Upload 21 files

.gitattributes

@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+tokenizer.json filter=lfs diff=lfs merge=lfs -text

config.json

@@ -0,0 +1,28 @@
+{
+  "_name_or_path": "output_512_5_128_512_2024-09-14_00_34_39/checkpoint-1800",
+  "architectures": [
+    "XLMRobertaModel"
+  ],
+  "attention_probs_dropout_prob": 0.1,
+  "bos_token_id": 0,
+  "classifier_dropout": null,
+  "eos_token_id": 2,
+  "hidden_act": "gelu",
+  "hidden_dropout_prob": 0.1,
+  "hidden_size": 1024,
+  "initializer_range": 0.02,
+  "intermediate_size": 4096,
+  "layer_norm_eps": 1e-05,
+  "max_position_embeddings": 8194,
+  "model_type": "xlm-roberta",
+  "num_attention_heads": 16,
+  "num_hidden_layers": 24,
+  "output_past": true,
+  "pad_token_id": 1,
+  "position_embedding_type": "absolute",
+  "torch_dtype": "float32",
+  "transformers_version": "4.44.2",
+  "type_vocab_size": 1,
+  "use_cache": true,
+  "vocab_size": 250002
+}

config_sentence_transformers.json

@@ -0,0 +1,10 @@
+{
+  "__version__": {
+    "sentence_transformers": "3.0.1",
+    "transformers": "4.44.2",
+    "pytorch": "2.4.0+cu121"
+  },
+  "prompts": {},
+  "default_prompt_name": null,
+  "similarity_fn_name": null
+}

latest

@@ -0,0 +1 @@
+global_step1800

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dc2a3500e68193e6a4b729099f9948784494958e3b47d445d1dc5b77e292e573
+size 2271064456

modules.json

@@ -0,0 +1,20 @@
+[
+  {
+    "idx": 0,
+    "name": "0",
+    "path": "",
+    "type": "sentence_transformers.models.Transformer"
+  },
+  {
+    "idx": 1,
+    "name": "1",
+    "path": "1_Pooling",
+    "type": "sentence_transformers.models.Pooling"
+  },
+  {
+    "idx": 2,
+    "name": "2",
+    "path": "2_Normalize",
+    "type": "sentence_transformers.models.Normalize"
+  }
+]

rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b33a678e34c66e4da53e59cbd8dd47a71ae2d329209a412fb9c728f82d14ad5c
+size 15920

rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3a4f0546d7d1507f688beed3b4cea3bf5aeab8be5a5873319727b234f4b99591
+size 15984

rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5be152bbf9a1384690b2e53d0b9225070b15ff4253005a0a00d1d84fda18a657
+size 15920

rng_state_3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:de500c8dc8790507d03510ab8a988161ac6b457b77ed3e7875f16cfc409c5d0c
+size 15920

rng_state_4.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5e77a3fa840d0c94f0fc7592dfad9acd21eef0226b2aa18b01e64db3d365c8fc
+size 15920

rng_state_5.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:df35156e206ea5f264395ab1c05e517ee4ed00b5349f936e8262bbfc80a3a21c
+size 15920

rng_state_6.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:afc5a40fb09a92400220ff44760e3a2bf7c1dedfc915d2b27adcef332442bbdc
+size 15984

rng_state_7.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2b50067d819f8d4ed5350b6579df557cb626480d0f9d23fe4ae343de4d1c3ba8
+size 15920

sentence_bert_config.json

@@ -0,0 +1,4 @@
+{
+  "max_seq_length": 8192,
+  "do_lower_case": false
+}

sentencepiece.bpe.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cfc8146abe2a0488e9e2a0c56de7952f7c11ab059eca145a0a727afce0db2865
+size 5069051

special_tokens_map.json

@@ -0,0 +1,51 @@
+{
+  "bos_token": {
+    "content": "<s>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "cls_token": {
+    "content": "<s>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "eos_token": {
+    "content": "</s>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "mask_token": {
+    "content": "<mask>",
+    "lstrip": true,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": {
+    "content": "<pad>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "sep_token": {
+    "content": "</s>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "unk_token": {
+    "content": "<unk>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

tokenizer.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b74659c780d49afad7a7b9799868f75cbd3014fb6c34956e85a793028d38094a
+size 17098251

tokenizer_config.json

@@ -0,0 +1,55 @@
+{
+  "added_tokens_decoder": {
+    "0": {
+      "content": "<s>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "1": {
+      "content": "<pad>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "2": {
+      "content": "</s>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "3": {
+      "content": "<unk>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "250001": {
+      "content": "<mask>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    }
+  },
+  "bos_token": "<s>",
+  "clean_up_tokenization_spaces": true,
+  "cls_token": "<s>",
+  "eos_token": "</s>",
+  "mask_token": "<mask>",
+  "model_max_length": 8192,
+  "pad_token": "<pad>",
+  "sep_token": "</s>",
+  "sp_model_kwargs": {},
+  "tokenizer_class": "XLMRobertaTokenizer",
+  "unk_token": "<unk>"
+}

trainer_state.json

@@ -0,0 +1,1293 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.098106712564544,
+  "eval_steps": 500,
+  "global_step": 1800,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.01721170395869191,
+      "grad_norm": 0.29955029487609863,
+      "learning_rate": 2.0293089116901574e-06,
+      "loss": 0.6322,
+      "step": 10
+    },
+    {
+      "epoch": 0.03442340791738382,
+      "grad_norm": 0.06169761344790459,
+      "learning_rate": 2.6401917645771237e-06,
+      "loss": 0.4697,
+      "step": 20
+    },
+    {
+      "epoch": 0.05163511187607573,
+      "grad_norm": 0.051926977932453156,
+      "learning_rate": 2.9975353258495578e-06,
+      "loss": 0.5617,
+      "step": 30
+    },
+    {
+      "epoch": 0.06884681583476764,
+      "grad_norm": 0.07096195966005325,
+      "learning_rate": 3.25107461746409e-06,
+      "loss": 0.4301,
+      "step": 40
+    },
+    {
+      "epoch": 0.08605851979345955,
+      "grad_norm": 0.06899057328701019,
+      "learning_rate": 3.4477349704933476e-06,
+      "loss": 0.4905,
+      "step": 50
+    },
+    {
+      "epoch": 0.10327022375215146,
+      "grad_norm": 0.08537387102842331,
+      "learning_rate": 3.6084181787365237e-06,
+      "loss": 0.4551,
+      "step": 60
+    },
+    {
+      "epoch": 0.12048192771084337,
+      "grad_norm": 0.049780745059251785,
+      "learning_rate": 3.7442738955429737e-06,
+      "loss": 0.4058,
+      "step": 70
+    },
+    {
+      "epoch": 0.13769363166953527,
+      "grad_norm": 0.04421038553118706,
+      "learning_rate": 3.861957470351056e-06,
+      "loss": 0.6748,
+      "step": 80
+    },
+    {
+      "epoch": 0.1549053356282272,
+      "grad_norm": 1.9084473848342896,
+      "learning_rate": 3.965761740008958e-06,
+      "loss": 0.8719,
+      "step": 90
+    },
+    {
+      "epoch": 0.1721170395869191,
+      "grad_norm": 0.08046019077301025,
+      "learning_rate": 4.058617823380315e-06,
+      "loss": 0.4635,
+      "step": 100
+    },
+    {
+      "epoch": 0.18932874354561102,
+      "grad_norm": 0.21439455449581146,
+      "learning_rate": 4.142616368250685e-06,
+      "loss": 0.928,
+      "step": 110
+    },
+    {
+      "epoch": 0.20654044750430292,
+      "grad_norm": 0.06055545434355736,
+      "learning_rate": 4.21930103162349e-06,
+      "loss": 0.3721,
+      "step": 120
+    },
+    {
+      "epoch": 0.22375215146299485,
+      "grad_norm": 0.08670035004615784,
+      "learning_rate": 4.289844083644429e-06,
+      "loss": 0.7536,
+      "step": 130
+    },
+    {
+      "epoch": 0.24096385542168675,
+      "grad_norm": 0.06118405610322952,
+      "learning_rate": 4.355156748429939e-06,
+      "loss": 0.9829,
+      "step": 140
+    },
+    {
+      "epoch": 0.25817555938037867,
+      "grad_norm": 0.04853704199194908,
+      "learning_rate": 4.415961384652748e-06,
+      "loss": 0.4444,
+      "step": 150
+    },
+    {
+      "epoch": 0.27538726333907054,
+      "grad_norm": 0.03537767753005028,
+      "learning_rate": 4.472840323238023e-06,
+      "loss": 0.5064,
+      "step": 160
+    },
+    {
+      "epoch": 0.29259896729776247,
+      "grad_norm": 0.06154410541057587,
+      "learning_rate": 4.52626987322263e-06,
+      "loss": 0.5456,
+      "step": 170
+    },
+    {
+      "epoch": 0.3098106712564544,
+      "grad_norm": 0.052560485899448395,
+      "learning_rate": 4.576644592895925e-06,
+      "loss": 0.5106,
+      "step": 180
+    },
+    {
+      "epoch": 0.3270223752151463,
+      "grad_norm": 0.04913010448217392,
+      "learning_rate": 4.6242949899596115e-06,
+      "loss": 0.4026,
+      "step": 190
+    },
+    {
+      "epoch": 0.3442340791738382,
+      "grad_norm": 0.07974158972501755,
+      "learning_rate": 4.66950067626728e-06,
+      "loss": 0.4828,
+      "step": 200
+    },
+    {
+      "epoch": 0.3614457831325301,
+      "grad_norm": 0.03538183122873306,
+      "learning_rate": 4.712500309702374e-06,
+      "loss": 0.3549,
+      "step": 210
+    },
+    {
+      "epoch": 0.37865748709122204,
+      "grad_norm": 0.21638496220111847,
+      "learning_rate": 4.753499221137652e-06,
+      "loss": 0.4912,
+      "step": 220
+    },
+    {
+      "epoch": 0.3958691910499139,
+      "grad_norm": 0.03895362466573715,
+      "learning_rate": 4.792675344617211e-06,
+      "loss": 0.3846,
+      "step": 230
+    },
+    {
+      "epoch": 0.41308089500860584,
+      "grad_norm": 0.03565879911184311,
+      "learning_rate": 4.830183884510456e-06,
+      "loss": 0.8434,
+      "step": 240
+    },
+    {
+      "epoch": 0.43029259896729777,
+      "grad_norm": 0.03526683151721954,
+      "learning_rate": 4.866161029296539e-06,
+      "loss": 0.3603,
+      "step": 250
+    },
+    {
+      "epoch": 0.4475043029259897,
+      "grad_norm": 0.064102903008461,
+      "learning_rate": 4.900726936531396e-06,
+      "loss": 0.5178,
+      "step": 260
+    },
+    {
+      "epoch": 0.46471600688468157,
+      "grad_norm": 0.06982860714197159,
+      "learning_rate": 4.9339881541683585e-06,
+      "loss": 0.3712,
+      "step": 270
+    },
+    {
+      "epoch": 0.4819277108433735,
+      "grad_norm": 0.0654272809624672,
+      "learning_rate": 4.966039601316906e-06,
+      "loss": 0.9119,
+      "step": 280
+    },
+    {
+      "epoch": 0.4991394148020654,
+      "grad_norm": 0.04955059662461281,
+      "learning_rate": 4.9969662012643525e-06,
+      "loss": 0.3874,
+      "step": 290
+    },
+    {
+      "epoch": 0.5163511187607573,
+      "grad_norm": 1.0234352350234985,
+      "learning_rate": 4.984697781178272e-06,
+      "loss": 0.8952,
+      "step": 300
+    },
+    {
+      "epoch": 0.5335628227194492,
+      "grad_norm": 0.03769606724381447,
+      "learning_rate": 4.96557000765111e-06,
+      "loss": 0.3347,
+      "step": 310
+    },
+    {
+      "epoch": 0.5507745266781411,
+      "grad_norm": 0.11739111691713333,
+      "learning_rate": 4.946442234123948e-06,
+      "loss": 0.3677,
+      "step": 320
+    },
+    {
+      "epoch": 0.5679862306368331,
+      "grad_norm": 0.04959660395979881,
+      "learning_rate": 4.927314460596787e-06,
+      "loss": 1.1762,
+      "step": 330
+    },
+    {
+      "epoch": 0.5851979345955249,
+      "grad_norm": 0.1042531356215477,
+      "learning_rate": 4.908186687069626e-06,
+      "loss": 0.4252,
+      "step": 340
+    },
+    {
+      "epoch": 0.6024096385542169,
+      "grad_norm": 0.05064910277724266,
+      "learning_rate": 4.889058913542464e-06,
+      "loss": 0.3836,
+      "step": 350
+    },
+    {
+      "epoch": 0.6196213425129088,
+      "grad_norm": 0.0689607635140419,
+      "learning_rate": 4.869931140015303e-06,
+      "loss": 0.7539,
+      "step": 360
+    },
+    {
+      "epoch": 0.6368330464716007,
+      "grad_norm": 0.23462702333927155,
+      "learning_rate": 4.850803366488141e-06,
+      "loss": 0.8236,
+      "step": 370
+    },
+    {
+      "epoch": 0.6540447504302926,
+      "grad_norm": 0.11018137633800507,
+      "learning_rate": 4.83167559296098e-06,
+      "loss": 0.4839,
+      "step": 380
+    },
+    {
+      "epoch": 0.6712564543889845,
+      "grad_norm": 0.0751522108912468,
+      "learning_rate": 4.812547819433818e-06,
+      "loss": 0.5791,
+      "step": 390
+    },
+    {
+      "epoch": 0.6884681583476764,
+      "grad_norm": 0.17227555811405182,
+      "learning_rate": 4.793420045906657e-06,
+      "loss": 0.7993,
+      "step": 400
+    },
+    {
+      "epoch": 0.7056798623063684,
+      "grad_norm": 0.0664035975933075,
+      "learning_rate": 4.7742922723794954e-06,
+      "loss": 0.387,
+      "step": 410
+    },
+    {
+      "epoch": 0.7228915662650602,
+      "grad_norm": 0.04762504622340202,
+      "learning_rate": 4.755164498852334e-06,
+      "loss": 0.5436,
+      "step": 420
+    },
+    {
+      "epoch": 0.7401032702237521,
+      "grad_norm": 0.03658389300107956,
+      "learning_rate": 4.736036725325173e-06,
+      "loss": 0.6715,
+      "step": 430
+    },
+    {
+      "epoch": 0.7573149741824441,
+      "grad_norm": 0.03955502808094025,
+      "learning_rate": 4.716908951798011e-06,
+      "loss": 0.4902,
+      "step": 440
+    },
+    {
+      "epoch": 0.774526678141136,
+      "grad_norm": 0.05926811322569847,
+      "learning_rate": 4.69778117827085e-06,
+      "loss": 0.7329,
+      "step": 450
+    },
+    {
+      "epoch": 0.7917383820998278,
+      "grad_norm": 0.26404136419296265,
+      "learning_rate": 4.678653404743688e-06,
+      "loss": 0.5748,
+      "step": 460
+    },
+    {
+      "epoch": 0.8089500860585198,
+      "grad_norm": 0.07195431739091873,
+      "learning_rate": 4.6595256312165265e-06,
+      "loss": 0.5501,
+      "step": 470
+    },
+    {
+      "epoch": 0.8261617900172117,
+      "grad_norm": 0.0486939400434494,
+      "learning_rate": 4.640397857689365e-06,
+      "loss": 0.4527,
+      "step": 480
+    },
+    {
+      "epoch": 0.8433734939759037,
+      "grad_norm": 0.05488497018814087,
+      "learning_rate": 4.621270084162204e-06,
+      "loss": 0.8637,
+      "step": 490
+    },
+    {
+      "epoch": 0.8605851979345955,
+      "grad_norm": 0.045418575406074524,
+      "learning_rate": 4.6021423106350425e-06,
+      "loss": 0.437,
+      "step": 500
+    },
+    {
+      "epoch": 0.8777969018932874,
+      "grad_norm": 0.04055708646774292,
+      "learning_rate": 4.583014537107881e-06,
+      "loss": 0.6466,
+      "step": 510
+    },
+    {
+      "epoch": 0.8950086058519794,
+      "grad_norm": 0.03856475651264191,
+      "learning_rate": 4.563886763580719e-06,
+      "loss": 0.669,
+      "step": 520
+    },
+    {
+      "epoch": 0.9122203098106713,
+      "grad_norm": 0.035741958767175674,
+      "learning_rate": 4.5447589900535585e-06,
+      "loss": 0.3615,
+      "step": 530
+    },
+    {
+      "epoch": 0.9294320137693631,
+      "grad_norm": 0.04278489947319031,
+      "learning_rate": 4.525631216526396e-06,
+      "loss": 0.3849,
+      "step": 540
+    },
+    {
+      "epoch": 0.9466437177280551,
+      "grad_norm": 0.031775712966918945,
+      "learning_rate": 4.506503442999236e-06,
+      "loss": 0.6446,
+      "step": 550
+    },
+    {
+      "epoch": 0.963855421686747,
+      "grad_norm": 0.19989252090454102,
+      "learning_rate": 4.487375669472074e-06,
+      "loss": 0.6668,
+      "step": 560
+    },
+    {
+      "epoch": 0.9810671256454389,
+      "grad_norm": 0.04056662693619728,
+      "learning_rate": 4.468247895944912e-06,
+      "loss": 0.4243,
+      "step": 570
+    },
+    {
+      "epoch": 0.9982788296041308,
+      "grad_norm": 0.06392610818147659,
+      "learning_rate": 4.449120122417751e-06,
+      "loss": 0.3431,
+      "step": 580
+    },
+    {
+      "epoch": 1.0154905335628228,
+      "grad_norm": 0.03935154527425766,
+      "learning_rate": 4.42999234889059e-06,
+      "loss": 0.5167,
+      "step": 590
+    },
+    {
+      "epoch": 1.0327022375215147,
+      "grad_norm": 0.05566889047622681,
+      "learning_rate": 4.410864575363428e-06,
+      "loss": 0.4372,
+      "step": 600
+    },
+    {
+      "epoch": 1.0499139414802066,
+      "grad_norm": 0.07127536088228226,
+      "learning_rate": 4.391736801836267e-06,
+      "loss": 1.4152,
+      "step": 610
+    },
+    {
+      "epoch": 1.0671256454388984,
+      "grad_norm": 0.04618392139673233,
+      "learning_rate": 4.372609028309105e-06,
+      "loss": 0.601,
+      "step": 620
+    },
+    {
+      "epoch": 1.0843373493975903,
+      "grad_norm": 0.04588570445775986,
+      "learning_rate": 4.3534812547819434e-06,
+      "loss": 0.4723,
+      "step": 630
+    },
+    {
+      "epoch": 1.1015490533562822,
+      "grad_norm": 0.03991321101784706,
+      "learning_rate": 4.334353481254782e-06,
+      "loss": 0.4807,
+      "step": 640
+    },
+    {
+      "epoch": 1.1187607573149743,
+      "grad_norm": 0.2501582205295563,
+      "learning_rate": 4.315225707727621e-06,
+      "loss": 0.8098,
+      "step": 650
+    },
+    {
+      "epoch": 1.1359724612736661,
+      "grad_norm": 0.042163778096437454,
+      "learning_rate": 4.296097934200459e-06,
+      "loss": 0.4158,
+      "step": 660
+    },
+    {
+      "epoch": 1.153184165232358,
+      "grad_norm": 0.04054609313607216,
+      "learning_rate": 4.276970160673298e-06,
+      "loss": 0.3728,
+      "step": 670
+    },
+    {
+      "epoch": 1.1703958691910499,
+      "grad_norm": 0.0925000011920929,
+      "learning_rate": 4.257842387146137e-06,
+      "loss": 0.4251,
+      "step": 680
+    },
+    {
+      "epoch": 1.1876075731497417,
+      "grad_norm": 0.06017041206359863,
+      "learning_rate": 4.2387146136189745e-06,
+      "loss": 0.4782,
+      "step": 690
+    },
+    {
+      "epoch": 1.2048192771084336,
+      "grad_norm": 0.040517594665288925,
+      "learning_rate": 4.219586840091814e-06,
+      "loss": 0.4354,
+      "step": 700
+    },
+    {
+      "epoch": 1.2220309810671257,
+      "grad_norm": 0.04731125384569168,
+      "learning_rate": 4.200459066564652e-06,
+      "loss": 0.4969,
+      "step": 710
+    },
+    {
+      "epoch": 1.2392426850258176,
+      "grad_norm": 0.050880610942840576,
+      "learning_rate": 4.1813312930374905e-06,
+      "loss": 0.492,
+      "step": 720
+    },
+    {
+      "epoch": 1.2564543889845095,
+      "grad_norm": 0.04548948258161545,
+      "learning_rate": 4.162203519510329e-06,
+      "loss": 0.3914,
+      "step": 730
+    },
+    {
+      "epoch": 1.2736660929432013,
+      "grad_norm": 0.03825736418366432,
+      "learning_rate": 4.143075745983168e-06,
+      "loss": 0.3921,
+      "step": 740
+    },
+    {
+      "epoch": 1.2908777969018934,
+      "grad_norm": 0.046227287501096725,
+      "learning_rate": 4.1239479724560065e-06,
+      "loss": 0.4632,
+      "step": 750
+    },
+    {
+      "epoch": 1.3080895008605853,
+      "grad_norm": 0.04002716392278671,
+      "learning_rate": 4.104820198928845e-06,
+      "loss": 0.7436,
+      "step": 760
+    },
+    {
+      "epoch": 1.3253012048192772,
+      "grad_norm": 0.04381329566240311,
+      "learning_rate": 4.085692425401683e-06,
+      "loss": 0.5388,
+      "step": 770
+    },
+    {
+      "epoch": 1.342512908777969,
+      "grad_norm": 0.09227538853883743,
+      "learning_rate": 4.0665646518745225e-06,
+      "loss": 0.7008,
+      "step": 780
+    },
+    {
+      "epoch": 1.359724612736661,
+      "grad_norm": 0.0453125424683094,
+      "learning_rate": 4.04743687834736e-06,
+      "loss": 0.4813,
+      "step": 790
+    },
+    {
+      "epoch": 1.3769363166953528,
+      "grad_norm": 0.20484060049057007,
+      "learning_rate": 4.0283091048202e-06,
+      "loss": 0.6594,
+      "step": 800
+    },
+    {
+      "epoch": 1.3941480206540446,
+      "grad_norm": 0.05485668033361435,
+      "learning_rate": 4.009181331293038e-06,
+      "loss": 0.6538,
+      "step": 810
+    },
+    {
+      "epoch": 1.4113597246127367,
+      "grad_norm": 0.04452645406126976,
+      "learning_rate": 3.990053557765876e-06,
+      "loss": 0.3713,
+      "step": 820
+    },
+    {
+      "epoch": 1.4285714285714286,
+      "grad_norm": 0.03632510080933571,
+      "learning_rate": 3.970925784238715e-06,
+      "loss": 0.3395,
+      "step": 830
+    },
+    {
+      "epoch": 1.4457831325301205,
+      "grad_norm": 0.0884113535284996,
+      "learning_rate": 3.951798010711554e-06,
+      "loss": 0.3602,
+      "step": 840
+    },
+    {
+      "epoch": 1.4629948364888123,
+      "grad_norm": 0.1275469958782196,
+      "learning_rate": 3.932670237184392e-06,
+      "loss": 0.4533,
+      "step": 850
+    },
+    {
+      "epoch": 1.4802065404475044,
+      "grad_norm": 0.03843805938959122,
+      "learning_rate": 3.913542463657231e-06,
+      "loss": 0.7519,
+      "step": 860
+    },
+    {
+      "epoch": 1.4974182444061963,
+      "grad_norm": 0.03635178506374359,
+      "learning_rate": 3.89441469013007e-06,
+      "loss": 0.388,
+      "step": 870
+    },
+    {
+      "epoch": 1.5146299483648882,
+      "grad_norm": 0.039031002670526505,
+      "learning_rate": 3.875286916602907e-06,
+      "loss": 0.4425,
+      "step": 880
+    },
+    {
+      "epoch": 1.53184165232358,
+      "grad_norm": 0.04110798239707947,
+      "learning_rate": 3.856159143075746e-06,
+      "loss": 0.4095,
+      "step": 890
+    },
+    {
+      "epoch": 1.549053356282272,
+      "grad_norm": 0.04002736508846283,
+      "learning_rate": 3.837031369548585e-06,
+      "loss": 0.6104,
+      "step": 900
+    },
+    {
+      "epoch": 1.5662650602409638,
+      "grad_norm": 0.03314425051212311,
+      "learning_rate": 3.817903596021423e-06,
+      "loss": 0.5594,
+      "step": 910
+    },
+    {
+      "epoch": 1.5834767641996557,
+      "grad_norm": 0.03947990760207176,
+      "learning_rate": 3.798775822494262e-06,
+      "loss": 0.4931,
+      "step": 920
+    },
+    {
+      "epoch": 1.6006884681583475,
+      "grad_norm": 0.05939627066254616,
+      "learning_rate": 3.7796480489671007e-06,
+      "loss": 0.5127,
+      "step": 930
+    },
+    {
+      "epoch": 1.6179001721170396,
+      "grad_norm": 0.03439631685614586,
+      "learning_rate": 3.760520275439939e-06,
+      "loss": 0.4139,
+      "step": 940
+    },
+    {
+      "epoch": 1.6351118760757315,
+      "grad_norm": 0.06566853076219559,
+      "learning_rate": 3.7413925019127776e-06,
+      "loss": 0.6641,
+      "step": 950
+    },
+    {
+      "epoch": 1.6523235800344234,
+      "grad_norm": 0.06731946766376495,
+      "learning_rate": 3.7222647283856163e-06,
+      "loss": 0.6865,
+      "step": 960
+    },
+    {
+      "epoch": 1.6695352839931155,
+      "grad_norm": 0.03529343381524086,
+      "learning_rate": 3.703136954858455e-06,
+      "loss": 0.6395,
+      "step": 970
+    },
+    {
+      "epoch": 1.6867469879518073,
+      "grad_norm": 0.09028229117393494,
+      "learning_rate": 3.684009181331293e-06,
+      "loss": 0.774,
+      "step": 980
+    },
+    {
+      "epoch": 1.7039586919104992,
+      "grad_norm": 0.04828124865889549,
+      "learning_rate": 3.664881407804132e-06,
+      "loss": 0.4953,
+      "step": 990
+    },
+    {
+      "epoch": 1.721170395869191,
+      "grad_norm": 0.050330750644207,
+      "learning_rate": 3.6457536342769705e-06,
+      "loss": 0.6435,
+      "step": 1000
+    },
+    {
+      "epoch": 1.738382099827883,
+      "grad_norm": 0.03781217709183693,
+      "learning_rate": 3.6266258607498087e-06,
+      "loss": 0.4538,
+      "step": 1010
+    },
+    {
+      "epoch": 1.7555938037865748,
+      "grad_norm": 0.053586967289447784,
+      "learning_rate": 3.607498087222648e-06,
+      "loss": 0.384,
+      "step": 1020
+    },
+    {
+      "epoch": 1.7728055077452667,
+      "grad_norm": 0.04280597344040871,
+      "learning_rate": 3.588370313695486e-06,
+      "loss": 0.385,
+      "step": 1030
+    },
+    {
+      "epoch": 1.7900172117039586,
+      "grad_norm": 0.05530484393239021,
+      "learning_rate": 3.5692425401683243e-06,
+      "loss": 0.732,
+      "step": 1040
+    },
+    {
+      "epoch": 1.8072289156626506,
+      "grad_norm": 0.05707624554634094,
+      "learning_rate": 3.5501147666411634e-06,
+      "loss": 0.4075,
+      "step": 1050
+    },
+    {
+      "epoch": 1.8244406196213425,
+      "grad_norm": 0.07795403897762299,
+      "learning_rate": 3.5309869931140016e-06,
+      "loss": 1.0486,
+      "step": 1060
+    },
+    {
+      "epoch": 1.8416523235800344,
+      "grad_norm": 0.08253274112939835,
+      "learning_rate": 3.5118592195868407e-06,
+      "loss": 0.7014,
+      "step": 1070
+    },
+    {
+      "epoch": 1.8588640275387265,
+      "grad_norm": 0.037665221840143204,
+      "learning_rate": 3.492731446059679e-06,
+      "loss": 0.5129,
+      "step": 1080
+    },
+    {
+      "epoch": 1.8760757314974184,
+      "grad_norm": 0.08074070513248444,
+      "learning_rate": 3.473603672532517e-06,
+      "loss": 0.6965,
+      "step": 1090
+    },
+    {
+      "epoch": 1.8932874354561102,
+      "grad_norm": 0.053863946348428726,
+      "learning_rate": 3.4544758990053563e-06,
+      "loss": 0.3608,
+      "step": 1100
+    },
+    {
+      "epoch": 1.910499139414802,
+      "grad_norm": 0.03980562463402748,
+      "learning_rate": 3.4353481254781945e-06,
+      "loss": 0.3408,
+      "step": 1110
+    },
+    {
+      "epoch": 1.927710843373494,
+      "grad_norm": 0.03091476857662201,
+      "learning_rate": 3.4162203519510336e-06,
+      "loss": 0.4147,
+      "step": 1120
+    },
+    {
+      "epoch": 1.9449225473321858,
+      "grad_norm": 0.05423520505428314,
+      "learning_rate": 3.399005355776588e-06,
+      "loss": 0.501,
+      "step": 1130
+    },
+    {
+      "epoch": 1.9621342512908777,
+      "grad_norm": 0.056222882121801376,
+      "learning_rate": 3.379877582249426e-06,
+      "loss": 0.6646,
+      "step": 1140
+    },
+    {
+      "epoch": 1.9793459552495696,
+      "grad_norm": 0.04780727997422218,
+      "learning_rate": 3.360749808722265e-06,
+      "loss": 0.4433,
+      "step": 1150
+    },
+    {
+      "epoch": 1.9965576592082617,
+      "grad_norm": 0.0465485118329525,
+      "learning_rate": 3.3416220351951034e-06,
+      "loss": 0.4117,
+      "step": 1160
+    },
+    {
+      "epoch": 2.0137693631669533,
+      "grad_norm": 0.038410015404224396,
+      "learning_rate": 3.3224942616679424e-06,
+      "loss": 0.9719,
+      "step": 1170
+    },
+    {
+      "epoch": 2.0309810671256456,
+      "grad_norm": 0.03839205205440521,
+      "learning_rate": 3.3033664881407807e-06,
+      "loss": 0.5383,
+      "step": 1180
+    },
+    {
+      "epoch": 2.0481927710843375,
+      "grad_norm": 0.05250284820795059,
+      "learning_rate": 3.284238714613619e-06,
+      "loss": 0.5573,
+      "step": 1190
+    },
+    {
+      "epoch": 2.0654044750430294,
+      "grad_norm": 0.05850391089916229,
+      "learning_rate": 3.265110941086458e-06,
+      "loss": 0.3652,
+      "step": 1200
+    },
+    {
+      "epoch": 2.0826161790017212,
+      "grad_norm": 0.03551226481795311,
+      "learning_rate": 3.2459831675592962e-06,
+      "loss": 1.1687,
+      "step": 1210
+    },
+    {
+      "epoch": 2.099827882960413,
+      "grad_norm": 0.035683631896972656,
+      "learning_rate": 3.226855394032135e-06,
+      "loss": 0.3377,
+      "step": 1220
+    },
+    {
+      "epoch": 2.117039586919105,
+      "grad_norm": 0.05406322330236435,
+      "learning_rate": 3.2077276205049736e-06,
+      "loss": 0.4614,
+      "step": 1230
+    },
+    {
+      "epoch": 2.134251290877797,
+      "grad_norm": 0.030787965282797813,
+      "learning_rate": 3.188599846977812e-06,
+      "loss": 0.3771,
+      "step": 1240
+    },
+    {
+      "epoch": 2.1514629948364887,
+      "grad_norm": 0.04496818408370018,
+      "learning_rate": 3.169472073450651e-06,
+      "loss": 0.4846,
+      "step": 1250
+    },
+    {
+      "epoch": 2.1686746987951806,
+      "grad_norm": 0.03633632883429527,
+      "learning_rate": 3.150344299923489e-06,
+      "loss": 0.3549,
+      "step": 1260
+    },
+    {
+      "epoch": 2.1858864027538725,
+      "grad_norm": 0.033117033541202545,
+      "learning_rate": 3.1312165263963278e-06,
+      "loss": 0.4224,
+      "step": 1270
+    },
+    {
+      "epoch": 2.2030981067125643,
+      "grad_norm": 0.04940853640437126,
+      "learning_rate": 3.1120887528691664e-06,
+      "loss": 0.6976,
+      "step": 1280
+    },
+    {
+      "epoch": 2.2203098106712567,
+      "grad_norm": 0.03474991396069527,
+      "learning_rate": 3.092960979342005e-06,
+      "loss": 0.5837,
+      "step": 1290
+    },
+    {
+      "epoch": 2.2375215146299485,
+      "grad_norm": 0.08616980165243149,
+      "learning_rate": 3.0738332058148433e-06,
+      "loss": 0.5885,
+      "step": 1300
+    },
+    {
+      "epoch": 2.2547332185886404,
+      "grad_norm": 0.04921899363398552,
+      "learning_rate": 3.054705432287682e-06,
+      "loss": 0.4007,
+      "step": 1310
+    },
+    {
+      "epoch": 2.2719449225473323,
+      "grad_norm": 0.033128101378679276,
+      "learning_rate": 3.0355776587605207e-06,
+      "loss": 0.3948,
+      "step": 1320
+    },
+    {
+      "epoch": 2.289156626506024,
+      "grad_norm": 0.0420563630759716,
+      "learning_rate": 3.016449885233359e-06,
+      "loss": 0.6675,
+      "step": 1330
+    },
+    {
+      "epoch": 2.306368330464716,
+      "grad_norm": 0.04620426893234253,
+      "learning_rate": 2.997322111706198e-06,
+      "loss": 0.3454,
+      "step": 1340
+    },
+    {
+      "epoch": 2.323580034423408,
+      "grad_norm": 0.031115278601646423,
+      "learning_rate": 2.9781943381790362e-06,
+      "loss": 0.4697,
+      "step": 1350
+    },
+    {
+      "epoch": 2.3407917383820998,
+      "grad_norm": 0.03716883435845375,
+      "learning_rate": 2.9590665646518745e-06,
+      "loss": 0.7016,
+      "step": 1360
+    },
+    {
+      "epoch": 2.3580034423407916,
+      "grad_norm": 0.2217116802930832,
+      "learning_rate": 2.9399387911247135e-06,
+      "loss": 0.6504,
+      "step": 1370
+    },
+    {
+      "epoch": 2.3752151462994835,
+      "grad_norm": 0.08799983561038971,
+      "learning_rate": 2.9208110175975518e-06,
+      "loss": 0.3518,
+      "step": 1380
+    },
+    {
+      "epoch": 2.3924268502581754,
+      "grad_norm": 0.03414052352309227,
+      "learning_rate": 2.901683244070391e-06,
+      "loss": 0.5522,
+      "step": 1390
+    },
+    {
+      "epoch": 2.4096385542168672,
+      "grad_norm": 0.14305748045444489,
+      "learning_rate": 2.882555470543229e-06,
+      "loss": 0.7692,
+      "step": 1400
+    },
+    {
+      "epoch": 2.4268502581755595,
+      "grad_norm": 0.04776856303215027,
+      "learning_rate": 2.8634276970160673e-06,
+      "loss": 0.4163,
+      "step": 1410
+    },
+    {
+      "epoch": 2.4440619621342514,
+      "grad_norm": 0.06117096543312073,
+      "learning_rate": 2.8442999234889064e-06,
+      "loss": 0.3797,
+      "step": 1420
+    },
+    {
+      "epoch": 2.4612736660929433,
+      "grad_norm": 0.1437849998474121,
+      "learning_rate": 2.8251721499617447e-06,
+      "loss": 0.3978,
+      "step": 1430
+    },
+    {
+      "epoch": 2.478485370051635,
+      "grad_norm": 0.03535407409071922,
+      "learning_rate": 2.8060443764345833e-06,
+      "loss": 0.7543,
+      "step": 1440
+    },
+    {
+      "epoch": 2.495697074010327,
+      "grad_norm": 0.034573543816804886,
+      "learning_rate": 2.786916602907422e-06,
+      "loss": 0.4385,
+      "step": 1450
+    },
+    {
+      "epoch": 2.512908777969019,
+      "grad_norm": 0.05264075845479965,
+      "learning_rate": 2.7677888293802602e-06,
+      "loss": 0.5788,
+      "step": 1460
+    },
+    {
+      "epoch": 2.5301204819277108,
+      "grad_norm": 0.047263339161872864,
+      "learning_rate": 2.748661055853099e-06,
+      "loss": 0.5397,
+      "step": 1470
+    },
+    {
+      "epoch": 2.5473321858864026,
+      "grad_norm": 0.03852943331003189,
+      "learning_rate": 2.7295332823259375e-06,
+      "loss": 0.3995,
+      "step": 1480
+    },
+    {
+      "epoch": 2.5645438898450945,
+      "grad_norm": 0.04756772890686989,
+      "learning_rate": 2.710405508798776e-06,
+      "loss": 0.5136,
+      "step": 1490
+    },
+    {
+      "epoch": 2.581755593803787,
+      "grad_norm": 0.07750029861927032,
+      "learning_rate": 2.6912777352716144e-06,
+      "loss": 0.8293,
+      "step": 1500
+    },
+    {
+      "epoch": 2.5989672977624787,
+      "grad_norm": 0.047012392431497574,
+      "learning_rate": 2.672149961744453e-06,
+      "loss": 0.5485,
+      "step": 1510
+    },
+    {
+      "epoch": 2.6161790017211706,
+      "grad_norm": 0.04318179562687874,
+      "learning_rate": 2.6530221882172918e-06,
+      "loss": 0.4112,
+      "step": 1520
+    },
+    {
+      "epoch": 2.6333907056798624,
+      "grad_norm": 0.06012555584311485,
+      "learning_rate": 2.63389441469013e-06,
+      "loss": 0.7031,
+      "step": 1530
+    },
+    {
+      "epoch": 2.6506024096385543,
+      "grad_norm": 0.03384987264871597,
+      "learning_rate": 2.614766641162969e-06,
+      "loss": 0.439,
+      "step": 1540
+    },
+    {
+      "epoch": 2.667814113597246,
+      "grad_norm": 0.05770883336663246,
+      "learning_rate": 2.5956388676358073e-06,
+      "loss": 0.3991,
+      "step": 1550
+    },
+    {
+      "epoch": 2.685025817555938,
+      "grad_norm": 0.05510050430893898,
+      "learning_rate": 2.5765110941086456e-06,
+      "loss": 0.9784,
+      "step": 1560
+    },
+    {
+      "epoch": 2.70223752151463,
+      "grad_norm": 0.055017050355672836,
+      "learning_rate": 2.5573833205814846e-06,
+      "loss": 0.3796,
+      "step": 1570
+    },
+    {
+      "epoch": 2.719449225473322,
+      "grad_norm": 0.04332127049565315,
+      "learning_rate": 2.538255547054323e-06,
+      "loss": 0.433,
+      "step": 1580
+    },
+    {
+      "epoch": 2.7366609294320137,
+      "grad_norm": 0.060054711997509,
+      "learning_rate": 2.519127773527162e-06,
+      "loss": 0.2799,
+      "step": 1590
+    },
+    {
+      "epoch": 2.7538726333907055,
+      "grad_norm": 0.0340825691819191,
+      "learning_rate": 2.5e-06,
+      "loss": 0.6797,
+      "step": 1600
+    },
+    {
+      "epoch": 2.7710843373493974,
+      "grad_norm": 0.22405555844306946,
+      "learning_rate": 2.480872226472839e-06,
+      "loss": 0.6071,
+      "step": 1610
+    },
+    {
+      "epoch": 2.7882960413080893,
+      "grad_norm": 0.04493927210569382,
+      "learning_rate": 2.4617444529456775e-06,
+      "loss": 0.4004,
+      "step": 1620
+    },
+    {
+      "epoch": 2.805507745266781,
+      "grad_norm": 0.06454917788505554,
+      "learning_rate": 2.4426166794185158e-06,
+      "loss": 0.3903,
+      "step": 1630
+    },
+    {
+      "epoch": 2.8227194492254735,
+      "grad_norm": 0.07336492091417313,
+      "learning_rate": 2.4234889058913544e-06,
+      "loss": 0.9157,
+      "step": 1640
+    },
+    {
+      "epoch": 2.8399311531841653,
+      "grad_norm": 0.08775831758975983,
+      "learning_rate": 2.404361132364193e-06,
+      "loss": 0.4865,
+      "step": 1650
+    },
+    {
+      "epoch": 2.857142857142857,
+      "grad_norm": 0.03372660651803017,
+      "learning_rate": 2.3852333588370317e-06,
+      "loss": 0.3975,
+      "step": 1660
+    },
+    {
+      "epoch": 2.874354561101549,
+      "grad_norm": 0.034449730068445206,
+      "learning_rate": 2.3661055853098704e-06,
+      "loss": 0.3927,
+      "step": 1670
+    },
+    {
+      "epoch": 2.891566265060241,
+      "grad_norm": 0.02975647896528244,
+      "learning_rate": 2.3469778117827086e-06,
+      "loss": 0.3664,
+      "step": 1680
+    },
+    {
+      "epoch": 2.908777969018933,
+      "grad_norm": 0.037901297211647034,
+      "learning_rate": 2.3278500382555473e-06,
+      "loss": 0.3973,
+      "step": 1690
+    },
+    {
+      "epoch": 2.9259896729776247,
+      "grad_norm": 0.05662724748253822,
+      "learning_rate": 2.308722264728386e-06,
+      "loss": 0.4422,
+      "step": 1700
+    },
+    {
+      "epoch": 2.9432013769363166,
+      "grad_norm": 0.044157788157463074,
+      "learning_rate": 2.289594491201224e-06,
+      "loss": 0.4324,
+      "step": 1710
+    },
+    {
+      "epoch": 2.960413080895009,
+      "grad_norm": 0.04280713573098183,
+      "learning_rate": 2.270466717674063e-06,
+      "loss": 0.5674,
+      "step": 1720
+    },
+    {
+      "epoch": 2.9776247848537007,
+      "grad_norm": 0.04871043935418129,
+      "learning_rate": 2.2513389441469015e-06,
+      "loss": 0.3223,
+      "step": 1730
+    },
+    {
+      "epoch": 2.9948364888123926,
+      "grad_norm": 0.036149609833955765,
+      "learning_rate": 2.2322111706197398e-06,
+      "loss": 0.6471,
+      "step": 1740
+    },
+    {
+      "epoch": 3.0120481927710845,
+      "grad_norm": 0.02951321005821228,
+      "learning_rate": 2.2130833970925784e-06,
+      "loss": 0.3926,
+      "step": 1750
+    },
+    {
+      "epoch": 3.0292598967297764,
+      "grad_norm": 0.04006199911236763,
+      "learning_rate": 2.193955623565417e-06,
+      "loss": 0.6222,
+      "step": 1760
+    },
+    {
+      "epoch": 3.0464716006884682,
+      "grad_norm": 0.03238508850336075,
+      "learning_rate": 2.1748278500382557e-06,
+      "loss": 0.4144,
+      "step": 1770
+    },
+    {
+      "epoch": 3.06368330464716,
+      "grad_norm": 0.035425204783678055,
+      "learning_rate": 2.1557000765110944e-06,
+      "loss": 0.3745,
+      "step": 1780
+    },
+    {
+      "epoch": 3.080895008605852,
+      "grad_norm": 0.08181657642126083,
+      "learning_rate": 2.1365723029839326e-06,
+      "loss": 0.4049,
+      "step": 1790
+    },
+    {
+      "epoch": 3.098106712564544,
+      "grad_norm": 0.03448079526424408,
+      "learning_rate": 2.1174445294567713e-06,
+      "loss": 0.5435,
+      "step": 1800
+    }
+  ],
+  "logging_steps": 10,
+  "max_steps": 2905,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 5,
+  "save_steps": 300,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 0.0,
+  "train_batch_size": 1,
+  "trial_name": null,
+  "trial_params": null
+}

training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:91116a9334865a57d085ece2efd2157dd49fc30fd2d8b9308f8a36934a61ff70
+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)