Upload folder using huggingface_hub

added_tokens.json

@@ -0,0 +1,5 @@
+{
+  "<|endoftext|>": 151643,
+  "<|im_end|>": 151645,
+  "<|im_start|>": 151644
+}

config.json

@@ -0,0 +1,38 @@
+{
+  "_name_or_path": "/home/azureuser/nanoLLaVA/checkpoint-5426",
+  "architectures": [
+    "BunnyQwenForCausalLM"
+  ],
+  "attention_dropout": 0.0,
+  "bos_token_id": 151645,
+  "eos_token_id": 151645,
+  "freeze_mm_mlp_adapter": false,
+  "hidden_act": "silu",
+  "hidden_size": 1024,
+  "image_aspect_ratio": "pad",
+  "initializer_range": 0.02,
+  "intermediate_size": 2816,
+  "max_position_embeddings": 32768,
+  "max_window_layers": 21,
+  "mm_hidden_size": 1152,
+  "mm_projector_lr": null,
+  "mm_projector_type": "mlp2x_gelu",
+  "mm_vision_tower": "google/siglip-so400m-patch14-384",
+  "model_type": "bunny-qwen",
+  "num_attention_heads": 16,
+  "num_hidden_layers": 24,
+  "num_key_value_heads": 16,
+  "rms_norm_eps": 1e-06,
+  "rope_theta": 1000000.0,
+  "sliding_window": 4096,
+  "tie_word_embeddings": false,
+  "tokenizer_model_max_length": 4096,
+  "tokenizer_padding_side": "right",
+  "torch_dtype": "bfloat16",
+  "transformers_version": "4.39.2",
+  "tune_mm_mlp_adapter": false,
+  "use_cache": false,
+  "use_mm_proj": true,
+  "use_sliding_window": false,
+  "vocab_size": 151936
+}

generation_config.json

@@ -0,0 +1,9 @@
+{
+  "bos_token_id": 151645,
+  "do_sample": true,
+  "eos_token_id": 151645,
+  "max_length": 4096,
+  "temperature": 0.7,
+  "top_p": 0.8,
+  "transformers_version": "4.39.2"
+}

latest

@@ -0,0 +1 @@
+global_step100

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6319126507046d2e020963da2e09e5943432eabb7ede807dab4e71bb858f5407
+size 2100155752

rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bbe0d720c4c75a6a04213fa3b64bacbe794718a53e2b56ebb67a1a795014dfad
+size 15024

rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:72452d3138d0ca2ff89429e3294a834ae7a68e8596fc757735ca56ae52509d57
+size 15024

rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f36e306fb8ebcf53a167bfd6c9af74db410a269ada1e619e3e816f5269543b9d
+size 15024

rng_state_3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bb47ce0c6f815a6f8302b0e3819b4c2315ca71dae3138d97fdceb765cdd0a039
+size 15024

scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:55aee30756eafb6867aac7a1030ef864e1bebc1c415125c9942410da063354fe
+size 1064

special_tokens_map.json

@@ -0,0 +1,27 @@
+{
+  "additional_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>"
+  ],
+  "bos_token": {
+    "content": "<|im_end|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "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
+  }
+}

tokenizer_config.json

@@ -0,0 +1,44 @@
+{
+  "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
+    }
+  },
+  "additional_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>"
+  ],
+  "bos_token": "<|im_end|>",
+  "chat_template": "{% for message in messages %}{{'<|im_start|>' + message['role'] + '\n' + message['content'] + '<|im_end|>' + '\n'}}{% endfor %}{% if add_generation_prompt %}{{ '<|im_start|>assistant\n' }}{% endif %}",
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|im_end|>",
+  "errors": "replace",
+  "model_max_length": 4096,
+  "pad_token": "<|endoftext|>",
+  "padding_side": "right",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}

trainer_state.json

@@ -0,0 +1,721 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.03685447090800203,
+  "eval_steps": 500,
+  "global_step": 100,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0,
+      "grad_norm": 28.375805268212044,
+      "learning_rate": 2.439024390243903e-07,
+      "loss": 1.9916,
+      "step": 1
+    },
+    {
+      "epoch": 0.0,
+      "grad_norm": 35.06631979328207,
+      "learning_rate": 4.878048780487805e-07,
+      "loss": 2.0906,
+      "step": 2
+    },
+    {
+      "epoch": 0.0,
+      "grad_norm": 38.398944345049074,
+      "learning_rate": 7.317073170731707e-07,
+      "loss": 2.1697,
+      "step": 3
+    },
+    {
+      "epoch": 0.0,
+      "grad_norm": 31.786491229547707,
+      "learning_rate": 9.75609756097561e-07,
+      "loss": 1.885,
+      "step": 4
+    },
+    {
+      "epoch": 0.0,
+      "grad_norm": 36.04697641354884,
+      "learning_rate": 1.2195121951219514e-06,
+      "loss": 2.055,
+      "step": 5
+    },
+    {
+      "epoch": 0.0,
+      "grad_norm": 34.00945696139665,
+      "learning_rate": 1.4634146341463414e-06,
+      "loss": 1.9784,
+      "step": 6
+    },
+    {
+      "epoch": 0.0,
+      "grad_norm": 26.542180094892565,
+      "learning_rate": 1.707317073170732e-06,
+      "loss": 1.9204,
+      "step": 7
+    },
+    {
+      "epoch": 0.0,
+      "grad_norm": 31.922434847983876,
+      "learning_rate": 1.951219512195122e-06,
+      "loss": 1.9556,
+      "step": 8
+    },
+    {
+      "epoch": 0.0,
+      "grad_norm": 29.799359182326604,
+      "learning_rate": 2.1951219512195125e-06,
+      "loss": 2.0238,
+      "step": 9
+    },
+    {
+      "epoch": 0.0,
+      "grad_norm": 24.788943060794754,
+      "learning_rate": 2.4390243902439027e-06,
+      "loss": 2.0543,
+      "step": 10
+    },
+    {
+      "epoch": 0.0,
+      "grad_norm": 23.62857327313063,
+      "learning_rate": 2.682926829268293e-06,
+      "loss": 1.974,
+      "step": 11
+    },
+    {
+      "epoch": 0.0,
+      "grad_norm": 25.440796117241586,
+      "learning_rate": 2.926829268292683e-06,
+      "loss": 1.9558,
+      "step": 12
+    },
+    {
+      "epoch": 0.0,
+      "grad_norm": 11.507873994835268,
+      "learning_rate": 3.1707317073170736e-06,
+      "loss": 1.554,
+      "step": 13
+    },
+    {
+      "epoch": 0.01,
+      "grad_norm": 10.11918265638141,
+      "learning_rate": 3.414634146341464e-06,
+      "loss": 1.5119,
+      "step": 14
+    },
+    {
+      "epoch": 0.01,
+      "grad_norm": 12.356401968625594,
+      "learning_rate": 3.6585365853658537e-06,
+      "loss": 1.6321,
+      "step": 15
+    },
+    {
+      "epoch": 0.01,
+      "grad_norm": 11.132578216935597,
+      "learning_rate": 3.902439024390244e-06,
+      "loss": 1.5888,
+      "step": 16
+    },
+    {
+      "epoch": 0.01,
+      "grad_norm": 12.240345185091753,
+      "learning_rate": 4.146341463414634e-06,
+      "loss": 1.5076,
+      "step": 17
+    },
+    {
+      "epoch": 0.01,
+      "grad_norm": 11.968138590460152,
+      "learning_rate": 4.390243902439025e-06,
+      "loss": 1.3815,
+      "step": 18
+    },
+    {
+      "epoch": 0.01,
+      "grad_norm": 8.169350991545556,
+      "learning_rate": 4.634146341463416e-06,
+      "loss": 1.3101,
+      "step": 19
+    },
+    {
+      "epoch": 0.01,
+      "grad_norm": 9.505811804013986,
+      "learning_rate": 4.8780487804878055e-06,
+      "loss": 1.2818,
+      "step": 20
+    },
+    {
+      "epoch": 0.01,
+      "grad_norm": 5.796640993993368,
+      "learning_rate": 5.121951219512195e-06,
+      "loss": 1.2677,
+      "step": 21
+    },
+    {
+      "epoch": 0.01,
+      "grad_norm": 4.764801673099041,
+      "learning_rate": 5.365853658536586e-06,
+      "loss": 1.2417,
+      "step": 22
+    },
+    {
+      "epoch": 0.01,
+      "grad_norm": 3.8410501167446562,
+      "learning_rate": 5.609756097560977e-06,
+      "loss": 1.2282,
+      "step": 23
+    },
+    {
+      "epoch": 0.01,
+      "grad_norm": 3.2269719491020132,
+      "learning_rate": 5.853658536585366e-06,
+      "loss": 1.1735,
+      "step": 24
+    },
+    {
+      "epoch": 0.01,
+      "grad_norm": 2.7362981827784276,
+      "learning_rate": 6.0975609756097564e-06,
+      "loss": 1.1771,
+      "step": 25
+    },
+    {
+      "epoch": 0.01,
+      "grad_norm": 2.3380480940076693,
+      "learning_rate": 6.341463414634147e-06,
+      "loss": 1.1862,
+      "step": 26
+    },
+    {
+      "epoch": 0.01,
+      "grad_norm": 2.3527882605489214,
+      "learning_rate": 6.585365853658538e-06,
+      "loss": 1.0824,
+      "step": 27
+    },
+    {
+      "epoch": 0.01,
+      "grad_norm": 2.120763742750798,
+      "learning_rate": 6.829268292682928e-06,
+      "loss": 1.0907,
+      "step": 28
+    },
+    {
+      "epoch": 0.01,
+      "grad_norm": 1.7082373070011942,
+      "learning_rate": 7.0731707317073175e-06,
+      "loss": 1.0936,
+      "step": 29
+    },
+    {
+      "epoch": 0.01,
+      "grad_norm": 1.771684587866832,
+      "learning_rate": 7.317073170731707e-06,
+      "loss": 1.232,
+      "step": 30
+    },
+    {
+      "epoch": 0.01,
+      "grad_norm": 1.6792419401200112,
+      "learning_rate": 7.560975609756098e-06,
+      "loss": 1.0939,
+      "step": 31
+    },
+    {
+      "epoch": 0.01,
+      "grad_norm": 1.5870734246620417,
+      "learning_rate": 7.804878048780489e-06,
+      "loss": 1.0344,
+      "step": 32
+    },
+    {
+      "epoch": 0.01,
+      "grad_norm": 1.5321693507507685,
+      "learning_rate": 8.048780487804879e-06,
+      "loss": 1.1247,
+      "step": 33
+    },
+    {
+      "epoch": 0.01,
+      "grad_norm": 1.5384110247679068,
+      "learning_rate": 8.292682926829268e-06,
+      "loss": 1.0561,
+      "step": 34
+    },
+    {
+      "epoch": 0.01,
+      "grad_norm": 1.6803740230388178,
+      "learning_rate": 8.536585365853658e-06,
+      "loss": 1.0912,
+      "step": 35
+    },
+    {
+      "epoch": 0.01,
+      "grad_norm": 1.7687493200257336,
+      "learning_rate": 8.78048780487805e-06,
+      "loss": 1.0656,
+      "step": 36
+    },
+    {
+      "epoch": 0.01,
+      "grad_norm": 1.696255950031143,
+      "learning_rate": 9.02439024390244e-06,
+      "loss": 1.0381,
+      "step": 37
+    },
+    {
+      "epoch": 0.01,
+      "grad_norm": 1.5093882845959572,
+      "learning_rate": 9.268292682926831e-06,
+      "loss": 1.1241,
+      "step": 38
+    },
+    {
+      "epoch": 0.01,
+      "grad_norm": 1.6975518192922392,
+      "learning_rate": 9.51219512195122e-06,
+      "loss": 0.9997,
+      "step": 39
+    },
+    {
+      "epoch": 0.01,
+      "grad_norm": 1.5343961293035455,
+      "learning_rate": 9.756097560975611e-06,
+      "loss": 1.0369,
+      "step": 40
+    },
+    {
+      "epoch": 0.02,
+      "grad_norm": 1.4138966954537764,
+      "learning_rate": 1e-05,
+      "loss": 1.014,
+      "step": 41
+    },
+    {
+      "epoch": 0.02,
+      "grad_norm": 1.176007866432356,
+      "learning_rate": 1.024390243902439e-05,
+      "loss": 1.0021,
+      "step": 42
+    },
+    {
+      "epoch": 0.02,
+      "grad_norm": 1.2381411186803413,
+      "learning_rate": 1.0487804878048782e-05,
+      "loss": 1.1223,
+      "step": 43
+    },
+    {
+      "epoch": 0.02,
+      "grad_norm": 1.320371206343042,
+      "learning_rate": 1.0731707317073172e-05,
+      "loss": 1.1278,
+      "step": 44
+    },
+    {
+      "epoch": 0.02,
+      "grad_norm": 1.169780930584551,
+      "learning_rate": 1.0975609756097562e-05,
+      "loss": 0.9749,
+      "step": 45
+    },
+    {
+      "epoch": 0.02,
+      "grad_norm": 1.2215812178026708,
+      "learning_rate": 1.1219512195121953e-05,
+      "loss": 1.1331,
+      "step": 46
+    },
+    {
+      "epoch": 0.02,
+      "grad_norm": 1.2706297862727212,
+      "learning_rate": 1.1463414634146342e-05,
+      "loss": 1.0683,
+      "step": 47
+    },
+    {
+      "epoch": 0.02,
+      "grad_norm": 1.1432329777320256,
+      "learning_rate": 1.1707317073170731e-05,
+      "loss": 1.1807,
+      "step": 48
+    },
+    {
+      "epoch": 0.02,
+      "grad_norm": 1.2331341664649251,
+      "learning_rate": 1.1951219512195123e-05,
+      "loss": 1.1285,
+      "step": 49
+    },
+    {
+      "epoch": 0.02,
+      "grad_norm": 1.1741530257255526,
+      "learning_rate": 1.2195121951219513e-05,
+      "loss": 1.0448,
+      "step": 50
+    },
+    {
+      "epoch": 0.02,
+      "grad_norm": 1.1161714391272533,
+      "learning_rate": 1.2439024390243903e-05,
+      "loss": 1.0628,
+      "step": 51
+    },
+    {
+      "epoch": 0.02,
+      "grad_norm": 1.1971235536055034,
+      "learning_rate": 1.2682926829268294e-05,
+      "loss": 1.0049,
+      "step": 52
+    },
+    {
+      "epoch": 0.02,
+      "grad_norm": 1.045851128188208,
+      "learning_rate": 1.2926829268292684e-05,
+      "loss": 1.1882,
+      "step": 53
+    },
+    {
+      "epoch": 0.02,
+      "grad_norm": 1.0698067786575107,
+      "learning_rate": 1.3170731707317076e-05,
+      "loss": 0.9463,
+      "step": 54
+    },
+    {
+      "epoch": 0.02,
+      "grad_norm": 1.172591425775864,
+      "learning_rate": 1.3414634146341466e-05,
+      "loss": 1.0636,
+      "step": 55
+    },
+    {
+      "epoch": 0.02,
+      "grad_norm": 1.1106016799757537,
+      "learning_rate": 1.3658536585365855e-05,
+      "loss": 0.9946,
+      "step": 56
+    },
+    {
+      "epoch": 0.02,
+      "grad_norm": 1.086675627063963,
+      "learning_rate": 1.3902439024390244e-05,
+      "loss": 1.1333,
+      "step": 57
+    },
+    {
+      "epoch": 0.02,
+      "grad_norm": 1.1240451271188707,
+      "learning_rate": 1.4146341463414635e-05,
+      "loss": 1.0376,
+      "step": 58
+    },
+    {
+      "epoch": 0.02,
+      "grad_norm": 1.0978604882977046,
+      "learning_rate": 1.4390243902439025e-05,
+      "loss": 1.0286,
+      "step": 59
+    },
+    {
+      "epoch": 0.02,
+      "grad_norm": 1.1371703935435598,
+      "learning_rate": 1.4634146341463415e-05,
+      "loss": 1.0489,
+      "step": 60
+    },
+    {
+      "epoch": 0.02,
+      "grad_norm": 1.081855866398565,
+      "learning_rate": 1.4878048780487806e-05,
+      "loss": 1.0347,
+      "step": 61
+    },
+    {
+      "epoch": 0.02,
+      "grad_norm": 1.1018052800440352,
+      "learning_rate": 1.5121951219512196e-05,
+      "loss": 1.0503,
+      "step": 62
+    },
+    {
+      "epoch": 0.02,
+      "grad_norm": 1.2132015921868264,
+      "learning_rate": 1.5365853658536586e-05,
+      "loss": 0.962,
+      "step": 63
+    },
+    {
+      "epoch": 0.02,
+      "grad_norm": 1.1035822755607994,
+      "learning_rate": 1.5609756097560978e-05,
+      "loss": 1.0749,
+      "step": 64
+    },
+    {
+      "epoch": 0.02,
+      "grad_norm": 1.0944024225848128,
+      "learning_rate": 1.585365853658537e-05,
+      "loss": 1.0126,
+      "step": 65
+    },
+    {
+      "epoch": 0.02,
+      "grad_norm": 1.1408025250647078,
+      "learning_rate": 1.6097560975609757e-05,
+      "loss": 0.9841,
+      "step": 66
+    },
+    {
+      "epoch": 0.02,
+      "grad_norm": 1.0549980708871876,
+      "learning_rate": 1.6341463414634145e-05,
+      "loss": 1.0857,
+      "step": 67
+    },
+    {
+      "epoch": 0.03,
+      "grad_norm": 1.1088465780550794,
+      "learning_rate": 1.6585365853658537e-05,
+      "loss": 1.0189,
+      "step": 68
+    },
+    {
+      "epoch": 0.03,
+      "grad_norm": 1.1030941169394408,
+      "learning_rate": 1.682926829268293e-05,
+      "loss": 0.9477,
+      "step": 69
+    },
+    {
+      "epoch": 0.03,
+      "grad_norm": 1.0906659313859568,
+      "learning_rate": 1.7073170731707317e-05,
+      "loss": 1.1326,
+      "step": 70
+    },
+    {
+      "epoch": 0.03,
+      "grad_norm": 1.095023300471396,
+      "learning_rate": 1.7317073170731708e-05,
+      "loss": 1.0073,
+      "step": 71
+    },
+    {
+      "epoch": 0.03,
+      "grad_norm": 1.0603304396820843,
+      "learning_rate": 1.75609756097561e-05,
+      "loss": 1.0442,
+      "step": 72
+    },
+    {
+      "epoch": 0.03,
+      "grad_norm": 1.0658804001104982,
+      "learning_rate": 1.7804878048780488e-05,
+      "loss": 1.0266,
+      "step": 73
+    },
+    {
+      "epoch": 0.03,
+      "grad_norm": 1.0581434225632516,
+      "learning_rate": 1.804878048780488e-05,
+      "loss": 0.9944,
+      "step": 74
+    },
+    {
+      "epoch": 0.03,
+      "grad_norm": 1.0325391690636108,
+      "learning_rate": 1.829268292682927e-05,
+      "loss": 0.9795,
+      "step": 75
+    },
+    {
+      "epoch": 0.03,
+      "grad_norm": 1.0137734048869766,
+      "learning_rate": 1.8536585365853663e-05,
+      "loss": 0.9792,
+      "step": 76
+    },
+    {
+      "epoch": 0.03,
+      "grad_norm": 1.0835263870224308,
+      "learning_rate": 1.878048780487805e-05,
+      "loss": 1.0396,
+      "step": 77
+    },
+    {
+      "epoch": 0.03,
+      "grad_norm": 0.9724143527075428,
+      "learning_rate": 1.902439024390244e-05,
+      "loss": 0.9221,
+      "step": 78
+    },
+    {
+      "epoch": 0.03,
+      "grad_norm": 0.9990983369485978,
+      "learning_rate": 1.926829268292683e-05,
+      "loss": 0.9228,
+      "step": 79
+    },
+    {
+      "epoch": 0.03,
+      "grad_norm": 1.1063583109588462,
+      "learning_rate": 1.9512195121951222e-05,
+      "loss": 1.0854,
+      "step": 80
+    },
+    {
+      "epoch": 0.03,
+      "grad_norm": 1.0512858412634352,
+      "learning_rate": 1.975609756097561e-05,
+      "loss": 1.0148,
+      "step": 81
+    },
+    {
+      "epoch": 0.03,
+      "grad_norm": 1.0536440636090694,
+      "learning_rate": 2e-05,
+      "loss": 1.0287,
+      "step": 82
+    },
+    {
+      "epoch": 0.03,
+      "grad_norm": 1.0803096330003004,
+      "learning_rate": 1.999999287101006e-05,
+      "loss": 1.1047,
+      "step": 83
+    },
+    {
+      "epoch": 0.03,
+      "grad_norm": 1.0547983394060736,
+      "learning_rate": 1.99999714840504e-05,
+      "loss": 0.9816,
+      "step": 84
+    },
+    {
+      "epoch": 0.03,
+      "grad_norm": 0.9807015967414549,
+      "learning_rate": 1.9999935839151513e-05,
+      "loss": 0.9548,
+      "step": 85
+    },
+    {
+      "epoch": 0.03,
+      "grad_norm": 1.0711912458397403,
+      "learning_rate": 1.999988593636423e-05,
+      "loss": 0.9682,
+      "step": 86
+    },
+    {
+      "epoch": 0.03,
+      "grad_norm": 0.9722241890066263,
+      "learning_rate": 1.999982177575969e-05,
+      "loss": 1.0405,
+      "step": 87
+    },
+    {
+      "epoch": 0.03,
+      "grad_norm": 1.028798727103559,
+      "learning_rate": 1.999974335742938e-05,
+      "loss": 1.1811,
+      "step": 88
+    },
+    {
+      "epoch": 0.03,
+      "grad_norm": 1.1021829611351164,
+      "learning_rate": 1.999965068148511e-05,
+      "loss": 0.9051,
+      "step": 89
+    },
+    {
+      "epoch": 0.03,
+      "grad_norm": 1.0334726008966082,
+      "learning_rate": 1.9999543748059012e-05,
+      "loss": 1.0484,
+      "step": 90
+    },
+    {
+      "epoch": 0.03,
+      "grad_norm": 1.066221360284238,
+      "learning_rate": 1.9999422557303553e-05,
+      "loss": 1.0307,
+      "step": 91
+    },
+    {
+      "epoch": 0.03,
+      "grad_norm": 1.054488043267414,
+      "learning_rate": 1.999928710939153e-05,
+      "loss": 1.1487,
+      "step": 92
+    },
+    {
+      "epoch": 0.03,
+      "grad_norm": 1.1079480240688568,
+      "learning_rate": 1.9999137404516062e-05,
+      "loss": 0.9761,
+      "step": 93
+    },
+    {
+      "epoch": 0.03,
+      "grad_norm": 1.070030561141793,
+      "learning_rate": 1.99989734428906e-05,
+      "loss": 1.0988,
+      "step": 94
+    },
+    {
+      "epoch": 0.04,
+      "grad_norm": 1.0564078940337887,
+      "learning_rate": 1.9998795224748916e-05,
+      "loss": 1.0195,
+      "step": 95
+    },
+    {
+      "epoch": 0.04,
+      "grad_norm": 1.1531641916007334,
+      "learning_rate": 1.9998602750345113e-05,
+      "loss": 1.0807,
+      "step": 96
+    },
+    {
+      "epoch": 0.04,
+      "grad_norm": 1.116837722786228,
+      "learning_rate": 1.9998396019953627e-05,
+      "loss": 0.9621,
+      "step": 97
+    },
+    {
+      "epoch": 0.04,
+      "grad_norm": 1.0677298051233879,
+      "learning_rate": 1.9998175033869205e-05,
+      "loss": 0.9151,
+      "step": 98
+    },
+    {
+      "epoch": 0.04,
+      "grad_norm": 1.1242469864164173,
+      "learning_rate": 1.9997939792406936e-05,
+      "loss": 0.9819,
+      "step": 99
+    },
+    {
+      "epoch": 0.04,
+      "grad_norm": 1.16872009332317,
+      "learning_rate": 1.9997690295902225e-05,
+      "loss": 1.1617,
+      "step": 100
+    }
+  ],
+  "logging_steps": 1.0,
+  "max_steps": 2713,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 1,
+  "save_steps": 100,
+  "total_flos": 134491228471296.0,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}

training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:738f0ea8e09d651424e0acfd276ca112425fe1fa15d2370f5ef6bafeda192889
+size 6968

zero_to_fp32.py

@@ -0,0 +1,604 @@
+#!/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 . pytorch_model.bin
+
+import argparse
+import torch
+import glob
+import math
+import os
+import re
+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 param_shapes.items():
+
+        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_file, 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_file``: path to the pytorch fp32 state_dict output file (e.g. path/pytorch_model.bin)
+        - ``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
+    """
+
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag, exclude_frozen_parameters)
+    print(f"Saving fp32 state dict to {output_file}")
+    torch.save(state_dict, output_file)
+
+
+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_file",
+        type=str,
+        help="path to the pytorch fp32 state_dict output file (e.g. path/checkpoint-12/pytorch_model.bin)")
+    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_file,
+                                               tag=args.tag,
+                                               exclude_frozen_parameters=args.exclude_frozen_parameters)