feat: Add trained model

Signed-off-by: Julep AI <developers@julep.ai>

config.json

@@ -0,0 +1,32 @@
+{
+  "_name_or_path": "prakharz/DIAL-T0",
+  "architectures": [
+    "T5ForConditionalGeneration"
+  ],
+  "d_ff": 5120,
+  "d_kv": 64,
+  "d_model": 2048,
+  "decoder_start_token_id": 0,
+  "dense_act_fn": "gelu_new",
+  "dropout_rate": 0.1,
+  "eos_token_id": 1,
+  "feed_forward_proj": "gated-gelu",
+  "gradient_checkpointing": false,
+  "initializer_factor": 1.0,
+  "is_encoder_decoder": true,
+  "is_gated_act": true,
+  "layer_norm_epsilon": 1e-06,
+  "model_type": "t5",
+  "num_decoder_layers": 24,
+  "num_heads": 32,
+  "num_layers": 24,
+  "output_past": true,
+  "pad_token_id": 0,
+  "relative_attention_max_distance": 128,
+  "relative_attention_num_buckets": 32,
+  "tie_word_embeddings": false,
+  "torch_dtype": "bfloat16",
+  "transformers_version": "4.25.1",
+  "use_cache": false,
+  "vocab_size": 32100
+}

global_step1600/bf16_zero_pp_rank_0_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cb2bf10ab79a734abf48b4526e85617b23cb1c5b7111ba3ad379dbac908bad78
+size 8548942587

global_step1600/bf16_zero_pp_rank_1_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a5a8b042a08e277e9607ae79d179bc99de7f521b92e835ee4da73b2c46aeb054
+size 8548943675

global_step1600/bf16_zero_pp_rank_2_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b94e350dee7dc475cd000bdcb40cb1defaa8bb627d8fd31733e726003f4393a5
+size 8548944827

global_step1600/bf16_zero_pp_rank_3_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ce62491926e9ac6b1c1cc2b7dad2e51995e342c008162e81bff5549bc55054f7
+size 8548943867

global_step1600/mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4a5eefa82e897593644630ca4a35533b7d933cb8a29d64ca442390b9cb02df76
+size 5699481819

latest

@@ -0,0 +1 @@
+global_step1600

pytorch_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:19fb7c9a42c8115a0b2518631b5bbb17dea18f382fc7c1533c4c534df2229d15
+size 5699366451

rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c41b81c8227fa28b27fd97d2ad7721930a51b65a4edde92cf2f4fab8c936ef99
+size 14583

rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:191c7a58927c9f14ad9603e587b64634de332f567bbc5010a7f0d63bda38db1b
+size 14583

rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:030bb0a6f5114351f55cfc1c6138a702b0c7aabf96b28f202d0656ade264bc2d
+size 14583

rng_state_3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:26bd79bd6efb5789c1f2bbb43bb094cf34184e47e809cc4029a946738f0927d9
+size 14583

special_tokens_map.json

@@ -0,0 +1,107 @@
+{
+  "additional_special_tokens": [
+    "<extra_id_0>",
+    "<extra_id_1>",
+    "<extra_id_2>",
+    "<extra_id_3>",
+    "<extra_id_4>",
+    "<extra_id_5>",
+    "<extra_id_6>",
+    "<extra_id_7>",
+    "<extra_id_8>",
+    "<extra_id_9>",
+    "<extra_id_10>",
+    "<extra_id_11>",
+    "<extra_id_12>",
+    "<extra_id_13>",
+    "<extra_id_14>",
+    "<extra_id_15>",
+    "<extra_id_16>",
+    "<extra_id_17>",
+    "<extra_id_18>",
+    "<extra_id_19>",
+    "<extra_id_20>",
+    "<extra_id_21>",
+    "<extra_id_22>",
+    "<extra_id_23>",
+    "<extra_id_24>",
+    "<extra_id_25>",
+    "<extra_id_26>",
+    "<extra_id_27>",
+    "<extra_id_28>",
+    "<extra_id_29>",
+    "<extra_id_30>",
+    "<extra_id_31>",
+    "<extra_id_32>",
+    "<extra_id_33>",
+    "<extra_id_34>",
+    "<extra_id_35>",
+    "<extra_id_36>",
+    "<extra_id_37>",
+    "<extra_id_38>",
+    "<extra_id_39>",
+    "<extra_id_40>",
+    "<extra_id_41>",
+    "<extra_id_42>",
+    "<extra_id_43>",
+    "<extra_id_44>",
+    "<extra_id_45>",
+    "<extra_id_46>",
+    "<extra_id_47>",
+    "<extra_id_48>",
+    "<extra_id_49>",
+    "<extra_id_50>",
+    "<extra_id_51>",
+    "<extra_id_52>",
+    "<extra_id_53>",
+    "<extra_id_54>",
+    "<extra_id_55>",
+    "<extra_id_56>",
+    "<extra_id_57>",
+    "<extra_id_58>",
+    "<extra_id_59>",
+    "<extra_id_60>",
+    "<extra_id_61>",
+    "<extra_id_62>",
+    "<extra_id_63>",
+    "<extra_id_64>",
+    "<extra_id_65>",
+    "<extra_id_66>",
+    "<extra_id_67>",
+    "<extra_id_68>",
+    "<extra_id_69>",
+    "<extra_id_70>",
+    "<extra_id_71>",
+    "<extra_id_72>",
+    "<extra_id_73>",
+    "<extra_id_74>",
+    "<extra_id_75>",
+    "<extra_id_76>",
+    "<extra_id_77>",
+    "<extra_id_78>",
+    "<extra_id_79>",
+    "<extra_id_80>",
+    "<extra_id_81>",
+    "<extra_id_82>",
+    "<extra_id_83>",
+    "<extra_id_84>",
+    "<extra_id_85>",
+    "<extra_id_86>",
+    "<extra_id_87>",
+    "<extra_id_88>",
+    "<extra_id_89>",
+    "<extra_id_90>",
+    "<extra_id_91>",
+    "<extra_id_92>",
+    "<extra_id_93>",
+    "<extra_id_94>",
+    "<extra_id_95>",
+    "<extra_id_96>",
+    "<extra_id_97>",
+    "<extra_id_98>",
+    "<extra_id_99>"
+  ],
+  "eos_token": "</s>",
+  "pad_token": "<pad>",
+  "unk_token": "<unk>"
+}

tokenizer_config.json

@@ -0,0 +1,113 @@
+{
+  "additional_special_tokens": [
+    "<extra_id_0>",
+    "<extra_id_1>",
+    "<extra_id_2>",
+    "<extra_id_3>",
+    "<extra_id_4>",
+    "<extra_id_5>",
+    "<extra_id_6>",
+    "<extra_id_7>",
+    "<extra_id_8>",
+    "<extra_id_9>",
+    "<extra_id_10>",
+    "<extra_id_11>",
+    "<extra_id_12>",
+    "<extra_id_13>",
+    "<extra_id_14>",
+    "<extra_id_15>",
+    "<extra_id_16>",
+    "<extra_id_17>",
+    "<extra_id_18>",
+    "<extra_id_19>",
+    "<extra_id_20>",
+    "<extra_id_21>",
+    "<extra_id_22>",
+    "<extra_id_23>",
+    "<extra_id_24>",
+    "<extra_id_25>",
+    "<extra_id_26>",
+    "<extra_id_27>",
+    "<extra_id_28>",
+    "<extra_id_29>",
+    "<extra_id_30>",
+    "<extra_id_31>",
+    "<extra_id_32>",
+    "<extra_id_33>",
+    "<extra_id_34>",
+    "<extra_id_35>",
+    "<extra_id_36>",
+    "<extra_id_37>",
+    "<extra_id_38>",
+    "<extra_id_39>",
+    "<extra_id_40>",
+    "<extra_id_41>",
+    "<extra_id_42>",
+    "<extra_id_43>",
+    "<extra_id_44>",
+    "<extra_id_45>",
+    "<extra_id_46>",
+    "<extra_id_47>",
+    "<extra_id_48>",
+    "<extra_id_49>",
+    "<extra_id_50>",
+    "<extra_id_51>",
+    "<extra_id_52>",
+    "<extra_id_53>",
+    "<extra_id_54>",
+    "<extra_id_55>",
+    "<extra_id_56>",
+    "<extra_id_57>",
+    "<extra_id_58>",
+    "<extra_id_59>",
+    "<extra_id_60>",
+    "<extra_id_61>",
+    "<extra_id_62>",
+    "<extra_id_63>",
+    "<extra_id_64>",
+    "<extra_id_65>",
+    "<extra_id_66>",
+    "<extra_id_67>",
+    "<extra_id_68>",
+    "<extra_id_69>",
+    "<extra_id_70>",
+    "<extra_id_71>",
+    "<extra_id_72>",
+    "<extra_id_73>",
+    "<extra_id_74>",
+    "<extra_id_75>",
+    "<extra_id_76>",
+    "<extra_id_77>",
+    "<extra_id_78>",
+    "<extra_id_79>",
+    "<extra_id_80>",
+    "<extra_id_81>",
+    "<extra_id_82>",
+    "<extra_id_83>",
+    "<extra_id_84>",
+    "<extra_id_85>",
+    "<extra_id_86>",
+    "<extra_id_87>",
+    "<extra_id_88>",
+    "<extra_id_89>",
+    "<extra_id_90>",
+    "<extra_id_91>",
+    "<extra_id_92>",
+    "<extra_id_93>",
+    "<extra_id_94>",
+    "<extra_id_95>",
+    "<extra_id_96>",
+    "<extra_id_97>",
+    "<extra_id_98>",
+    "<extra_id_99>"
+  ],
+  "eos_token": "</s>",
+  "extra_ids": 100,
+  "model_max_length": 512,
+  "name_or_path": "prakharz/DIAL-T0",
+  "pad_token": "<pad>",
+  "sp_model_kwargs": {},
+  "special_tokens_map_file": "/home/ubuntu/.cache/huggingface/transformers/303fbee39a17e96552ac07e02b70ba62ff0ad760609687e3a1b92b4ad2dff58c.c94798918c92ded6aeef2d2f0e666d2cc4145eca1aa6e1336fde07f2e13e2f46",
+  "tokenizer_class": "T5Tokenizer",
+  "unk_token": "<unk>"
+}

trainer_state.json

@@ -0,0 +1,444 @@
+{
+  "best_metric": 85.9274,
+  "best_model_checkpoint": "./output/checkpoint-1600",
+  "epoch": 1.463719512195122,
+  "global_step": 1600,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.9496336996337e-05,
+      "loss": 0.7035,
+      "step": 25
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.8923992673992674e-05,
+      "loss": 0.3973,
+      "step": 50
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.8351648351648355e-05,
+      "loss": 0.3092,
+      "step": 75
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.7779304029304036e-05,
+      "loss": 0.2813,
+      "step": 100
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.720695970695971e-05,
+      "loss": 0.2603,
+      "step": 125
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.6634615384615384e-05,
+      "loss": 0.2371,
+      "step": 150
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.606227106227106e-05,
+      "loss": 0.2334,
+      "step": 175
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.548992673992674e-05,
+      "loss": 0.2296,
+      "step": 200
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.491758241758242e-05,
+      "loss": 0.2219,
+      "step": 225
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.4345238095238095e-05,
+      "loss": 0.2078,
+      "step": 250
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.3772893772893776e-05,
+      "loss": 0.2123,
+      "step": 275
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.320054945054946e-05,
+      "loss": 0.2063,
+      "step": 300
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.262820512820513e-05,
+      "loss": 0.2002,
+      "step": 325
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 4.2055860805860806e-05,
+      "loss": 0.2102,
+      "step": 350
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 4.148351648351649e-05,
+      "loss": 0.2003,
+      "step": 375
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 4.091117216117216e-05,
+      "loss": 0.2019,
+      "step": 400
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 4.033882783882784e-05,
+      "loss": 0.1944,
+      "step": 425
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.9766483516483516e-05,
+      "loss": 0.1935,
+      "step": 450
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.91941391941392e-05,
+      "loss": 0.1974,
+      "step": 475
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 3.862179487179488e-05,
+      "loss": 0.1922,
+      "step": 500
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 3.804945054945055e-05,
+      "loss": 0.188,
+      "step": 525
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 3.747710622710623e-05,
+      "loss": 0.1945,
+      "step": 550
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 3.690476190476191e-05,
+      "loss": 0.1897,
+      "step": 575
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 3.633241758241758e-05,
+      "loss": 0.1797,
+      "step": 600
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.576007326007326e-05,
+      "loss": 0.1752,
+      "step": 625
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.5187728937728944e-05,
+      "loss": 0.1797,
+      "step": 650
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.461538461538462e-05,
+      "loss": 0.1785,
+      "step": 675
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.404304029304029e-05,
+      "loss": 0.185,
+      "step": 700
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 3.347069597069597e-05,
+      "loss": 0.1751,
+      "step": 725
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 3.289835164835165e-05,
+      "loss": 0.1799,
+      "step": 750
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.232600732600733e-05,
+      "loss": 0.1801,
+      "step": 775
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.1753663003663003e-05,
+      "loss": 0.1748,
+      "step": 800
+    },
+    {
+      "epoch": 0.73,
+      "eval_Bleu_1": 0.8552,
+      "eval_Bleu_2": 0.0009,
+      "eval_Bleu_3": 0.0001,
+      "eval_Bleu_4": 0.0,
+      "eval_ROUGE_L": 0.8552,
+      "eval_accuracy": 85.52,
+      "eval_f1": 85.52,
+      "eval_gen_len": 2.2084,
+      "eval_loss": 0.17816422879695892,
+      "eval_precision": 85.52,
+      "eval_recall": 85.52,
+      "eval_rouge1": 85.6587,
+      "eval_rouge2": 3.7183,
+      "eval_rougeL": 85.6528,
+      "eval_rougeLsum": 85.6348,
+      "eval_runtime": 1004.3102,
+      "eval_samples_per_second": 34.705,
+      "eval_steps_per_second": 4.338,
+      "step": 800
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1181318681318685e-05,
+      "loss": 0.1698,
+      "step": 825
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.0608974358974366e-05,
+      "loss": 0.1763,
+      "step": 850
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.0036630036630036e-05,
+      "loss": 0.1655,
+      "step": 875
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.9464285714285718e-05,
+      "loss": 0.1703,
+      "step": 900
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.8891941391941392e-05,
+      "loss": 0.174,
+      "step": 925
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.831959706959707e-05,
+      "loss": 0.1723,
+      "step": 950
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.774725274725275e-05,
+      "loss": 0.1739,
+      "step": 975
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.7174908424908425e-05,
+      "loss": 0.1673,
+      "step": 1000
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.6602564102564102e-05,
+      "loss": 0.1647,
+      "step": 1025
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.6030219780219783e-05,
+      "loss": 0.1739,
+      "step": 1050
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.5457875457875458e-05,
+      "loss": 0.1703,
+      "step": 1075
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 2.488553113553114e-05,
+      "loss": 0.1727,
+      "step": 1100
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 2.4313186813186813e-05,
+      "loss": 0.1603,
+      "step": 1125
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 2.374084249084249e-05,
+      "loss": 0.1511,
+      "step": 1150
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 2.3168498168498172e-05,
+      "loss": 0.1544,
+      "step": 1175
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 2.2596153846153846e-05,
+      "loss": 0.1517,
+      "step": 1200
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 2.2023809523809524e-05,
+      "loss": 0.1556,
+      "step": 1225
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 2.14514652014652e-05,
+      "loss": 0.1688,
+      "step": 1250
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 2.0879120879120882e-05,
+      "loss": 0.1658,
+      "step": 1275
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 2.0306776556776557e-05,
+      "loss": 0.1587,
+      "step": 1300
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.9734432234432234e-05,
+      "loss": 0.1664,
+      "step": 1325
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.9162087912087915e-05,
+      "loss": 0.1558,
+      "step": 1350
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.858974358974359e-05,
+      "loss": 0.1591,
+      "step": 1375
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.8017399267399267e-05,
+      "loss": 0.1558,
+      "step": 1400
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.7445054945054945e-05,
+      "loss": 0.1583,
+      "step": 1425
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.6872710622710626e-05,
+      "loss": 0.155,
+      "step": 1450
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.63003663003663e-05,
+      "loss": 0.1644,
+      "step": 1475
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.5728021978021978e-05,
+      "loss": 0.1574,
+      "step": 1500
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.5155677655677656e-05,
+      "loss": 0.1555,
+      "step": 1525
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.4583333333333335e-05,
+      "loss": 0.1542,
+      "step": 1550
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.4010989010989013e-05,
+      "loss": 0.1537,
+      "step": 1575
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.3438644688644688e-05,
+      "loss": 0.1493,
+      "step": 1600
+    },
+    {
+      "epoch": 1.46,
+      "eval_Bleu_1": 0.8593,
+      "eval_Bleu_2": 0.0009,
+      "eval_Bleu_3": 0.0001,
+      "eval_Bleu_4": 0.0,
+      "eval_ROUGE_L": 0.8593,
+      "eval_accuracy": 85.9274,
+      "eval_f1": 85.9274,
+      "eval_gen_len": 2.2038,
+      "eval_loss": 0.1681133359670639,
+      "eval_precision": 85.9274,
+      "eval_recall": 85.9274,
+      "eval_rouge1": 86.0601,
+      "eval_rouge2": 3.5662,
+      "eval_rougeL": 86.0539,
+      "eval_rougeLsum": 86.0496,
+      "eval_runtime": 1001.4875,
+      "eval_samples_per_second": 34.803,
+      "eval_steps_per_second": 4.351,
+      "step": 1600
+    }
+  ],
+  "max_steps": 2186,
+  "num_train_epochs": 2,
+  "total_flos": 2.7769320388994007e+18,
+  "trial_name": null,
+  "trial_params": null
+}

training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2850b9e779255d3efce03dff445e52ea503bbb829fd1b72d24779361b420c4a9
+size 4603

zero_to_fp32.py

@@ -0,0 +1,482 @@
+#!/usr/bin/env python
+
+# This script extracts fp32 consolidated weights from a zero 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
+
+# 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)
+
+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_optim_files(checkpoint_dir):
+    # XXX: need to test that this simple glob rule works for multi-node setup too
+    optim_files = sorted(glob.glob(os.path.join(checkpoint_dir,
+                                                "*_optim_states.pt")),
+                         key=natural_keys)
+
+    if len(optim_files) == 0:
+        raise FileNotFoundError(
+            f"can't find '*_optim_states.pt' files in directory '{checkpoint_dir}'")
+
+    return optim_files
+
+
+def parse_model_state(file):
+    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]
+
+    ds_version = state_dict.get(DS_VERSION, None)
+
+    return buffers, param_shapes, ds_version
+
+
+def parse_optim_states(files, ds_checkpoint_dir):
+
+    total_files = len(files)
+    state_dicts = []
+    for f in files:
+        state_dicts.append(torch.load(f, map_location=device))
+
+    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):
+    """
+    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_file = get_model_state_file(ds_checkpoint_dir, zero_stage)
+    buffers, param_shapes, ds_version = parse_model_state(model_file)
+    print(f'Parsing checkpoint created by deepspeed=={ds_version}')
+
+    if zero_stage == 2:
+        return _get_fp32_state_dict_from_zero2_checkpoint(world_size,
+                                                          param_shapes,
+                                                          fp32_flat_groups,
+                                                          buffers)
+    elif zero_stage == 3:
+        return _get_fp32_state_dict_from_zero3_checkpoint(world_size,
+                                                          param_shapes,
+                                                          fp32_flat_groups,
+                                                          buffers)
+
+
+def _get_fp32_state_dict_from_zero2_checkpoint(world_size,
+                                               param_shapes,
+                                               fp32_flat_groups,
+                                               buffers):
+
+    # 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.")
+
+    state_dict = OrderedDict()
+
+    # buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    # 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()
+            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"
+    )
+
+    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 _get_fp32_state_dict_from_zero3_checkpoint(world_size,
+                                               param_shapes,
+                                               fp32_flat_groups,
+                                               buffers):
+
+    # 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
+
+    avail_numel = fp32_flat_groups[0].numel() * world_size
+    # 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
+        print(f"Have {avail_numel} numels to process.")
+        print(f"Need {wanted_numel} numels in {wanted_params} params.")
+
+    state_dict = OrderedDict()
+
+    # buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    # 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"{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 fp32 state dict with {total_params} params {total_numel} elements"
+    )
+
+    return state_dict
+
+
+def get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag=None):
+    """
+    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``
+
+    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)
+
+
+def convert_zero_checkpoint_to_fp32_state_dict(checkpoint_dir, output_file, tag=None):
+    """
+    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``
+    """
+
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag)
+    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("-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)