Upload checkpoint-2000

checkpoint-2000/added_tokens.json

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

checkpoint-2000/config.json

@@ -0,0 +1,84 @@
+{
+  "architectures": [
+    "Qwen2_5_VLForConditionalGenerationWithVGGT"
+  ],
+  "attention_dropout": 0.0,
+  "bos_token_id": 151643,
+  "eos_token_id": 151645,
+  "feature_fusion_method": "deepstack_language_add",
+  "fusion_num_layers": 1,
+  "geometry_encoder_layers": [
+    11,
+    17,
+    21
+  ],
+  "geometry_encoder_type": "vggt",
+  "geometry_fusion_layers": [
+    0,
+    1,
+    2
+  ],
+  "geometry_merger_type": "mlp",
+  "hidden_act": "silu",
+  "hidden_size": 2048,
+  "image_token_id": 151655,
+  "include_camera_token": false,
+  "initializer_range": 0.02,
+  "intermediate_size": 11008,
+  "max_position_embeddings": 128000,
+  "max_window_layers": 70,
+  "model_type": "qwen2_5_vl",
+  "num_attention_heads": 16,
+  "num_hidden_layers": 36,
+  "num_key_value_heads": 2,
+  "pos_encoding_type": "none",
+  "reference_frame": "first",
+  "rms_norm_eps": 1e-06,
+  "rope_scaling": {
+    "mrope_section": [
+      16,
+      24,
+      24
+    ],
+    "rope_type": "default",
+    "type": "default"
+  },
+  "rope_theta": 1000000.0,
+  "sliding_window": 32768,
+  "tie_word_embeddings": true,
+  "torch_dtype": "bfloat16",
+  "transformers_version": "4.50.0",
+  "use_cache": false,
+  "use_geometry_encoder": true,
+  "use_sliding_window": false,
+  "video_token_id": 151656,
+  "vision_config": {
+    "depth": 32,
+    "fullatt_block_indexes": [
+      7,
+      15,
+      23,
+      31
+    ],
+    "hidden_act": "silu",
+    "hidden_size": 1280,
+    "in_channels": 3,
+    "in_chans": 3,
+    "intermediate_size": 3420,
+    "model_type": "qwen2_5_vl",
+    "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": 2,
+    "torch_dtype": "bfloat16",
+    "window_size": 112
+  },
+  "vision_end_token_id": 151653,
+  "vision_language_fusion_layers": null,
+  "vision_start_token_id": 151652,
+  "vision_token_id": 151654,
+  "vocab_size": 151936
+}

checkpoint-2000/generation_config.json

@@ -0,0 +1,12 @@
+{
+  "bos_token_id": 151643,
+  "do_sample": true,
+  "eos_token_id": [
+    151645,
+    151643
+  ],
+  "pad_token_id": 151643,
+  "repetition_penalty": 1.05,
+  "temperature": 1e-06,
+  "transformers_version": "4.50.0"
+}

checkpoint-2000/global_step2000/bf16_zero_pp_rank_0_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5c234173e1b7b1021f813ad8d9645e50fc8bf356bcadf1bd76db0fd50e6be38a
+size 4817695729

checkpoint-2000/global_step2000/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:b3304202060334f26c68ba9dd1742e53cd312dea416e1ab5d6b95c7e6aa43a4f
+size 4817698545

checkpoint-2000/global_step2000/bf16_zero_pp_rank_2_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:252b55983d8583dc38413ba9e848746cc32ceb3b886321316e7567a19fe15d9f
+size 4817699121

checkpoint-2000/global_step2000/bf16_zero_pp_rank_3_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5c1e518697e2da90f0f519619ed6d635835d80a4ffa287a9fec73ac1a3de67fb
+size 4817699569

checkpoint-2000/global_step2000/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:d32a45e076dacafe7da70027cc350fbb2c2d77e878651ed67bc02b99554e1c7d
+size 4817698993

checkpoint-2000/global_step2000/bf16_zero_pp_rank_5_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7fab3ab4ae9cd875c425ca5e50cc00d4ac0ca6acdd593c668429bb9cf105b400
+size 4817699249

checkpoint-2000/global_step2000/bf16_zero_pp_rank_6_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8ef20cf06f792651c18eed21a686a41c8b098ba6aba48915ef26c2ba0fd1db45
+size 4817699057

checkpoint-2000/global_step2000/bf16_zero_pp_rank_7_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9fb931aba92e7a88e07c39ecf12ae7f8056e2872a265c3b3e2125e548bcae715
+size 4817699377

checkpoint-2000/global_step2000/mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e6000c9215dda6e3e9cb43e9f6cc65e1e1f30797a2f4e7b5111de8fcee3f48d6
+size 12736254037

checkpoint-2000/latest

@@ -0,0 +1 @@
+global_step2000

checkpoint-2000/model-00001-of-00003.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4eda8acf90dd0dad671385f9648f10ba812f3f3525693f2273de18d794759eb8
+size 4962140952

checkpoint-2000/model-00002-of-00003.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3fcf624e817ede52821989d86116e5cb8e91a2f441928e65b3294bda61ddf3eb
+size 4869006472

checkpoint-2000/model-00003-of-00003.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d96a44998e4ddefae712b8571b47ba7bc13984a4a612769ee319f4b00a0d9e7a
+size 622329984

checkpoint-2000/rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:93366b78442d88be019e36f3ea689d21ed498b09e95d530b448530a143cc3f36
+size 16325

checkpoint-2000/rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d721cb58d3bd65db773d5c1f43927f5aad0aeb38d44f1201f9142e93b467b6ed
+size 16325

checkpoint-2000/rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f8cd3dabf2f4ae849586a5b78bf256d3ee47426368adf39ef0cfa60f97e31328
+size 16325

checkpoint-2000/rng_state_3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1a6d3c132b137b6645ecb4941c37da51f062228a8f875e480ad1ab2786df8a46
+size 16325

checkpoint-2000/rng_state_4.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:70602ef413ff4a3a2b3c27c82c937c3edccab869afd9899600831facd1efa37d
+size 16325

checkpoint-2000/rng_state_5.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5da8055d6f4011e5ec500f5e23cc87cbfed3e0e1aac3d8d331d42698da8e6aa6
+size 16325

checkpoint-2000/rng_state_6.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f3eca2914b1b5c55d14aafdc6a7517c617a86a987344214e557c3c5b9d74bc10
+size 16325

checkpoint-2000/rng_state_7.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:26f3977ddd6e391a32d3384758177f0212c6b30f62092335828dc6f5a279396f
+size 16325

checkpoint-2000/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:55c8ab00eff0e3985fe0711a73c280a2872463cb5d58817c6c4c1710d7d62f57
+size 1465

checkpoint-2000/special_tokens_map.json

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

checkpoint-2000/tokenizer_config.json

@@ -0,0 +1,209 @@
+{
+  "add_bos_token": false,
+  "add_prefix_space": false,
+  "added_tokens_decoder": {
+    "151643": {
+      "content": "<|endoftext|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151644": {
+      "content": "<|im_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151645": {
+      "content": "<|im_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151646": {
+      "content": "<|object_ref_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151647": {
+      "content": "<|object_ref_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151648": {
+      "content": "<|box_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151649": {
+      "content": "<|box_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151650": {
+      "content": "<|quad_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151651": {
+      "content": "<|quad_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151652": {
+      "content": "<|vision_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151653": {
+      "content": "<|vision_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151654": {
+      "content": "<|vision_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151655": {
+      "content": "<|image_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151656": {
+      "content": "<|video_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151657": {
+      "content": "<tool_call>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151658": {
+      "content": "</tool_call>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151659": {
+      "content": "<|fim_prefix|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151660": {
+      "content": "<|fim_middle|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151661": {
+      "content": "<|fim_suffix|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151662": {
+      "content": "<|fim_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151663": {
+      "content": "<|repo_name|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151664": {
+      "content": "<|file_sep|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    }
+  },
+  "additional_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "bos_token": null,
+  "chat_template": "{% 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\nYou are a helpful assistant.<|im_end|>\n{% endif %}<|im_start|>{{ message['role'] }}\n{% if message['content'] is string %}{{ message['content'] }}<|im_end|>\n{% 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_start|><|image_pad|><|vision_end|>{% 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_start|><|video_pad|><|vision_end|>{% elif 'text' in content %}{{ content['text'] }}{% endif %}{% endfor %}<|im_end|>\n{% endif %}{% endfor %}{% if add_generation_prompt %}<|im_start|>assistant\n{% endif %}",
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|im_end|>",
+  "errors": "replace",
+  "extra_special_tokens": {},
+  "model_max_length": 12800,
+  "pad_token": "<|endoftext|>",
+  "padding_side": "right",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}

checkpoint-2000/trainer_state.json

@@ -0,0 +1,1434 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.604366548311551,
+  "eval_steps": 500,
+  "global_step": 2000,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0030218327415577548,
+      "grad_norm": 165.52227783203125,
+      "learning_rate": 1.0000000000000002e-06,
+      "loss": 13.7981,
+      "step": 10
+    },
+    {
+      "epoch": 0.0060436654831155096,
+      "grad_norm": 78.14161682128906,
+      "learning_rate": 2.0000000000000003e-06,
+      "loss": 13.0602,
+      "step": 20
+    },
+    {
+      "epoch": 0.009065498224673264,
+      "grad_norm": 51.4779167175293,
+      "learning_rate": 3e-06,
+      "loss": 12.098,
+      "step": 30
+    },
+    {
+      "epoch": 0.012087330966231019,
+      "grad_norm": 48.51755905151367,
+      "learning_rate": 4.000000000000001e-06,
+      "loss": 10.3148,
+      "step": 40
+    },
+    {
+      "epoch": 0.015109163707788774,
+      "grad_norm": 45.85222244262695,
+      "learning_rate": 5e-06,
+      "loss": 9.5523,
+      "step": 50
+    },
+    {
+      "epoch": 0.01813099644934653,
+      "grad_norm": 40.89993667602539,
+      "learning_rate": 6e-06,
+      "loss": 8.3467,
+      "step": 60
+    },
+    {
+      "epoch": 0.021152829190904283,
+      "grad_norm": 31.405731201171875,
+      "learning_rate": 7e-06,
+      "loss": 6.9178,
+      "step": 70
+    },
+    {
+      "epoch": 0.024174661932462038,
+      "grad_norm": 53.975521087646484,
+      "learning_rate": 8.000000000000001e-06,
+      "loss": 6.779,
+      "step": 80
+    },
+    {
+      "epoch": 0.027196494674019793,
+      "grad_norm": 30.481246948242188,
+      "learning_rate": 9e-06,
+      "loss": 7.0495,
+      "step": 90
+    },
+    {
+      "epoch": 0.030218327415577548,
+      "grad_norm": 37.02349090576172,
+      "learning_rate": 1e-05,
+      "loss": 6.7093,
+      "step": 100
+    },
+    {
+      "epoch": 0.0332401601571353,
+      "grad_norm": 37.39162063598633,
+      "learning_rate": 9.999760394462267e-06,
+      "loss": 7.3573,
+      "step": 110
+    },
+    {
+      "epoch": 0.03626199289869306,
+      "grad_norm": 38.217613220214844,
+      "learning_rate": 9.999041600813393e-06,
+      "loss": 7.008,
+      "step": 120
+    },
+    {
+      "epoch": 0.03928382564025081,
+      "grad_norm": 31.28360366821289,
+      "learning_rate": 9.997843687944153e-06,
+      "loss": 6.2504,
+      "step": 130
+    },
+    {
+      "epoch": 0.04230565838180857,
+      "grad_norm": 29.08894920349121,
+      "learning_rate": 9.996166770665168e-06,
+      "loss": 5.5792,
+      "step": 140
+    },
+    {
+      "epoch": 0.04532749112336632,
+      "grad_norm": 31.739566802978516,
+      "learning_rate": 9.994011009695908e-06,
+      "loss": 5.6811,
+      "step": 150
+    },
+    {
+      "epoch": 0.048349323864924076,
+      "grad_norm": 30.99082374572754,
+      "learning_rate": 9.991376611649278e-06,
+      "loss": 6.0987,
+      "step": 160
+    },
+    {
+      "epoch": 0.05137115660648183,
+      "grad_norm": 26.928142547607422,
+      "learning_rate": 9.988263829011821e-06,
+      "loss": 5.4118,
+      "step": 170
+    },
+    {
+      "epoch": 0.054392989348039586,
+      "grad_norm": 25.255813598632812,
+      "learning_rate": 9.984672960119523e-06,
+      "loss": 5.3659,
+      "step": 180
+    },
+    {
+      "epoch": 0.05741482208959734,
+      "grad_norm": 39.82838439941406,
+      "learning_rate": 9.980604349129212e-06,
+      "loss": 5.763,
+      "step": 190
+    },
+    {
+      "epoch": 0.060436654831155096,
+      "grad_norm": 26.258312225341797,
+      "learning_rate": 9.976058385985575e-06,
+      "loss": 5.1983,
+      "step": 200
+    },
+    {
+      "epoch": 0.06345848757271286,
+      "grad_norm": 34.27275466918945,
+      "learning_rate": 9.971035506383791e-06,
+      "loss": 5.533,
+      "step": 210
+    },
+    {
+      "epoch": 0.0664803203142706,
+      "grad_norm": 29.04474449157715,
+      "learning_rate": 9.96553619172777e-06,
+      "loss": 5.0581,
+      "step": 220
+    },
+    {
+      "epoch": 0.06950215305582837,
+      "grad_norm": 30.255725860595703,
+      "learning_rate": 9.959560969084004e-06,
+      "loss": 5.3482,
+      "step": 230
+    },
+    {
+      "epoch": 0.07252398579738611,
+      "grad_norm": 27.225391387939453,
+      "learning_rate": 9.953110411131073e-06,
+      "loss": 4.7629,
+      "step": 240
+    },
+    {
+      "epoch": 0.07554581853894388,
+      "grad_norm": 28.110713958740234,
+      "learning_rate": 9.946185136104736e-06,
+      "loss": 5.4251,
+      "step": 250
+    },
+    {
+      "epoch": 0.07856765128050162,
+      "grad_norm": 30.360767364501953,
+      "learning_rate": 9.938785807738692e-06,
+      "loss": 4.8019,
+      "step": 260
+    },
+    {
+      "epoch": 0.08158948402205939,
+      "grad_norm": 31.589569091796875,
+      "learning_rate": 9.930913135200964e-06,
+      "loss": 5.4093,
+      "step": 270
+    },
+    {
+      "epoch": 0.08461131676361713,
+      "grad_norm": 23.243560791015625,
+      "learning_rate": 9.922567873025924e-06,
+      "loss": 5.2431,
+      "step": 280
+    },
+    {
+      "epoch": 0.0876331495051749,
+      "grad_norm": 37.121124267578125,
+      "learning_rate": 9.913750821041988e-06,
+      "loss": 4.5542,
+      "step": 290
+    },
+    {
+      "epoch": 0.09065498224673264,
+      "grad_norm": 25.293624877929688,
+      "learning_rate": 9.904462824294945e-06,
+      "loss": 4.5975,
+      "step": 300
+    },
+    {
+      "epoch": 0.0936768149882904,
+      "grad_norm": 26.541584014892578,
+      "learning_rate": 9.894704772966978e-06,
+      "loss": 4.4755,
+      "step": 310
+    },
+    {
+      "epoch": 0.09669864772984815,
+      "grad_norm": 23.836275100708008,
+      "learning_rate": 9.884477602291343e-06,
+      "loss": 4.4776,
+      "step": 320
+    },
+    {
+      "epoch": 0.09972048047140591,
+      "grad_norm": 33.217247009277344,
+      "learning_rate": 9.873782292462727e-06,
+      "loss": 4.3572,
+      "step": 330
+    },
+    {
+      "epoch": 0.10274231321296366,
+      "grad_norm": 27.012453079223633,
+      "learning_rate": 9.862619868543323e-06,
+      "loss": 8.1775,
+      "step": 340
+    },
+    {
+      "epoch": 0.10576414595452142,
+      "grad_norm": 37.925636291503906,
+      "learning_rate": 9.850991400364557e-06,
+      "loss": 5.1596,
+      "step": 350
+    },
+    {
+      "epoch": 0.10878597869607917,
+      "grad_norm": 25.5006046295166,
+      "learning_rate": 9.838898002424586e-06,
+      "loss": 6.0762,
+      "step": 360
+    },
+    {
+      "epoch": 0.11180781143763693,
+      "grad_norm": 30.08086585998535,
+      "learning_rate": 9.826340833781448e-06,
+      "loss": 5.8416,
+      "step": 370
+    },
+    {
+      "epoch": 0.11482964417919468,
+      "grad_norm": 19.61676788330078,
+      "learning_rate": 9.813321097942005e-06,
+      "loss": 5.0978,
+      "step": 380
+    },
+    {
+      "epoch": 0.11785147692075244,
+      "grad_norm": 27.508745193481445,
+      "learning_rate": 9.79984004274658e-06,
+      "loss": 4.8911,
+      "step": 390
+    },
+    {
+      "epoch": 0.12087330966231019,
+      "grad_norm": 21.87676239013672,
+      "learning_rate": 9.785898960249365e-06,
+      "loss": 4.2318,
+      "step": 400
+    },
+    {
+      "epoch": 0.12389514240386795,
+      "grad_norm": 19.91452407836914,
+      "learning_rate": 9.771499186594586e-06,
+      "loss": 5.0726,
+      "step": 410
+    },
+    {
+      "epoch": 0.12691697514542571,
+      "grad_norm": 30.188919067382812,
+      "learning_rate": 9.756642101888449e-06,
+      "loss": 5.1973,
+      "step": 420
+    },
+    {
+      "epoch": 0.12993880788698345,
+      "grad_norm": 36.87270736694336,
+      "learning_rate": 9.74132913006686e-06,
+      "loss": 3.6298,
+      "step": 430
+    },
+    {
+      "epoch": 0.1329606406285412,
+      "grad_norm": 20.732437133789062,
+      "learning_rate": 9.725561738758956e-06,
+      "loss": 3.5142,
+      "step": 440
+    },
+    {
+      "epoch": 0.13598247337009897,
+      "grad_norm": 22.868412017822266,
+      "learning_rate": 9.709341439146452e-06,
+      "loss": 5.1262,
+      "step": 450
+    },
+    {
+      "epoch": 0.13900430611165673,
+      "grad_norm": 25.75754737854004,
+      "learning_rate": 9.692669785818787e-06,
+      "loss": 6.4161,
+      "step": 460
+    },
+    {
+      "epoch": 0.14202613885321447,
+      "grad_norm": 35.30986785888672,
+      "learning_rate": 9.675548376624149e-06,
+      "loss": 5.7558,
+      "step": 470
+    },
+    {
+      "epoch": 0.14504797159477223,
+      "grad_norm": 27.885366439819336,
+      "learning_rate": 9.657978852516318e-06,
+      "loss": 5.6992,
+      "step": 480
+    },
+    {
+      "epoch": 0.14806980433633,
+      "grad_norm": 21.603235244750977,
+      "learning_rate": 9.639962897397405e-06,
+      "loss": 4.2,
+      "step": 490
+    },
+    {
+      "epoch": 0.15109163707788775,
+      "grad_norm": 21.881452560424805,
+      "learning_rate": 9.621502237956452e-06,
+      "loss": 5.0617,
+      "step": 500
+    },
+    {
+      "epoch": 0.1541134698194455,
+      "grad_norm": 21.701580047607422,
+      "learning_rate": 9.602598643503957e-06,
+      "loss": 3.2664,
+      "step": 510
+    },
+    {
+      "epoch": 0.15713530256100325,
+      "grad_norm": 29.036582946777344,
+      "learning_rate": 9.583253925802283e-06,
+      "loss": 4.2414,
+      "step": 520
+    },
+    {
+      "epoch": 0.160157135302561,
+      "grad_norm": 22.403575897216797,
+      "learning_rate": 9.563469938892023e-06,
+      "loss": 4.8757,
+      "step": 530
+    },
+    {
+      "epoch": 0.16317896804411877,
+      "grad_norm": 23.00878143310547,
+      "learning_rate": 9.543248578914309e-06,
+      "loss": 3.2053,
+      "step": 540
+    },
+    {
+      "epoch": 0.1662008007856765,
+      "grad_norm": 22.494098663330078,
+      "learning_rate": 9.522591783929069e-06,
+      "loss": 4.8134,
+      "step": 550
+    },
+    {
+      "epoch": 0.16922263352723427,
+      "grad_norm": 20.914226531982422,
+      "learning_rate": 9.501501533729297e-06,
+      "loss": 4.1195,
+      "step": 560
+    },
+    {
+      "epoch": 0.17224446626879203,
+      "grad_norm": 33.715003967285156,
+      "learning_rate": 9.479979849651287e-06,
+      "loss": 5.765,
+      "step": 570
+    },
+    {
+      "epoch": 0.1752662990103498,
+      "grad_norm": 23.223495483398438,
+      "learning_rate": 9.45802879438091e-06,
+      "loss": 6.3336,
+      "step": 580
+    },
+    {
+      "epoch": 0.17828813175190752,
+      "grad_norm": 23.889699935913086,
+      "learning_rate": 9.43565047175593e-06,
+      "loss": 4.1538,
+      "step": 590
+    },
+    {
+      "epoch": 0.1813099644934653,
+      "grad_norm": 24.332725524902344,
+      "learning_rate": 9.412847026564359e-06,
+      "loss": 3.9682,
+      "step": 600
+    },
+    {
+      "epoch": 0.18433179723502305,
+      "grad_norm": 21.96092414855957,
+      "learning_rate": 9.389620644338893e-06,
+      "loss": 4.1644,
+      "step": 610
+    },
+    {
+      "epoch": 0.1873536299765808,
+      "grad_norm": 36.027503967285156,
+      "learning_rate": 9.365973551147453e-06,
+      "loss": 4.7194,
+      "step": 620
+    },
+    {
+      "epoch": 0.19037546271813854,
+      "grad_norm": 22.030109405517578,
+      "learning_rate": 9.341908013379832e-06,
+      "loss": 4.7219,
+      "step": 630
+    },
+    {
+      "epoch": 0.1933972954596963,
+      "grad_norm": 23.482004165649414,
+      "learning_rate": 9.317426337530477e-06,
+      "loss": 4.0362,
+      "step": 640
+    },
+    {
+      "epoch": 0.19641912820125407,
+      "grad_norm": 22.311803817749023,
+      "learning_rate": 9.292530869977432e-06,
+      "loss": 5.5712,
+      "step": 650
+    },
+    {
+      "epoch": 0.19944096094281183,
+      "grad_norm": 22.5808162689209,
+      "learning_rate": 9.26722399675745e-06,
+      "loss": 3.225,
+      "step": 660
+    },
+    {
+      "epoch": 0.20246279368436956,
+      "grad_norm": 20.662601470947266,
+      "learning_rate": 9.24150814333732e-06,
+      "loss": 3.9247,
+      "step": 670
+    },
+    {
+      "epoch": 0.20548462642592732,
+      "grad_norm": 19.702299118041992,
+      "learning_rate": 9.215385774381395e-06,
+      "loss": 6.2027,
+      "step": 680
+    },
+    {
+      "epoch": 0.2085064591674851,
+      "grad_norm": 41.593204498291016,
+      "learning_rate": 9.188859393515382e-06,
+      "loss": 4.8472,
+      "step": 690
+    },
+    {
+      "epoch": 0.21152829190904285,
+      "grad_norm": 24.661819458007812,
+      "learning_rate": 9.16193154308638e-06,
+      "loss": 6.0424,
+      "step": 700
+    },
+    {
+      "epoch": 0.21455012465060058,
+      "grad_norm": 35.47103500366211,
+      "learning_rate": 9.13460480391922e-06,
+      "loss": 6.2253,
+      "step": 710
+    },
+    {
+      "epoch": 0.21757195739215834,
+      "grad_norm": 37.96540069580078,
+      "learning_rate": 9.106881795069116e-06,
+      "loss": 6.4773,
+      "step": 720
+    },
+    {
+      "epoch": 0.2205937901337161,
+      "grad_norm": 18.79058074951172,
+      "learning_rate": 9.078765173570649e-06,
+      "loss": 3.1354,
+      "step": 730
+    },
+    {
+      "epoch": 0.22361562287527387,
+      "grad_norm": 18.278860092163086,
+      "learning_rate": 9.0502576341831e-06,
+      "loss": 4.0845,
+      "step": 740
+    },
+    {
+      "epoch": 0.2266374556168316,
+      "grad_norm": 21.75337791442871,
+      "learning_rate": 9.02136190913219e-06,
+      "loss": 5.4407,
+      "step": 750
+    },
+    {
+      "epoch": 0.22965928835838936,
+      "grad_norm": 36.95095443725586,
+      "learning_rate": 8.99208076784822e-06,
+      "loss": 4.6502,
+      "step": 760
+    },
+    {
+      "epoch": 0.23268112109994712,
+      "grad_norm": 14.346115112304688,
+      "learning_rate": 8.962417016700624e-06,
+      "loss": 3.0298,
+      "step": 770
+    },
+    {
+      "epoch": 0.2357029538415049,
+      "grad_norm": 22.408119201660156,
+      "learning_rate": 8.932373498729026e-06,
+      "loss": 4.6357,
+      "step": 780
+    },
+    {
+      "epoch": 0.23872478658306262,
+      "grad_norm": 23.731843948364258,
+      "learning_rate": 8.901953093370734e-06,
+      "loss": 4.0057,
+      "step": 790
+    },
+    {
+      "epoch": 0.24174661932462038,
+      "grad_norm": 21.797941207885742,
+      "learning_rate": 8.871158716184784e-06,
+      "loss": 3.9041,
+      "step": 800
+    },
+    {
+      "epoch": 0.24476845206617814,
+      "grad_norm": 32.5623893737793,
+      "learning_rate": 8.839993318572497e-06,
+      "loss": 4.8501,
+      "step": 810
+    },
+    {
+      "epoch": 0.2477902848077359,
+      "grad_norm": 18.512962341308594,
+      "learning_rate": 8.808459887494617e-06,
+      "loss": 3.05,
+      "step": 820
+    },
+    {
+      "epoch": 0.25081211754929367,
+      "grad_norm": 16.462116241455078,
+      "learning_rate": 8.77656144518502e-06,
+      "loss": 3.8237,
+      "step": 830
+    },
+    {
+      "epoch": 0.25383395029085143,
+      "grad_norm": 18.127500534057617,
+      "learning_rate": 8.744301048861083e-06,
+      "loss": 2.9264,
+      "step": 840
+    },
+    {
+      "epoch": 0.25685578303240914,
+      "grad_norm": 22.66571044921875,
+      "learning_rate": 8.711681790430646e-06,
+      "loss": 2.9888,
+      "step": 850
+    },
+    {
+      "epoch": 0.2598776157739669,
+      "grad_norm": 24.76766586303711,
+      "learning_rate": 8.678706796195694e-06,
+      "loss": 4.7451,
+      "step": 860
+    },
+    {
+      "epoch": 0.26289944851552466,
+      "grad_norm": 20.290781021118164,
+      "learning_rate": 8.645379226552712e-06,
+      "loss": 3.7028,
+      "step": 870
+    },
+    {
+      "epoch": 0.2659212812570824,
+      "grad_norm": 19.00446891784668,
+      "learning_rate": 8.611702275689805e-06,
+      "loss": 4.6546,
+      "step": 880
+    },
+    {
+      "epoch": 0.2689431139986402,
+      "grad_norm": 34.68561935424805,
+      "learning_rate": 8.577679171280538e-06,
+      "loss": 4.4799,
+      "step": 890
+    },
+    {
+      "epoch": 0.27196494674019794,
+      "grad_norm": 18.584571838378906,
+      "learning_rate": 8.543313174174601e-06,
+      "loss": 5.2755,
+      "step": 900
+    },
+    {
+      "epoch": 0.2749867794817557,
+      "grad_norm": 23.914329528808594,
+      "learning_rate": 8.508607578085281e-06,
+      "loss": 3.846,
+      "step": 910
+    },
+    {
+      "epoch": 0.27800861222331347,
+      "grad_norm": 21.329547882080078,
+      "learning_rate": 8.473565709273786e-06,
+      "loss": 3.8483,
+      "step": 920
+    },
+    {
+      "epoch": 0.2810304449648712,
+      "grad_norm": 14.663247108459473,
+      "learning_rate": 8.438190926230439e-06,
+      "loss": 3.8175,
+      "step": 930
+    },
+    {
+      "epoch": 0.28405227770642894,
+      "grad_norm": 32.22755432128906,
+      "learning_rate": 8.40248661935281e-06,
+      "loss": 4.3899,
+      "step": 940
+    },
+    {
+      "epoch": 0.2870741104479867,
+      "grad_norm": 24.638607025146484,
+      "learning_rate": 8.366456210620756e-06,
+      "loss": 3.1177,
+      "step": 950
+    },
+    {
+      "epoch": 0.29009594318954446,
+      "grad_norm": 33.2247428894043,
+      "learning_rate": 8.330103153268464e-06,
+      "loss": 3.7424,
+      "step": 960
+    },
+    {
+      "epoch": 0.2931177759311022,
+      "grad_norm": 20.71030616760254,
+      "learning_rate": 8.29343093145347e-06,
+      "loss": 3.6926,
+      "step": 970
+    },
+    {
+      "epoch": 0.29613960867266,
+      "grad_norm": 21.19091033935547,
+      "learning_rate": 8.25644305992275e-06,
+      "loss": 3.6266,
+      "step": 980
+    },
+    {
+      "epoch": 0.29916144141421774,
+      "grad_norm": 19.403636932373047,
+      "learning_rate": 8.21914308367584e-06,
+      "loss": 4.561,
+      "step": 990
+    },
+    {
+      "epoch": 0.3021832741557755,
+      "grad_norm": 20.385461807250977,
+      "learning_rate": 8.181534577625088e-06,
+      "loss": 3.7729,
+      "step": 1000
+    },
+    {
+      "epoch": 0.3052051068973332,
+      "grad_norm": 16.03277587890625,
+      "learning_rate": 8.143621146253022e-06,
+      "loss": 4.5904,
+      "step": 1010
+    },
+    {
+      "epoch": 0.308226939638891,
+      "grad_norm": 17.92051887512207,
+      "learning_rate": 8.105406423266884e-06,
+      "loss": 4.6473,
+      "step": 1020
+    },
+    {
+      "epoch": 0.31124877238044873,
+      "grad_norm": 19.62211036682129,
+      "learning_rate": 8.066894071250374e-06,
+      "loss": 4.4546,
+      "step": 1030
+    },
+    {
+      "epoch": 0.3142706051220065,
+      "grad_norm": 18.912364959716797,
+      "learning_rate": 8.02808778131262e-06,
+      "loss": 3.8032,
+      "step": 1040
+    },
+    {
+      "epoch": 0.31729243786356426,
+      "grad_norm": 17.066823959350586,
+      "learning_rate": 7.988991272734407e-06,
+      "loss": 4.4699,
+      "step": 1050
+    },
+    {
+      "epoch": 0.320314270605122,
+      "grad_norm": 19.543676376342773,
+      "learning_rate": 7.94960829261172e-06,
+      "loss": 4.4354,
+      "step": 1060
+    },
+    {
+      "epoch": 0.3233361033466798,
+      "grad_norm": 21.18368911743164,
+      "learning_rate": 7.909942615496613e-06,
+      "loss": 4.7008,
+      "step": 1070
+    },
+    {
+      "epoch": 0.32635793608823754,
+      "grad_norm": 29.533164978027344,
+      "learning_rate": 7.869998043035442e-06,
+      "loss": 5.3891,
+      "step": 1080
+    },
+    {
+      "epoch": 0.32937976882979525,
+      "grad_norm": 14.957313537597656,
+      "learning_rate": 7.829778403604512e-06,
+      "loss": 5.0378,
+      "step": 1090
+    },
+    {
+      "epoch": 0.332401601571353,
+      "grad_norm": 33.31967544555664,
+      "learning_rate": 7.789287551943158e-06,
+      "loss": 6.0914,
+      "step": 1100
+    },
+    {
+      "epoch": 0.3354234343129108,
+      "grad_norm": 17.046133041381836,
+      "learning_rate": 7.748529368784293e-06,
+      "loss": 4.5389,
+      "step": 1110
+    },
+    {
+      "epoch": 0.33844526705446853,
+      "grad_norm": 16.102283477783203,
+      "learning_rate": 7.707507760482473e-06,
+      "loss": 6.1424,
+      "step": 1120
+    },
+    {
+      "epoch": 0.3414670997960263,
+      "grad_norm": 16.310462951660156,
+      "learning_rate": 7.666226658639507e-06,
+      "loss": 3.7652,
+      "step": 1130
+    },
+    {
+      "epoch": 0.34448893253758406,
+      "grad_norm": 22.625951766967773,
+      "learning_rate": 7.624690019727636e-06,
+      "loss": 3.6742,
+      "step": 1140
+    },
+    {
+      "epoch": 0.3475107652791418,
+      "grad_norm": 21.400938034057617,
+      "learning_rate": 7.58290182471034e-06,
+      "loss": 4.5513,
+      "step": 1150
+    },
+    {
+      "epoch": 0.3505325980206996,
+      "grad_norm": 14.196311950683594,
+      "learning_rate": 7.5408660786607976e-06,
+      "loss": 3.6687,
+      "step": 1160
+    },
+    {
+      "epoch": 0.3535544307622573,
+      "grad_norm": 19.97701644897461,
+      "learning_rate": 7.498586810378019e-06,
+      "loss": 2.9966,
+      "step": 1170
+    },
+    {
+      "epoch": 0.35657626350381505,
+      "grad_norm": 18.132966995239258,
+      "learning_rate": 7.456068072000731e-06,
+      "loss": 2.809,
+      "step": 1180
+    },
+    {
+      "epoch": 0.3595980962453728,
+      "grad_norm": 22.550369262695312,
+      "learning_rate": 7.4133139386190026e-06,
+      "loss": 4.5377,
+      "step": 1190
+    },
+    {
+      "epoch": 0.3626199289869306,
+      "grad_norm": 15.473170280456543,
+      "learning_rate": 7.3703285078836796e-06,
+      "loss": 5.1951,
+      "step": 1200
+    },
+    {
+      "epoch": 0.36564176172848833,
+      "grad_norm": 17.210092544555664,
+      "learning_rate": 7.3271158996136625e-06,
+      "loss": 3.7087,
+      "step": 1210
+    },
+    {
+      "epoch": 0.3686635944700461,
+      "grad_norm": 15.88957691192627,
+      "learning_rate": 7.283680255401049e-06,
+      "loss": 4.4085,
+      "step": 1220
+    },
+    {
+      "epoch": 0.37168542721160386,
+      "grad_norm": 20.273908615112305,
+      "learning_rate": 7.240025738214193e-06,
+      "loss": 6.2167,
+      "step": 1230
+    },
+    {
+      "epoch": 0.3747072599531616,
+      "grad_norm": 31.755516052246094,
+      "learning_rate": 7.196156531998718e-06,
+      "loss": 4.532,
+      "step": 1240
+    },
+    {
+      "epoch": 0.3777290926947193,
+      "grad_norm": 20.531906127929688,
+      "learning_rate": 7.152076841276527e-06,
+      "loss": 3.6233,
+      "step": 1250
+    },
+    {
+      "epoch": 0.3807509254362771,
+      "grad_norm": 18.760116577148438,
+      "learning_rate": 7.1077908907428154e-06,
+      "loss": 3.7953,
+      "step": 1260
+    },
+    {
+      "epoch": 0.38377275817783485,
+      "grad_norm": 31.519561767578125,
+      "learning_rate": 7.063302924861182e-06,
+      "loss": 3.8545,
+      "step": 1270
+    },
+    {
+      "epoch": 0.3867945909193926,
+      "grad_norm": 23.91765022277832,
+      "learning_rate": 7.018617207456821e-06,
+      "loss": 3.6115,
+      "step": 1280
+    },
+    {
+      "epoch": 0.3898164236609504,
+      "grad_norm": 16.870439529418945,
+      "learning_rate": 6.973738021307872e-06,
+      "loss": 3.6868,
+      "step": 1290
+    },
+    {
+      "epoch": 0.39283825640250813,
+      "grad_norm": 18.010969161987305,
+      "learning_rate": 6.9286696677349455e-06,
+      "loss": 5.9382,
+      "step": 1300
+    },
+    {
+      "epoch": 0.3958600891440659,
+      "grad_norm": 16.40668296813965,
+      "learning_rate": 6.883416466188881e-06,
+      "loss": 3.7045,
+      "step": 1310
+    },
+    {
+      "epoch": 0.39888192188562366,
+      "grad_norm": 15.330764770507812,
+      "learning_rate": 6.837982753836755e-06,
+      "loss": 2.8419,
+      "step": 1320
+    },
+    {
+      "epoch": 0.40190375462718136,
+      "grad_norm": 33.37950897216797,
+      "learning_rate": 6.7923728851461955e-06,
+      "loss": 6.0184,
+      "step": 1330
+    },
+    {
+      "epoch": 0.4049255873687391,
+      "grad_norm": 21.508163452148438,
+      "learning_rate": 6.74659123146805e-06,
+      "loss": 3.6311,
+      "step": 1340
+    },
+    {
+      "epoch": 0.4079474201102969,
+      "grad_norm": 16.043455123901367,
+      "learning_rate": 6.70064218061742e-06,
+      "loss": 2.8293,
+      "step": 1350
+    },
+    {
+      "epoch": 0.41096925285185465,
+      "grad_norm": 16.867509841918945,
+      "learning_rate": 6.654530136453119e-06,
+      "loss": 4.3584,
+      "step": 1360
+    },
+    {
+      "epoch": 0.4139910855934124,
+      "grad_norm": 19.85649871826172,
+      "learning_rate": 6.608259518455599e-06,
+      "loss": 5.2523,
+      "step": 1370
+    },
+    {
+      "epoch": 0.4170129183349702,
+      "grad_norm": 14.44941520690918,
+      "learning_rate": 6.5618347613033875e-06,
+      "loss": 5.0827,
+      "step": 1380
+    },
+    {
+      "epoch": 0.42003475107652793,
+      "grad_norm": 18.058303833007812,
+      "learning_rate": 6.5152603144480406e-06,
+      "loss": 5.9903,
+      "step": 1390
+    },
+    {
+      "epoch": 0.4230565838180857,
+      "grad_norm": 30.503751754760742,
+      "learning_rate": 6.468540641687716e-06,
+      "loss": 4.5515,
+      "step": 1400
+    },
+    {
+      "epoch": 0.4260784165596434,
+      "grad_norm": 18.618654251098633,
+      "learning_rate": 6.421680220739337e-06,
+      "loss": 3.9462,
+      "step": 1410
+    },
+    {
+      "epoch": 0.42910024930120116,
+      "grad_norm": 32.41435623168945,
+      "learning_rate": 6.374683542809447e-06,
+      "loss": 7.8369,
+      "step": 1420
+    },
+    {
+      "epoch": 0.4321220820427589,
+      "grad_norm": 19.87618637084961,
+      "learning_rate": 6.327555112163761e-06,
+      "loss": 4.3855,
+      "step": 1430
+    },
+    {
+      "epoch": 0.4351439147843167,
+      "grad_norm": 17.212278366088867,
+      "learning_rate": 6.280299445695469e-06,
+      "loss": 5.1971,
+      "step": 1440
+    },
+    {
+      "epoch": 0.43816574752587445,
+      "grad_norm": 21.317800521850586,
+      "learning_rate": 6.232921072492319e-06,
+      "loss": 4.3216,
+      "step": 1450
+    },
+    {
+      "epoch": 0.4411875802674322,
+      "grad_norm": 23.228479385375977,
+      "learning_rate": 6.185424533402543e-06,
+      "loss": 4.3291,
+      "step": 1460
+    },
+    {
+      "epoch": 0.44420941300899,
+      "grad_norm": 19.224565505981445,
+      "learning_rate": 6.13781438059966e-06,
+      "loss": 3.5382,
+      "step": 1470
+    },
+    {
+      "epoch": 0.44723124575054773,
+      "grad_norm": 33.251380920410156,
+      "learning_rate": 6.090095177146178e-06,
+      "loss": 5.1775,
+      "step": 1480
+    },
+    {
+      "epoch": 0.45025307849210544,
+      "grad_norm": 16.563169479370117,
+      "learning_rate": 6.042271496556255e-06,
+      "loss": 2.8536,
+      "step": 1490
+    },
+    {
+      "epoch": 0.4532749112336632,
+      "grad_norm": 22.154621124267578,
+      "learning_rate": 5.994347922357372e-06,
+      "loss": 3.8333,
+      "step": 1500
+    },
+    {
+      "epoch": 0.45629674397522096,
+      "grad_norm": 20.189830780029297,
+      "learning_rate": 5.946329047651037e-06,
+      "loss": 3.5955,
+      "step": 1510
+    },
+    {
+      "epoch": 0.4593185767167787,
+      "grad_norm": 30.116336822509766,
+      "learning_rate": 5.8982194746725686e-06,
+      "loss": 2.7101,
+      "step": 1520
+    },
+    {
+      "epoch": 0.4623404094583365,
+      "grad_norm": 21.328214645385742,
+      "learning_rate": 5.850023814350007e-06,
+      "loss": 4.2863,
+      "step": 1530
+    },
+    {
+      "epoch": 0.46536224219989425,
+      "grad_norm": 15.175004005432129,
+      "learning_rate": 5.801746685862197e-06,
+      "loss": 6.1059,
+      "step": 1540
+    },
+    {
+      "epoch": 0.468384074941452,
+      "grad_norm": 15.263406753540039,
+      "learning_rate": 5.753392716196069e-06,
+      "loss": 2.8679,
+      "step": 1550
+    },
+    {
+      "epoch": 0.4714059076830098,
+      "grad_norm": 14.840780258178711,
+      "learning_rate": 5.704966539703185e-06,
+      "loss": 3.6484,
+      "step": 1560
+    },
+    {
+      "epoch": 0.4744277404245675,
+      "grad_norm": 16.718647003173828,
+      "learning_rate": 5.656472797655571e-06,
+      "loss": 4.4473,
+      "step": 1570
+    },
+    {
+      "epoch": 0.47744957316612524,
+      "grad_norm": 15.991732597351074,
+      "learning_rate": 5.60791613780088e-06,
+      "loss": 2.7819,
+      "step": 1580
+    },
+    {
+      "epoch": 0.480471405907683,
+      "grad_norm": 21.31011390686035,
+      "learning_rate": 5.5593012139169525e-06,
+      "loss": 4.3017,
+      "step": 1590
+    },
+    {
+      "epoch": 0.48349323864924076,
+      "grad_norm": 19.14992332458496,
+      "learning_rate": 5.510632685365777e-06,
+      "loss": 4.443,
+      "step": 1600
+    },
+    {
+      "epoch": 0.4865150713907985,
+      "grad_norm": 16.440427780151367,
+      "learning_rate": 5.461915216646938e-06,
+      "loss": 2.8006,
+      "step": 1610
+    },
+    {
+      "epoch": 0.4895369041323563,
+      "grad_norm": 22.039566040039062,
+      "learning_rate": 5.41315347695055e-06,
+      "loss": 4.2488,
+      "step": 1620
+    },
+    {
+      "epoch": 0.49255873687391405,
+      "grad_norm": 17.224102020263672,
+      "learning_rate": 5.364352139709758e-06,
+      "loss": 4.9229,
+      "step": 1630
+    },
+    {
+      "epoch": 0.4955805696154718,
+      "grad_norm": 15.041423797607422,
+      "learning_rate": 5.315515882152822e-06,
+      "loss": 3.5541,
+      "step": 1640
+    },
+    {
+      "epoch": 0.4986024023570295,
+      "grad_norm": 16.443140029907227,
+      "learning_rate": 5.266649384854842e-06,
+      "loss": 4.2881,
+      "step": 1650
+    },
+    {
+      "epoch": 0.5016242350985873,
+      "grad_norm": 20.163755416870117,
+      "learning_rate": 5.217757331289165e-06,
+      "loss": 3.721,
+      "step": 1660
+    },
+    {
+      "epoch": 0.5046460678401451,
+      "grad_norm": 16.496013641357422,
+      "learning_rate": 5.168844407378506e-06,
+      "loss": 4.8698,
+      "step": 1670
+    },
+    {
+      "epoch": 0.5076679005817029,
+      "grad_norm": 12.180791854858398,
+      "learning_rate": 5.119915301045836e-06,
+      "loss": 2.9101,
+      "step": 1680
+    },
+    {
+      "epoch": 0.5106897333232605,
+      "grad_norm": 22.121103286743164,
+      "learning_rate": 5.070974701765089e-06,
+      "loss": 5.189,
+      "step": 1690
+    },
+    {
+      "epoch": 0.5137115660648183,
+      "grad_norm": 19.323467254638672,
+      "learning_rate": 5.022027300111712e-06,
+      "loss": 4.3854,
+      "step": 1700
+    },
+    {
+      "epoch": 0.516733398806376,
+      "grad_norm": 16.899927139282227,
+      "learning_rate": 4.973077787313099e-06,
+      "loss": 4.526,
+      "step": 1710
+    },
+    {
+      "epoch": 0.5197552315479338,
+      "grad_norm": 15.249711036682129,
+      "learning_rate": 4.924130854798983e-06,
+      "loss": 5.1185,
+      "step": 1720
+    },
+    {
+      "epoch": 0.5227770642894916,
+      "grad_norm": 14.765595436096191,
+      "learning_rate": 4.875191193751803e-06,
+      "loss": 2.7791,
+      "step": 1730
+    },
+    {
+      "epoch": 0.5257988970310493,
+      "grad_norm": 20.05950164794922,
+      "learning_rate": 4.826263494657077e-06,
+      "loss": 3.511,
+      "step": 1740
+    },
+    {
+      "epoch": 0.5288207297726071,
+      "grad_norm": 34.354034423828125,
+      "learning_rate": 4.777352446853863e-06,
+      "loss": 5.0201,
+      "step": 1750
+    },
+    {
+      "epoch": 0.5318425625141648,
+      "grad_norm": 20.756092071533203,
+      "learning_rate": 4.72846273808533e-06,
+      "loss": 3.5231,
+      "step": 1760
+    },
+    {
+      "epoch": 0.5348643952557226,
+      "grad_norm": 17.004915237426758,
+      "learning_rate": 4.679599054049458e-06,
+      "loss": 3.3843,
+      "step": 1770
+    },
+    {
+      "epoch": 0.5378862279972804,
+      "grad_norm": 21.614294052124023,
+      "learning_rate": 4.630766077949965e-06,
+      "loss": 5.98,
+      "step": 1780
+    },
+    {
+      "epoch": 0.5409080607388381,
+      "grad_norm": 32.01701354980469,
+      "learning_rate": 4.5819684900474484e-06,
+      "loss": 4.3486,
+      "step": 1790
+    },
+    {
+      "epoch": 0.5439298934803959,
+      "grad_norm": 14.644387245178223,
+      "learning_rate": 4.5332109672108245e-06,
+      "loss": 4.4888,
+      "step": 1800
+    },
+    {
+      "epoch": 0.5469517262219536,
+      "grad_norm": 13.374945640563965,
+      "learning_rate": 4.484498182469085e-06,
+      "loss": 3.6398,
+      "step": 1810
+    },
+    {
+      "epoch": 0.5499735589635114,
+      "grad_norm": 17.148927688598633,
+      "learning_rate": 4.435834804563422e-06,
+      "loss": 5.9183,
+      "step": 1820
+    },
+    {
+      "epoch": 0.5529953917050692,
+      "grad_norm": 18.936981201171875,
+      "learning_rate": 4.387225497499767e-06,
+      "loss": 3.6337,
+      "step": 1830
+    },
+    {
+      "epoch": 0.5560172244466269,
+      "grad_norm": 17.948814392089844,
+      "learning_rate": 4.3386749201017856e-06,
+      "loss": 3.4775,
+      "step": 1840
+    },
+    {
+      "epoch": 0.5590390571881846,
+      "grad_norm": 14.527155876159668,
+      "learning_rate": 4.290187725564356e-06,
+      "loss": 6.0241,
+      "step": 1850
+    },
+    {
+      "epoch": 0.5620608899297423,
+      "grad_norm": 14.412932395935059,
+      "learning_rate": 4.2417685610076135e-06,
+      "loss": 3.4722,
+      "step": 1860
+    },
+    {
+      "epoch": 0.5650827226713001,
+      "grad_norm": 14.845317840576172,
+      "learning_rate": 4.193422067031535e-06,
+      "loss": 4.3061,
+      "step": 1870
+    },
+    {
+      "epoch": 0.5681045554128579,
+      "grad_norm": 16.743505477905273,
+      "learning_rate": 4.145152877271196e-06,
+      "loss": 4.1782,
+      "step": 1880
+    },
+    {
+      "epoch": 0.5711263881544156,
+      "grad_norm": 17.867549896240234,
+      "learning_rate": 4.096965617952667e-06,
+      "loss": 4.4313,
+      "step": 1890
+    },
+    {
+      "epoch": 0.5741482208959734,
+      "grad_norm": 21.779705047607422,
+      "learning_rate": 4.048864907449619e-06,
+      "loss": 3.5466,
+      "step": 1900
+    },
+    {
+      "epoch": 0.5771700536375312,
+      "grad_norm": 18.34054183959961,
+      "learning_rate": 4.000855355840695e-06,
+      "loss": 3.5575,
+      "step": 1910
+    },
+    {
+      "epoch": 0.5801918863790889,
+      "grad_norm": 16.225175857543945,
+      "learning_rate": 3.952941564467665e-06,
+      "loss": 4.307,
+      "step": 1920
+    },
+    {
+      "epoch": 0.5832137191206467,
+      "grad_norm": 17.98233413696289,
+      "learning_rate": 3.905128125494427e-06,
+      "loss": 4.3588,
+      "step": 1930
+    },
+    {
+      "epoch": 0.5862355518622044,
+      "grad_norm": 32.549110412597656,
+      "learning_rate": 3.8574196214668876e-06,
+      "loss": 4.5427,
+      "step": 1940
+    },
+    {
+      "epoch": 0.5892573846037622,
+      "grad_norm": 29.19014549255371,
+      "learning_rate": 3.8098206248737486e-06,
+      "loss": 5.1964,
+      "step": 1950
+    },
+    {
+      "epoch": 0.59227921734532,
+      "grad_norm": 19.058385848999023,
+      "learning_rate": 3.7623356977082794e-06,
+      "loss": 2.615,
+      "step": 1960
+    },
+    {
+      "epoch": 0.5953010500868777,
+      "grad_norm": 17.663314819335938,
+      "learning_rate": 3.714969391031084e-06,
+      "loss": 4.3326,
+      "step": 1970
+    },
+    {
+      "epoch": 0.5983228828284355,
+      "grad_norm": 15.665877342224121,
+      "learning_rate": 3.6677262445339136e-06,
+      "loss": 3.6268,
+      "step": 1980
+    },
+    {
+      "epoch": 0.6013447155699932,
+      "grad_norm": 18.564617156982422,
+      "learning_rate": 3.6206107861045803e-06,
+      "loss": 2.6497,
+      "step": 1990
+    },
+    {
+      "epoch": 0.604366548311551,
+      "grad_norm": 13.690784454345703,
+      "learning_rate": 3.5736275313929826e-06,
+      "loss": 4.3688,
+      "step": 2000
+    }
+  ],
+  "logging_steps": 10,
+  "max_steps": 3309,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 1,
+  "save_steps": 1000,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 7.113757083291353e+18,
+  "train_batch_size": 1,
+  "trial_name": null,
+  "trial_params": null
+}

checkpoint-2000/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2fdf3cfd8ae404b89444346cfcb519afd77a9396cab7149d66c5223a19ec7f70
+size 7697

checkpoint-2000/zero_to_fp32.py

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