Upload folder using huggingface_hub

global_step700/bf16_zero_pp_rank_0_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5eeec11714b0fee5900be6302e5949a82d2928c45a552c3d1d6993694e059567
+size 25613049823

global_step700/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:d0e89aa7ae75adffe0676555363831ce6ecd7570b59b0963f6477098cb44900a
+size 25613049823

global_step700/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:b48ed2174a7b35d02653977a43bc7125ddf9e97e2ee706b365652ad3f12821a9
+size 25613049823

global_step700/bf16_zero_pp_rank_3_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:57df57eb9481893414a1e7e268894871130612875c5c128bf129ec2c1d6a0437
+size 25613049823

global_step700/zero_pp_rank_0_mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:991a6e32a3f05fd89870c84801e3c5e92452e33006fcd446aef3690ce5e180e0
+size 131199

global_step700/zero_pp_rank_1_mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:12b7c28a806d3ad014a060e21b2ee3c0bc2c98a0c333792f951585906dd3da32
+size 131135

global_step700/zero_pp_rank_2_mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:503c0eac64e9948c7a474f4c35247f3a112a6f8daa5f3dcd84caa7729432d04f
+size 131135

global_step700/zero_pp_rank_3_mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bb23420badf7250a50ee10795d14f55bb1a1cf2f1fe0c4b83bbfea9b065a690a
+size 131135

latest

@@ -0,0 +1 @@
+global_step700

rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:659b1cdee2219458dd84ce6a632a595465680b8080e5c44bd600ff97eca8d752
+size 15429

rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:86accf27064cdd503053e90476a6bd10de333d4ff0594535ad55ea13a473c91d
+size 15429

rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:18ca8d714ef40be035404c1957b5a4dee84e1f43980408393f8aa710552ee6f6
+size 15429

rng_state_3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2cfdebe99e40accc9c9d8f09c63136a14abda997d9b501969ec8e16e9d183179
+size 15429

scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3f5b0a08026753faf278ef077908a9ba17630202b616bbd23f16e30e79245d6a
+size 1465

trainer_state.json

@@ -0,0 +1,664 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.0,
+  "eval_steps": 500,
+  "global_step": 700,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.014295925661186561,
+      "grad_norm": 16.642337799072266,
+      "learning_rate": 1.9742857142857144e-05,
+      "loss": 4.2448,
+      "mean_token_accuracy": 0.44755197104532274,
+      "num_tokens": 63714.0,
+      "step": 10
+    },
+    {
+      "epoch": 0.028591851322373123,
+      "grad_norm": 12.869135856628418,
+      "learning_rate": 1.945714285714286e-05,
+      "loss": 1.9287,
+      "mean_token_accuracy": 0.5766903940588236,
+      "num_tokens": 128528.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04288777698355969,
+      "grad_norm": 18.376684188842773,
+      "learning_rate": 1.9171428571428573e-05,
+      "loss": 1.6956,
+      "mean_token_accuracy": 0.597195016592741,
+      "num_tokens": 191200.0,
+      "step": 30
+    },
+    {
+      "epoch": 0.057183702644746245,
+      "grad_norm": 17.71656036376953,
+      "learning_rate": 1.888571428571429e-05,
+      "loss": 1.6076,
+      "mean_token_accuracy": 0.6067132025957107,
+      "num_tokens": 255728.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.07147962830593281,
+      "grad_norm": 21.026283264160156,
+      "learning_rate": 1.86e-05,
+      "loss": 1.5728,
+      "mean_token_accuracy": 0.612850959226489,
+      "num_tokens": 319058.0,
+      "step": 50
+    },
+    {
+      "epoch": 0.08577555396711938,
+      "grad_norm": 46.10198974609375,
+      "learning_rate": 1.8314285714285714e-05,
+      "loss": 1.5977,
+      "mean_token_accuracy": 0.6111391615122557,
+      "num_tokens": 384900.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.10007147962830593,
+      "grad_norm": 14.742942810058594,
+      "learning_rate": 1.802857142857143e-05,
+      "loss": 1.5649,
+      "mean_token_accuracy": 0.6097237385809422,
+      "num_tokens": 450346.0,
+      "step": 70
+    },
+    {
+      "epoch": 0.11436740528949249,
+      "grad_norm": 43.62748718261719,
+      "learning_rate": 1.7742857142857143e-05,
+      "loss": 1.5184,
+      "mean_token_accuracy": 0.6210372049361468,
+      "num_tokens": 515018.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.12866333095067906,
+      "grad_norm": 15.469511032104492,
+      "learning_rate": 1.745714285714286e-05,
+      "loss": 1.4736,
+      "mean_token_accuracy": 0.6270900748670101,
+      "num_tokens": 576955.0,
+      "step": 90
+    },
+    {
+      "epoch": 0.14295925661186562,
+      "grad_norm": 19.448793411254883,
+      "learning_rate": 1.717142857142857e-05,
+      "loss": 1.4637,
+      "mean_token_accuracy": 0.6368957210332156,
+      "num_tokens": 641295.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.15725518227305219,
+      "grad_norm": 37.31778335571289,
+      "learning_rate": 1.6885714285714288e-05,
+      "loss": 1.5303,
+      "mean_token_accuracy": 0.6210926879197359,
+      "num_tokens": 706683.0,
+      "step": 110
+    },
+    {
+      "epoch": 0.17155110793423875,
+      "grad_norm": 9.722342491149902,
+      "learning_rate": 1.66e-05,
+      "loss": 1.4596,
+      "mean_token_accuracy": 0.6298462159931659,
+      "num_tokens": 771285.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.18584703359542531,
+      "grad_norm": 9.656769752502441,
+      "learning_rate": 1.6314285714285716e-05,
+      "loss": 1.5281,
+      "mean_token_accuracy": 0.6251190695911646,
+      "num_tokens": 840678.0,
+      "step": 130
+    },
+    {
+      "epoch": 0.20014295925661185,
+      "grad_norm": 9.608354568481445,
+      "learning_rate": 1.602857142857143e-05,
+      "loss": 1.4438,
+      "mean_token_accuracy": 0.6370445918291807,
+      "num_tokens": 905832.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.21443888491779842,
+      "grad_norm": 9.842904090881348,
+      "learning_rate": 1.5742857142857145e-05,
+      "loss": 1.5379,
+      "mean_token_accuracy": 0.6172758720815181,
+      "num_tokens": 972946.0,
+      "step": 150
+    },
+    {
+      "epoch": 0.22873481057898498,
+      "grad_norm": 18.17994499206543,
+      "learning_rate": 1.545714285714286e-05,
+      "loss": 1.4322,
+      "mean_token_accuracy": 0.6351331725716591,
+      "num_tokens": 1034427.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.24303073624017155,
+      "grad_norm": 8.876994132995605,
+      "learning_rate": 1.5171428571428572e-05,
+      "loss": 1.4343,
+      "mean_token_accuracy": 0.6313620086759328,
+      "num_tokens": 1101359.0,
+      "step": 170
+    },
+    {
+      "epoch": 0.2573266619013581,
+      "grad_norm": 10.895979881286621,
+      "learning_rate": 1.4885714285714288e-05,
+      "loss": 1.4537,
+      "mean_token_accuracy": 0.633945481479168,
+      "num_tokens": 1166538.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.27162258756254465,
+      "grad_norm": 12.30453872680664,
+      "learning_rate": 1.46e-05,
+      "loss": 1.5363,
+      "mean_token_accuracy": 0.6255367647856473,
+      "num_tokens": 1235118.0,
+      "step": 190
+    },
+    {
+      "epoch": 0.28591851322373124,
+      "grad_norm": 10.28065299987793,
+      "learning_rate": 1.4314285714285717e-05,
+      "loss": 1.4199,
+      "mean_token_accuracy": 0.635765865072608,
+      "num_tokens": 1300601.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.3002144388849178,
+      "grad_norm": 11.893675804138184,
+      "learning_rate": 1.402857142857143e-05,
+      "loss": 1.3881,
+      "mean_token_accuracy": 0.6449477795511485,
+      "num_tokens": 1362732.0,
+      "step": 210
+    },
+    {
+      "epoch": 0.31451036454610437,
+      "grad_norm": 11.485602378845215,
+      "learning_rate": 1.3742857142857144e-05,
+      "loss": 1.3647,
+      "mean_token_accuracy": 0.6486451178789139,
+      "num_tokens": 1424780.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.3288062902072909,
+      "grad_norm": 8.882689476013184,
+      "learning_rate": 1.3457142857142858e-05,
+      "loss": 1.3915,
+      "mean_token_accuracy": 0.6436454936861992,
+      "num_tokens": 1490808.0,
+      "step": 230
+    },
+    {
+      "epoch": 0.3431022158684775,
+      "grad_norm": 18.272981643676758,
+      "learning_rate": 1.3171428571428573e-05,
+      "loss": 1.4796,
+      "mean_token_accuracy": 0.6283955980092287,
+      "num_tokens": 1556933.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.35739814152966404,
+      "grad_norm": 11.947668075561523,
+      "learning_rate": 1.2885714285714285e-05,
+      "loss": 1.4398,
+      "mean_token_accuracy": 0.638329004868865,
+      "num_tokens": 1621053.0,
+      "step": 250
+    },
+    {
+      "epoch": 0.37169406719085063,
+      "grad_norm": 23.547773361206055,
+      "learning_rate": 1.2600000000000001e-05,
+      "loss": 1.3801,
+      "mean_token_accuracy": 0.6448216594755649,
+      "num_tokens": 1686747.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.38598999285203717,
+      "grad_norm": 16.82425880432129,
+      "learning_rate": 1.2314285714285716e-05,
+      "loss": 1.3891,
+      "mean_token_accuracy": 0.6475608512759209,
+      "num_tokens": 1751946.0,
+      "step": 270
+    },
+    {
+      "epoch": 0.4002859185132237,
+      "grad_norm": 11.931357383728027,
+      "learning_rate": 1.202857142857143e-05,
+      "loss": 1.3768,
+      "mean_token_accuracy": 0.6439446356147528,
+      "num_tokens": 1816123.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.4145818441744103,
+      "grad_norm": 14.375319480895996,
+      "learning_rate": 1.1742857142857144e-05,
+      "loss": 1.315,
+      "mean_token_accuracy": 0.6517576463520527,
+      "num_tokens": 1879227.0,
+      "step": 290
+    },
+    {
+      "epoch": 0.42887776983559683,
+      "grad_norm": 10.699817657470703,
+      "learning_rate": 1.1457142857142857e-05,
+      "loss": 1.3519,
+      "mean_token_accuracy": 0.6487406313419342,
+      "num_tokens": 1944238.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.4431736954967834,
+      "grad_norm": 12.067941665649414,
+      "learning_rate": 1.1171428571428573e-05,
+      "loss": 1.2784,
+      "mean_token_accuracy": 0.6627866499125957,
+      "num_tokens": 2007629.0,
+      "step": 310
+    },
+    {
+      "epoch": 0.45746962115796996,
+      "grad_norm": 15.550559997558594,
+      "learning_rate": 1.0885714285714286e-05,
+      "loss": 1.3495,
+      "mean_token_accuracy": 0.6514371998608113,
+      "num_tokens": 2076666.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.47176554681915656,
+      "grad_norm": 30.000173568725586,
+      "learning_rate": 1.0600000000000002e-05,
+      "loss": 1.3358,
+      "mean_token_accuracy": 0.6507035464048385,
+      "num_tokens": 2140860.0,
+      "step": 330
+    },
+    {
+      "epoch": 0.4860614724803431,
+      "grad_norm": 7.962319850921631,
+      "learning_rate": 1.0314285714285715e-05,
+      "loss": 1.3231,
+      "mean_token_accuracy": 0.6570919144898653,
+      "num_tokens": 2204773.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.5003573981415297,
+      "grad_norm": 24.023008346557617,
+      "learning_rate": 1.002857142857143e-05,
+      "loss": 1.3936,
+      "mean_token_accuracy": 0.6455658808350563,
+      "num_tokens": 2270923.0,
+      "step": 350
+    },
+    {
+      "epoch": 0.5146533238027162,
+      "grad_norm": 8.74783706665039,
+      "learning_rate": 9.742857142857143e-06,
+      "loss": 1.3383,
+      "mean_token_accuracy": 0.6552599217742682,
+      "num_tokens": 2337009.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.5289492494639028,
+      "grad_norm": 17.01344108581543,
+      "learning_rate": 9.457142857142858e-06,
+      "loss": 1.3524,
+      "mean_token_accuracy": 0.6488417606800795,
+      "num_tokens": 2405973.0,
+      "step": 370
+    },
+    {
+      "epoch": 0.5432451751250893,
+      "grad_norm": 9.353411674499512,
+      "learning_rate": 9.171428571428572e-06,
+      "loss": 1.2638,
+      "mean_token_accuracy": 0.6636200629174709,
+      "num_tokens": 2469824.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.557541100786276,
+      "grad_norm": 13.265799522399902,
+      "learning_rate": 8.885714285714286e-06,
+      "loss": 1.2254,
+      "mean_token_accuracy": 0.6689145911484957,
+      "num_tokens": 2535167.0,
+      "step": 390
+    },
+    {
+      "epoch": 0.5718370264474625,
+      "grad_norm": 19.46824836730957,
+      "learning_rate": 8.6e-06,
+      "loss": 1.3844,
+      "mean_token_accuracy": 0.6483918268233537,
+      "num_tokens": 2607759.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.586132952108649,
+      "grad_norm": 15.773782730102539,
+      "learning_rate": 8.314285714285715e-06,
+      "loss": 1.2708,
+      "mean_token_accuracy": 0.6655503377318382,
+      "num_tokens": 2670580.0,
+      "step": 410
+    },
+    {
+      "epoch": 0.6004288777698356,
+      "grad_norm": 8.917901039123535,
+      "learning_rate": 8.02857142857143e-06,
+      "loss": 1.2726,
+      "mean_token_accuracy": 0.6589578501880169,
+      "num_tokens": 2737272.0,
+      "step": 420
+    },
+    {
+      "epoch": 0.6147248034310222,
+      "grad_norm": 8.988587379455566,
+      "learning_rate": 7.742857142857144e-06,
+      "loss": 1.221,
+      "mean_token_accuracy": 0.664219357818365,
+      "num_tokens": 2803875.0,
+      "step": 430
+    },
+    {
+      "epoch": 0.6290207290922087,
+      "grad_norm": 12.661059379577637,
+      "learning_rate": 7.457142857142857e-06,
+      "loss": 1.2658,
+      "mean_token_accuracy": 0.662236025184393,
+      "num_tokens": 2869457.0,
+      "step": 440
+    },
+    {
+      "epoch": 0.6433166547533953,
+      "grad_norm": 8.545147895812988,
+      "learning_rate": 7.1714285714285725e-06,
+      "loss": 1.2778,
+      "mean_token_accuracy": 0.6622273363173008,
+      "num_tokens": 2931790.0,
+      "step": 450
+    },
+    {
+      "epoch": 0.6576125804145818,
+      "grad_norm": 20.769514083862305,
+      "learning_rate": 6.885714285714287e-06,
+      "loss": 1.2951,
+      "mean_token_accuracy": 0.6606701787561178,
+      "num_tokens": 2997229.0,
+      "step": 460
+    },
+    {
+      "epoch": 0.6719085060757684,
+      "grad_norm": 12.466110229492188,
+      "learning_rate": 6.600000000000001e-06,
+      "loss": 1.1754,
+      "mean_token_accuracy": 0.6822692640125751,
+      "num_tokens": 3063485.0,
+      "step": 470
+    },
+    {
+      "epoch": 0.686204431736955,
+      "grad_norm": 8.45051383972168,
+      "learning_rate": 6.314285714285715e-06,
+      "loss": 1.2102,
+      "mean_token_accuracy": 0.6759132348001003,
+      "num_tokens": 3127984.0,
+      "step": 480
+    },
+    {
+      "epoch": 0.7005003573981415,
+      "grad_norm": 12.029594421386719,
+      "learning_rate": 6.028571428571429e-06,
+      "loss": 1.3355,
+      "mean_token_accuracy": 0.6649406619369984,
+      "num_tokens": 3194219.0,
+      "step": 490
+    },
+    {
+      "epoch": 0.7147962830593281,
+      "grad_norm": 8.824553489685059,
+      "learning_rate": 5.742857142857143e-06,
+      "loss": 1.2317,
+      "mean_token_accuracy": 0.6705160938203335,
+      "num_tokens": 3259068.0,
+      "step": 500
+    },
+    {
+      "epoch": 0.7290922087205146,
+      "grad_norm": 16.150766372680664,
+      "learning_rate": 5.457142857142858e-06,
+      "loss": 1.1558,
+      "mean_token_accuracy": 0.6850677601993084,
+      "num_tokens": 3324070.0,
+      "step": 510
+    },
+    {
+      "epoch": 0.7433881343817013,
+      "grad_norm": 7.721499919891357,
+      "learning_rate": 5.171428571428571e-06,
+      "loss": 1.168,
+      "mean_token_accuracy": 0.6747931383550168,
+      "num_tokens": 3386885.0,
+      "step": 520
+    },
+    {
+      "epoch": 0.7576840600428878,
+      "grad_norm": 9.311972618103027,
+      "learning_rate": 4.885714285714286e-06,
+      "loss": 1.1645,
+      "mean_token_accuracy": 0.6775478422641754,
+      "num_tokens": 3448602.0,
+      "step": 530
+    },
+    {
+      "epoch": 0.7719799857040743,
+      "grad_norm": 9.636552810668945,
+      "learning_rate": 4.600000000000001e-06,
+      "loss": 1.2542,
+      "mean_token_accuracy": 0.6680241461843253,
+      "num_tokens": 3516481.0,
+      "step": 540
+    },
+    {
+      "epoch": 0.7862759113652609,
+      "grad_norm": 36.31599044799805,
+      "learning_rate": 4.314285714285714e-06,
+      "loss": 1.1866,
+      "mean_token_accuracy": 0.6768352195620537,
+      "num_tokens": 3580217.0,
+      "step": 550
+    },
+    {
+      "epoch": 0.8005718370264474,
+      "grad_norm": 7.471230506896973,
+      "learning_rate": 4.028571428571429e-06,
+      "loss": 1.1705,
+      "mean_token_accuracy": 0.6818295098841191,
+      "num_tokens": 3643021.0,
+      "step": 560
+    },
+    {
+      "epoch": 0.8148677626876341,
+      "grad_norm": 48.099830627441406,
+      "learning_rate": 3.742857142857143e-06,
+      "loss": 1.1602,
+      "mean_token_accuracy": 0.6852999441325665,
+      "num_tokens": 3710116.0,
+      "step": 570
+    },
+    {
+      "epoch": 0.8291636883488206,
+      "grad_norm": 13.096914291381836,
+      "learning_rate": 3.4571428571428574e-06,
+      "loss": 1.1942,
+      "mean_token_accuracy": 0.6752621583640576,
+      "num_tokens": 3775926.0,
+      "step": 580
+    },
+    {
+      "epoch": 0.8434596140100071,
+      "grad_norm": 11.580378532409668,
+      "learning_rate": 3.1714285714285714e-06,
+      "loss": 1.1277,
+      "mean_token_accuracy": 0.6849311918020249,
+      "num_tokens": 3840218.0,
+      "step": 590
+    },
+    {
+      "epoch": 0.8577555396711937,
+      "grad_norm": 9.58252239227295,
+      "learning_rate": 2.885714285714286e-06,
+      "loss": 1.187,
+      "mean_token_accuracy": 0.6740429483354091,
+      "num_tokens": 3904300.0,
+      "step": 600
+    },
+    {
+      "epoch": 0.8720514653323803,
+      "grad_norm": 9.778560638427734,
+      "learning_rate": 2.6e-06,
+      "loss": 1.2088,
+      "mean_token_accuracy": 0.6759266927838326,
+      "num_tokens": 3970409.0,
+      "step": 610
+    },
+    {
+      "epoch": 0.8863473909935669,
+      "grad_norm": 9.931038856506348,
+      "learning_rate": 2.3142857142857145e-06,
+      "loss": 1.1766,
+      "mean_token_accuracy": 0.6766778022050858,
+      "num_tokens": 4038742.0,
+      "step": 620
+    },
+    {
+      "epoch": 0.9006433166547534,
+      "grad_norm": 7.126023769378662,
+      "learning_rate": 2.028571428571429e-06,
+      "loss": 1.0968,
+      "mean_token_accuracy": 0.6913008309900761,
+      "num_tokens": 4103374.0,
+      "step": 630
+    },
+    {
+      "epoch": 0.9149392423159399,
+      "grad_norm": 7.73612642288208,
+      "learning_rate": 1.7428571428571432e-06,
+      "loss": 1.1254,
+      "mean_token_accuracy": 0.6863209947943687,
+      "num_tokens": 4170239.0,
+      "step": 640
+    },
+    {
+      "epoch": 0.9292351679771265,
+      "grad_norm": 6.532904148101807,
+      "learning_rate": 1.4571428571428573e-06,
+      "loss": 1.1586,
+      "mean_token_accuracy": 0.6804635964334012,
+      "num_tokens": 4237810.0,
+      "step": 650
+    },
+    {
+      "epoch": 0.9435310936383131,
+      "grad_norm": 7.370081901550293,
+      "learning_rate": 1.1714285714285715e-06,
+      "loss": 1.174,
+      "mean_token_accuracy": 0.6809860028326511,
+      "num_tokens": 4302937.0,
+      "step": 660
+    },
+    {
+      "epoch": 0.9578270192994996,
+      "grad_norm": 7.471885681152344,
+      "learning_rate": 8.857142857142857e-07,
+      "loss": 1.1755,
+      "mean_token_accuracy": 0.6858656518161297,
+      "num_tokens": 4368704.0,
+      "step": 670
+    },
+    {
+      "epoch": 0.9721229449606862,
+      "grad_norm": 9.739863395690918,
+      "learning_rate": 6.000000000000001e-07,
+      "loss": 1.1052,
+      "mean_token_accuracy": 0.6904201626777648,
+      "num_tokens": 4431384.0,
+      "step": 680
+    },
+    {
+      "epoch": 0.9864188706218727,
+      "grad_norm": 11.182050704956055,
+      "learning_rate": 3.1428571428571433e-07,
+      "loss": 1.1422,
+      "mean_token_accuracy": 0.688240597397089,
+      "num_tokens": 4500182.0,
+      "step": 690
+    },
+    {
+      "epoch": 1.0,
+      "grad_norm": 11.066879272460938,
+      "learning_rate": 2.8571428571428575e-08,
+      "loss": 1.131,
+      "mean_token_accuracy": 0.6868848518321389,
+      "num_tokens": 4559091.0,
+      "step": 700
+    }
+  ],
+  "logging_steps": 10,
+  "max_steps": 700,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 1,
+  "save_steps": 500,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": true
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 4791381278720.0,
+  "train_batch_size": 1,
+  "trial_name": null,
+  "trial_params": null
+}

zero_to_fp32.py

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