
+      "loss": 0.8489,
+      "step": 181
+    },
+    {
+      "epoch": 11.648,
+      "grad_norm": 0.11544518917798996,
+      "learning_rate": 8.744325086085248e-07,
+      "loss": 0.788,
+      "step": 182
+    },
+    {
+      "epoch": 11.712,
+      "grad_norm": 0.12407847493886948,
+      "learning_rate": 8.353937964495029e-07,
+      "loss": 0.8044,
+      "step": 183
+    },
+    {
+      "epoch": 11.776,
+      "grad_norm": 0.1282244175672531,
+      "learning_rate": 7.971669825215789e-07,
+      "loss": 0.7997,
+      "step": 184
+    },
+    {
+      "epoch": 11.84,
+      "grad_norm": 0.1248067319393158,
+      "learning_rate": 7.597595192178702e-07,
+      "loss": 0.835,
+      "step": 185
+    },
+    {
+      "epoch": 11.904,
+      "grad_norm": 0.120790995657444,
+      "learning_rate": 7.23178699197467e-07,
+      "loss": 0.8166,
+      "step": 186
+    },
+    {
+      "epoch": 11.968,
+      "grad_norm": 0.13572026789188385,
+      "learning_rate": 6.874316539637127e-07,
+      "loss": 0.8752,
+      "step": 187
+    },
+    {
+      "epoch": 12.032,
+      "grad_norm": 0.1664300560951233,
+      "learning_rate": 6.52525352473905e-07,
+      "loss": 1.0303,
+      "step": 188
+    },
+    {
+      "epoch": 12.096,
+      "grad_norm": 0.1228347048163414,
+      "learning_rate": 6.184665997806832e-07,
+      "loss": 0.8162,
+      "step": 189
+    },
+    {
+      "epoch": 12.16,
+      "grad_norm": 0.11765308678150177,
+      "learning_rate": 5.852620357053651e-07,
+      "loss": 0.7837,
+      "step": 190
+    },
+    {
+      "epoch": 12.224,
+      "grad_norm": 0.12184052914381027,
+      "learning_rate": 5.529181335435124e-07,
+      "loss": 0.7915,
+      "step": 191
+    },
+    {
+      "epoch": 12.288,
+      "grad_norm": 0.12436401098966599,
+      "learning_rate": 5.214411988029355e-07,
+      "loss": 0.8588,
+      "step": 192
+    },
+    {
+      "epoch": 12.352,
+      "grad_norm": 0.12295132875442505,
+      "learning_rate": 4.908373679744316e-07,
+      "loss": 0.8373,
+      "step": 193
+    },
+    {
+      "epoch": 12.416,
+      "grad_norm": 0.12551254034042358,
+      "learning_rate": 4.6111260733545714e-07,
+      "loss": 0.8477,
+      "step": 194
+    },
+    {
+      "epoch": 12.48,
+      "grad_norm": 0.12384471297264099,
+      "learning_rate": 4.322727117869951e-07,
+      "loss": 0.819,
+      "step": 195
+    },
+    {
+      "epoch": 12.544,
+      "grad_norm": 0.12367549538612366,
+      "learning_rate": 4.043233037238281e-07,
+      "loss": 0.7978,
+      "step": 196
+    },
+    {
+      "epoch": 12.608,
+      "grad_norm": 0.12734423577785492,
+      "learning_rate": 3.772698319384349e-07,
+      "loss": 0.9013,
+      "step": 197
+    },
+    {
+      "epoch": 12.672,
+      "grad_norm": 0.11762301623821259,
+      "learning_rate": 3.511175705587433e-07,
+      "loss": 0.7289,
+      "step": 198
+    },
+    {
+      "epoch": 12.736,
+      "grad_norm": 0.12712667882442474,
+      "learning_rate": 3.258716180199278e-07,
+      "loss": 0.9106,
+      "step": 199
+    },
+    {
+      "epoch": 12.8,
+      "grad_norm": 0.11422532051801682,
+      "learning_rate": 3.015368960704584e-07,
+      "loss": 0.7579,
+      "step": 200
+    },
+    {
+      "epoch": 12.864,
+      "grad_norm": 0.1257271021604538,
+      "learning_rate": 2.7811814881259503e-07,
+      "loss": 0.8527,
+      "step": 201
+    },
+    {
+      "epoch": 12.928,
+      "grad_norm": 0.1224028617143631,
+      "learning_rate": 2.556199417775174e-07,
+      "loss": 0.8717,
+      "step": 202
+    },
+    {
+      "epoch": 12.992,
+      "grad_norm": 0.1601700484752655,
+      "learning_rate": 2.3404666103526542e-07,
+      "loss": 0.9906,
+      "step": 203
+    },
+    {
+      "epoch": 13.056,
+      "grad_norm": 0.14647288620471954,
+      "learning_rate": 2.134025123396638e-07,
+      "loss": 0.9804,
+      "step": 204
+    },
+    {
+      "epoch": 13.12,
+      "grad_norm": 0.11194372177124023,
+      "learning_rate": 1.9369152030840553e-07,
+      "loss": 0.7664,
+      "step": 205
+    },
+    {
+      "epoch": 13.184,
+      "grad_norm": 0.13287296891212463,
+      "learning_rate": 1.7491752763844294e-07,
+      "loss": 0.8259,
+      "step": 206
+    },
+    {
+      "epoch": 13.248,
+      "grad_norm": 0.1248527243733406,
+      "learning_rate": 1.5708419435684463e-07,
+      "loss": 0.8892,
+      "step": 207
+    },
+    {
+      "epoch": 13.312,
+      "grad_norm": 0.1182282567024231,
+      "learning_rate": 1.4019499710726913e-07,
+      "loss": 0.7666,
+      "step": 208
+    },
+    {
+      "epoch": 13.376,
+      "grad_norm": 0.11040018498897552,
+      "learning_rate": 1.2425322847218368e-07,
+      "loss": 0.7489,
+      "step": 209
+    },
+    {
+      "epoch": 13.44,
+      "grad_norm": 0.11782634258270264,
+      "learning_rate": 1.0926199633097156e-07,
+      "loss": 0.8304,
+      "step": 210
+    },
+    {
+      "epoch": 13.504,
+      "grad_norm": 0.12430648505687714,
+      "learning_rate": 9.522422325404234e-08,
+      "loss": 0.941,
+      "step": 211
+    },
+    {
+      "epoch": 13.568,
+      "grad_norm": 0.11995483934879303,
+      "learning_rate": 8.214264593307097e-08,
+      "loss": 0.8013,
+      "step": 212
+    },
+    {
+      "epoch": 13.632,
+      "grad_norm": 0.12744170427322388,
+      "learning_rate": 7.001981464747565e-08,
+      "loss": 0.7954,
+      "step": 213
+    },
+    {
+      "epoch": 13.696,
+      "grad_norm": 0.12261604517698288,
+      "learning_rate": 5.8858092767236084e-08,
+      "loss": 0.8749,
+      "step": 214
+    },
+    {
+      "epoch": 13.76,
+      "grad_norm": 0.1196310818195343,
+      "learning_rate": 4.865965629214819e-08,
+      "loss": 0.7981,
+      "step": 215
+    },
+    {
+      "epoch": 13.824,
+      "grad_norm": 0.12484210729598999,
+      "learning_rate": 3.9426493427611177e-08,
+      "loss": 0.7978,
+      "step": 216
+    },
+    {
+      "epoch": 13.888,
+      "grad_norm": 0.12131179869174957,
+      "learning_rate": 3.1160404197018155e-08,
+      "loss": 0.7463,
+      "step": 217
+    },
+    {
+      "epoch": 13.952,
+      "grad_norm": 0.1267390251159668,
+      "learning_rate": 2.386300009084408e-08,
+      "loss": 0.8298,
+      "step": 218
+    },
+    {
+      "epoch": 14.016,
+      "grad_norm": 0.16648182272911072,
+      "learning_rate": 1.753570375247815e-08,
+      "loss": 1.154,
+      "step": 219
+    },
+    {
+      "epoch": 14.08,
+      "grad_norm": 0.11439771205186844,
+      "learning_rate": 1.2179748700879013e-08,
+      "loss": 0.7684,
+      "step": 220
+    },
+    {
+      "epoch": 14.144,
+      "grad_norm": 0.11785776168107986,
+      "learning_rate": 7.796179090094891e-09,
+      "loss": 0.7948,
+      "step": 221
+    },
+    {
+      "epoch": 14.208,
+      "grad_norm": 0.11341831833124161,
+      "learning_rate": 4.385849505708084e-09,
+      "loss": 0.7567,
+      "step": 222
+    },
+    {
+      "epoch": 14.272,
+      "grad_norm": 0.12832769751548767,
+      "learning_rate": 1.9494247982282386e-09,
+      "loss": 1.0097,
+      "step": 223
+    },
+    {
+      "epoch": 14.336,
+      "grad_norm": 0.11620064079761505,
+      "learning_rate": 4.87379953478806e-10,
+      "loss": 0.7495,
+      "step": 224
+    },
+    {
+      "epoch": 14.4,
+      "grad_norm": 0.12200789898633957,
+      "learning_rate": 0.0,
+      "loss": 0.8347,
+      "step": 225
+    }
+  ],
+  "logging_steps": 1,
+  "max_steps": 225,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 15,
+  "save_steps": 75,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": true
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 125627307065344.0,
+  "train_batch_size": 1,
+  "trial_name": null,
+  "trial_params": null
+}

s1k-1.1/full/checkpoint-225/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9a9f455dc7902de14b90e50b602aa3f85b31052d9b16e340282d8fc3c06c90fb
+size 7608

s1k-1.1/full/checkpoint-225/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)