Upload folder using huggingface_hub

config.json

@@ -0,0 +1,61 @@
+{
+  "architectures": [
+    "Qwen3Model"
+  ],
+  "attention_bias": false,
+  "attention_dropout": 0.0,
+  "bos_token_id": 151643,
+  "dtype": "bfloat16",
+  "eos_token_id": 151643,
+  "head_dim": 128,
+  "hidden_act": "silu",
+  "hidden_size": 1024,
+  "initializer_range": 0.02,
+  "intermediate_size": 3072,
+  "layer_types": [
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention"
+  ],
+  "max_position_embeddings": 32768,
+  "max_window_layers": 28,
+  "model_type": "qwen3",
+  "num_attention_heads": 16,
+  "num_hidden_layers": 28,
+  "num_key_value_heads": 8,
+  "pad_token_id": 0,
+  "rms_norm_eps": 1e-06,
+  "rope_scaling": null,
+  "rope_theta": 1000000,
+  "sliding_window": null,
+  "tie_word_embeddings": true,
+  "transformers_version": "4.56.2",
+  "use_cache": true,
+  "use_sliding_window": false,
+  "vocab_size": 151669
+}

latest

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

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:281e4c4a547e18d5363f12b1fd0a5822fb8e8ebfbed10ad3acba11fadd3f97b6
+size 1191586416

rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:33fe1a45c0111b18df213058c73c3a4e717295b975e92faf7b2e048e6504b3f3
+size 14917

rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4bf26da988f2f17ca0d366aece1dfdb5c3bcab91066168b7062b361b8c3ac2d6
+size 14917

trainer_state.json

@@ -0,0 +1,1434 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.273117937291103,
+  "eval_steps": 500,
+  "global_step": 2000,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.006366385484641095,
+      "grad_norm": 28.7295833201599,
+      "learning_rate": 0.00011219390703061137,
+      "loss": 6.6223,
+      "step": 10
+    },
+    {
+      "epoch": 0.01273277096928219,
+      "grad_norm": 15.081303427956621,
+      "learning_rate": 0.00014596763837756142,
+      "loss": 2.7812,
+      "step": 20
+    },
+    {
+      "epoch": 0.019099156453923283,
+      "grad_norm": 17.081283324526982,
+      "learning_rate": 0.00016572400472495783,
+      "loss": 2.6574,
+      "step": 30
+    },
+    {
+      "epoch": 0.02546554193856438,
+      "grad_norm": 15.201762160364256,
+      "learning_rate": 0.00017974136972451145,
+      "loss": 2.9039,
+      "step": 40
+    },
+    {
+      "epoch": 0.031831927423205474,
+      "grad_norm": 11.557044296106628,
+      "learning_rate": 0.0001906140827142727,
+      "loss": 2.9066,
+      "step": 50
+    },
+    {
+      "epoch": 0.03819831290784657,
+      "grad_norm": 14.915432761425704,
+      "learning_rate": 0.00019949773607190786,
+      "loss": 2.8914,
+      "step": 60
+    },
+    {
+      "epoch": 0.04456469839248767,
+      "grad_norm": 8.745902020340656,
+      "learning_rate": 0.0002070087579637228,
+      "loss": 2.9014,
+      "step": 70
+    },
+    {
+      "epoch": 0.05093108387712876,
+      "grad_norm": 8.857658675771324,
+      "learning_rate": 0.0002135151010714615,
+      "loss": 2.9293,
+      "step": 80
+    },
+    {
+      "epoch": 0.057297469361769854,
+      "grad_norm": 11.48593285000581,
+      "learning_rate": 0.0002192541024193043,
+      "loss": 2.8625,
+      "step": 90
+    },
+    {
+      "epoch": 0.06366385484641095,
+      "grad_norm": 7.815960040918993,
+      "learning_rate": 0.00022438781406122275,
+      "loss": 3.0063,
+      "step": 100
+    },
+    {
+      "epoch": 0.07003024033105204,
+      "grad_norm": 8.632620514774269,
+      "learning_rate": 0.00022903182113161202,
+      "loss": 2.9527,
+      "step": 110
+    },
+    {
+      "epoch": 0.07639662581569313,
+      "grad_norm": 7.955348207734164,
+      "learning_rate": 0.00023327146741885792,
+      "loss": 3.0082,
+      "step": 120
+    },
+    {
+      "epoch": 0.08276301130033424,
+      "grad_norm": 7.216941061016277,
+      "learning_rate": 0.00023717156393669215,
+      "loss": 2.7871,
+      "step": 130
+    },
+    {
+      "epoch": 0.08912939678497533,
+      "grad_norm": 5.822424121539524,
+      "learning_rate": 0.00024078248931067283,
+      "loss": 2.7844,
+      "step": 140
+    },
+    {
+      "epoch": 0.09549578226961643,
+      "grad_norm": 7.906464628684026,
+      "learning_rate": 0.00024414418040861915,
+      "loss": 2.8514,
+      "step": 150
+    },
+    {
+      "epoch": 0.10186216775425752,
+      "grad_norm": 7.514192904534888,
+      "learning_rate": 0.00024728883241841157,
+      "loss": 3.0086,
+      "step": 160
+    },
+    {
+      "epoch": 0.10822855323889861,
+      "grad_norm": 6.425203024033055,
+      "learning_rate": 0.0002502427789216415,
+      "loss": 2.8754,
+      "step": 170
+    },
+    {
+      "epoch": 0.11459493872353971,
+      "grad_norm": 7.088171053123606,
+      "learning_rate": 0.00025302783376625435,
+      "loss": 2.9094,
+      "step": 180
+    },
+    {
+      "epoch": 0.1209613242081808,
+      "grad_norm": 6.538392044337595,
+      "learning_rate": 0.00025566226965097254,
+      "loss": 2.9262,
+      "step": 190
+    },
+    {
+      "epoch": 0.1273277096928219,
+      "grad_norm": 8.672338929603391,
+      "learning_rate": 0.0002581615454081728,
+      "loss": 2.8391,
+      "step": 200
+    },
+    {
+      "epoch": 0.133694095177463,
+      "grad_norm": 7.204443295582082,
+      "learning_rate": 0.00026053885565806924,
+      "loss": 2.9121,
+      "step": 210
+    },
+    {
+      "epoch": 0.14006048066210408,
+      "grad_norm": 6.463760927732387,
+      "learning_rate": 0.0002628055524785621,
+      "loss": 2.9268,
+      "step": 220
+    },
+    {
+      "epoch": 0.1464268661467452,
+      "grad_norm": 6.869739493367925,
+      "learning_rate": 0.0002649714732657648,
+      "loss": 2.8965,
+      "step": 230
+    },
+    {
+      "epoch": 0.15279325163138627,
+      "grad_norm": 5.613452359830457,
+      "learning_rate": 0.00026704519876580795,
+      "loss": 2.9688,
+      "step": 240
+    },
+    {
+      "epoch": 0.15915963711602737,
+      "grad_norm": 5.7575322101372866,
+      "learning_rate": 0.000269034258397934,
+      "loss": 2.9008,
+      "step": 250
+    },
+    {
+      "epoch": 0.16552602260066848,
+      "grad_norm": 7.337481922102622,
+      "learning_rate": 0.00027094529528364224,
+      "loss": 2.9398,
+      "step": 260
+    },
+    {
+      "epoch": 0.17189240808530956,
+      "grad_norm": 6.95242288679829,
+      "learning_rate": 0.00027278420011365073,
+      "loss": 2.8973,
+      "step": 270
+    },
+    {
+      "epoch": 0.17825879356995067,
+      "grad_norm": 4.554710482953035,
+      "learning_rate": 0.00027455622065762283,
+      "loss": 2.9867,
+      "step": 280
+    },
+    {
+      "epoch": 0.18462517905459175,
+      "grad_norm": 7.670629856220385,
+      "learning_rate": 0.00027626605204863905,
+      "loss": 2.9191,
+      "step": 290
+    },
+    {
+      "epoch": 0.19099156453923286,
+      "grad_norm": 6.140158435864239,
+      "learning_rate": 0.0002779179117555692,
+      "loss": 2.8199,
+      "step": 300
+    },
+    {
+      "epoch": 0.19735795002387393,
+      "grad_norm": 6.06880330501044,
+      "learning_rate": 0.00027951560225766885,
+      "loss": 2.959,
+      "step": 310
+    },
+    {
+      "epoch": 0.20372433550851504,
+      "grad_norm": 5.932458703590213,
+      "learning_rate": 0.0002810625637653616,
+      "loss": 2.9984,
+      "step": 320
+    },
+    {
+      "epoch": 0.21009072099315615,
+      "grad_norm": 5.413884214713577,
+      "learning_rate": 0.0002825619188259585,
+      "loss": 2.8937,
+      "step": 330
+    },
+    {
+      "epoch": 0.21645710647779723,
+      "grad_norm": 5.6133458231383315,
+      "learning_rate": 0.00028401651026859154,
+      "loss": 2.8844,
+      "step": 340
+    },
+    {
+      "epoch": 0.22282349196243834,
+      "grad_norm": 6.157515185075721,
+      "learning_rate": 0.00028542893364738413,
+      "loss": 2.9137,
+      "step": 350
+    },
+    {
+      "epoch": 0.22918987744707942,
+      "grad_norm": 6.150435968994946,
+      "learning_rate": 0.0002868015651132044,
+      "loss": 2.8086,
+      "step": 360
+    },
+    {
+      "epoch": 0.23555626293172052,
+      "grad_norm": 6.1391865377467205,
+      "learning_rate": 0.00028813658546582825,
+      "loss": 2.9246,
+      "step": 370
+    },
+    {
+      "epoch": 0.2419226484163616,
+      "grad_norm": 4.720212170517568,
+      "learning_rate": 0.0002894360009979226,
+      "loss": 2.7273,
+      "step": 380
+    },
+    {
+      "epoch": 0.2482890339010027,
+      "grad_norm": 5.221438084968306,
+      "learning_rate": 0.0002907016616310386,
+      "loss": 2.8332,
+      "step": 390
+    },
+    {
+      "epoch": 0.2546554193856438,
+      "grad_norm": 5.397389581466363,
+      "learning_rate": 0.00029193527675512284,
+      "loss": 2.776,
+      "step": 400
+    },
+    {
+      "epoch": 0.26102180487028487,
+      "grad_norm": 7.525981812223918,
+      "learning_rate": 0.0002931384291118962,
+      "loss": 2.8727,
+      "step": 410
+    },
+    {
+      "epoch": 0.267388190354926,
+      "grad_norm": 5.7116155190893805,
+      "learning_rate": 0.00029431258700501927,
+      "loss": 2.9766,
+      "step": 420
+    },
+    {
+      "epoch": 0.2737545758395671,
+      "grad_norm": 7.012023587035556,
+      "learning_rate": 0.00029545911507334384,
+      "loss": 2.7902,
+      "step": 430
+    },
+    {
+      "epoch": 0.28012096132420816,
+      "grad_norm": 4.8865448559050435,
+      "learning_rate": 0.0002965792838255121,
+      "loss": 2.6873,
+      "step": 440
+    },
+    {
+      "epoch": 0.2864873468088493,
+      "grad_norm": 4.682656054807096,
+      "learning_rate": 0.0002976742781029656,
+      "loss": 2.8498,
+      "step": 450
+    },
+    {
+      "epoch": 0.2928537322934904,
+      "grad_norm": 4.114956894922335,
+      "learning_rate": 0.00029874520461271485,
+      "loss": 2.7461,
+      "step": 460
+    },
+    {
+      "epoch": 0.29922011777813146,
+      "grad_norm": 4.388435500348782,
+      "learning_rate": 0.0002997930986499357,
+      "loss": 2.9004,
+      "step": 470
+    },
+    {
+      "epoch": 0.30558650326277254,
+      "grad_norm": 4.06992963685517,
+      "learning_rate": 0.0002995048337656213,
+      "loss": 2.8203,
+      "step": 480
+    },
+    {
+      "epoch": 0.31195288874741367,
+      "grad_norm": 5.7122354924160375,
+      "learning_rate": 0.0002987974534307946,
+      "loss": 2.8477,
+      "step": 490
+    },
+    {
+      "epoch": 0.31831927423205475,
+      "grad_norm": 4.543743759310822,
+      "learning_rate": 0.0002980900730959679,
+      "loss": 2.8168,
+      "step": 500
+    },
+    {
+      "epoch": 0.32468565971669583,
+      "grad_norm": 5.2449822235737384,
+      "learning_rate": 0.0002973826927611412,
+      "loss": 2.7021,
+      "step": 510
+    },
+    {
+      "epoch": 0.33105204520133696,
+      "grad_norm": 6.42302117474529,
+      "learning_rate": 0.0002966753124263145,
+      "loss": 2.8545,
+      "step": 520
+    },
+    {
+      "epoch": 0.33741843068597804,
+      "grad_norm": 6.4116258183908945,
+      "learning_rate": 0.0002959679320914878,
+      "loss": 2.8436,
+      "step": 530
+    },
+    {
+      "epoch": 0.3437848161706191,
+      "grad_norm": 4.982523320381435,
+      "learning_rate": 0.00029526055175666113,
+      "loss": 2.7607,
+      "step": 540
+    },
+    {
+      "epoch": 0.3501512016552602,
+      "grad_norm": 4.815920229901925,
+      "learning_rate": 0.00029455317142183443,
+      "loss": 2.734,
+      "step": 550
+    },
+    {
+      "epoch": 0.35651758713990134,
+      "grad_norm": 9.300484207985622,
+      "learning_rate": 0.00029384579108700774,
+      "loss": 2.7889,
+      "step": 560
+    },
+    {
+      "epoch": 0.3628839726245424,
+      "grad_norm": 4.5559522343853285,
+      "learning_rate": 0.00029313841075218105,
+      "loss": 2.7531,
+      "step": 570
+    },
+    {
+      "epoch": 0.3692503581091835,
+      "grad_norm": 5.351641807589207,
+      "learning_rate": 0.0002924310304173544,
+      "loss": 2.7451,
+      "step": 580
+    },
+    {
+      "epoch": 0.37561674359382463,
+      "grad_norm": 5.696588702079196,
+      "learning_rate": 0.00029172365008252766,
+      "loss": 2.7189,
+      "step": 590
+    },
+    {
+      "epoch": 0.3819831290784657,
+      "grad_norm": 4.694068773418385,
+      "learning_rate": 0.00029101626974770096,
+      "loss": 2.6096,
+      "step": 600
+    },
+    {
+      "epoch": 0.3883495145631068,
+      "grad_norm": 9.013555425033168,
+      "learning_rate": 0.0002903088894128743,
+      "loss": 2.6605,
+      "step": 610
+    },
+    {
+      "epoch": 0.39471590004774787,
+      "grad_norm": 10.964569620769009,
+      "learning_rate": 0.00028960150907804763,
+      "loss": 2.7258,
+      "step": 620
+    },
+    {
+      "epoch": 0.401082285532389,
+      "grad_norm": 8.94321766980447,
+      "learning_rate": 0.00028889412874322093,
+      "loss": 2.6688,
+      "step": 630
+    },
+    {
+      "epoch": 0.4074486710170301,
+      "grad_norm": 10.742420289850541,
+      "learning_rate": 0.0002881867484083942,
+      "loss": 2.573,
+      "step": 640
+    },
+    {
+      "epoch": 0.41381505650167116,
+      "grad_norm": 8.890111021830462,
+      "learning_rate": 0.00028747936807356755,
+      "loss": 2.5846,
+      "step": 650
+    },
+    {
+      "epoch": 0.4201814419863123,
+      "grad_norm": 27.744833393643688,
+      "learning_rate": 0.00028677198773874085,
+      "loss": 2.5406,
+      "step": 660
+    },
+    {
+      "epoch": 0.4265478274709534,
+      "grad_norm": 8.283239589638123,
+      "learning_rate": 0.00028606460740391416,
+      "loss": 2.602,
+      "step": 670
+    },
+    {
+      "epoch": 0.43291421295559446,
+      "grad_norm": 7.738262298672388,
+      "learning_rate": 0.00028535722706908746,
+      "loss": 2.6947,
+      "step": 680
+    },
+    {
+      "epoch": 0.43928059844023554,
+      "grad_norm": 11.536106786052837,
+      "learning_rate": 0.00028464984673426077,
+      "loss": 2.6422,
+      "step": 690
+    },
+    {
+      "epoch": 0.44564698392487667,
+      "grad_norm": 7.93060742031869,
+      "learning_rate": 0.0002839424663994341,
+      "loss": 2.5859,
+      "step": 700
+    },
+    {
+      "epoch": 0.45201336940951775,
+      "grad_norm": 4.031662354294752,
+      "learning_rate": 0.0002832350860646074,
+      "loss": 2.5947,
+      "step": 710
+    },
+    {
+      "epoch": 0.45837975489415883,
+      "grad_norm": 5.258278074052536,
+      "learning_rate": 0.0002825277057297807,
+      "loss": 2.64,
+      "step": 720
+    },
+    {
+      "epoch": 0.4647461403787999,
+      "grad_norm": 4.323834597980534,
+      "learning_rate": 0.000281820325394954,
+      "loss": 2.6949,
+      "step": 730
+    },
+    {
+      "epoch": 0.47111252586344105,
+      "grad_norm": 5.0167173121086135,
+      "learning_rate": 0.0002811129450601273,
+      "loss": 2.6221,
+      "step": 740
+    },
+    {
+      "epoch": 0.4774789113480821,
+      "grad_norm": 5.933856548166221,
+      "learning_rate": 0.0002804055647253006,
+      "loss": 2.632,
+      "step": 750
+    },
+    {
+      "epoch": 0.4838452968327232,
+      "grad_norm": 5.090852518324657,
+      "learning_rate": 0.0002796981843904739,
+      "loss": 2.8242,
+      "step": 760
+    },
+    {
+      "epoch": 0.49021168231736434,
+      "grad_norm": 4.60199924467046,
+      "learning_rate": 0.0002789908040556472,
+      "loss": 2.6357,
+      "step": 770
+    },
+    {
+      "epoch": 0.4965780678020054,
+      "grad_norm": 4.62334978934682,
+      "learning_rate": 0.0002782834237208206,
+      "loss": 2.6455,
+      "step": 780
+    },
+    {
+      "epoch": 0.5029444532866465,
+      "grad_norm": 5.7098868621282834,
+      "learning_rate": 0.0002775760433859938,
+      "loss": 2.7822,
+      "step": 790
+    },
+    {
+      "epoch": 0.5093108387712876,
+      "grad_norm": 5.492909395839608,
+      "learning_rate": 0.00027686866305116713,
+      "loss": 2.5965,
+      "step": 800
+    },
+    {
+      "epoch": 0.5156772242559287,
+      "grad_norm": 5.007751033932614,
+      "learning_rate": 0.00027616128271634044,
+      "loss": 2.6355,
+      "step": 810
+    },
+    {
+      "epoch": 0.5220436097405697,
+      "grad_norm": 6.170648429353342,
+      "learning_rate": 0.0002754539023815138,
+      "loss": 2.6123,
+      "step": 820
+    },
+    {
+      "epoch": 0.5284099952252109,
+      "grad_norm": 4.235113061064731,
+      "learning_rate": 0.0002747465220466871,
+      "loss": 2.6453,
+      "step": 830
+    },
+    {
+      "epoch": 0.534776380709852,
+      "grad_norm": 4.538569485793573,
+      "learning_rate": 0.00027403914171186036,
+      "loss": 2.6713,
+      "step": 840
+    },
+    {
+      "epoch": 0.5411427661944931,
+      "grad_norm": 4.019460059963649,
+      "learning_rate": 0.0002733317613770337,
+      "loss": 2.5943,
+      "step": 850
+    },
+    {
+      "epoch": 0.5475091516791342,
+      "grad_norm": 6.254490760667432,
+      "learning_rate": 0.000272624381042207,
+      "loss": 2.5418,
+      "step": 860
+    },
+    {
+      "epoch": 0.5538755371637752,
+      "grad_norm": 4.520259916472461,
+      "learning_rate": 0.0002719170007073803,
+      "loss": 2.584,
+      "step": 870
+    },
+    {
+      "epoch": 0.5602419226484163,
+      "grad_norm": 5.065731135931019,
+      "learning_rate": 0.0002712096203725536,
+      "loss": 2.6723,
+      "step": 880
+    },
+    {
+      "epoch": 0.5666083081330574,
+      "grad_norm": 3.833075034277501,
+      "learning_rate": 0.00027050224003772694,
+      "loss": 2.5248,
+      "step": 890
+    },
+    {
+      "epoch": 0.5729746936176986,
+      "grad_norm": 5.507080157861899,
+      "learning_rate": 0.00026979485970290024,
+      "loss": 2.6379,
+      "step": 900
+    },
+    {
+      "epoch": 0.5793410791023397,
+      "grad_norm": 14.128248882331409,
+      "learning_rate": 0.00026908747936807355,
+      "loss": 2.5904,
+      "step": 910
+    },
+    {
+      "epoch": 0.5857074645869808,
+      "grad_norm": 6.1778278241357265,
+      "learning_rate": 0.00026838009903324686,
+      "loss": 2.5844,
+      "step": 920
+    },
+    {
+      "epoch": 0.5920738500716218,
+      "grad_norm": 8.064848357512728,
+      "learning_rate": 0.00026767271869842016,
+      "loss": 2.6055,
+      "step": 930
+    },
+    {
+      "epoch": 0.5984402355562629,
+      "grad_norm": 6.16438539022089,
+      "learning_rate": 0.00026696533836359347,
+      "loss": 2.5506,
+      "step": 940
+    },
+    {
+      "epoch": 0.604806621040904,
+      "grad_norm": 5.882124357483614,
+      "learning_rate": 0.0002662579580287668,
+      "loss": 2.4973,
+      "step": 950
+    },
+    {
+      "epoch": 0.6111730065255451,
+      "grad_norm": 8.373991467730166,
+      "learning_rate": 0.0002655505776939401,
+      "loss": 2.5428,
+      "step": 960
+    },
+    {
+      "epoch": 0.6175393920101863,
+      "grad_norm": 6.831051520016273,
+      "learning_rate": 0.0002648431973591134,
+      "loss": 2.4982,
+      "step": 970
+    },
+    {
+      "epoch": 0.6239057774948273,
+      "grad_norm": 5.074869694709693,
+      "learning_rate": 0.0002641358170242867,
+      "loss": 2.5465,
+      "step": 980
+    },
+    {
+      "epoch": 0.6302721629794684,
+      "grad_norm": 5.503484162830985,
+      "learning_rate": 0.00026342843668946,
+      "loss": 2.324,
+      "step": 990
+    },
+    {
+      "epoch": 0.6366385484641095,
+      "grad_norm": 4.385405886090266,
+      "learning_rate": 0.0002627210563546333,
+      "loss": 2.417,
+      "step": 1000
+    },
+    {
+      "epoch": 0.6430049339487506,
+      "grad_norm": 5.034207848822367,
+      "learning_rate": 0.0002620136760198066,
+      "loss": 2.6066,
+      "step": 1010
+    },
+    {
+      "epoch": 0.6493713194333917,
+      "grad_norm": 3.696840611162144,
+      "learning_rate": 0.00026130629568497997,
+      "loss": 2.5441,
+      "step": 1020
+    },
+    {
+      "epoch": 0.6557377049180327,
+      "grad_norm": 4.6314444028768875,
+      "learning_rate": 0.0002605989153501533,
+      "loss": 2.4752,
+      "step": 1030
+    },
+    {
+      "epoch": 0.6621040904026739,
+      "grad_norm": 3.4663340841035093,
+      "learning_rate": 0.0002598915350153265,
+      "loss": 2.4637,
+      "step": 1040
+    },
+    {
+      "epoch": 0.668470475887315,
+      "grad_norm": 8.334602958229153,
+      "learning_rate": 0.00025918415468049983,
+      "loss": 2.6082,
+      "step": 1050
+    },
+    {
+      "epoch": 0.6748368613719561,
+      "grad_norm": 4.575304861734687,
+      "learning_rate": 0.0002584767743456732,
+      "loss": 2.4842,
+      "step": 1060
+    },
+    {
+      "epoch": 0.6812032468565972,
+      "grad_norm": 3.618603659881246,
+      "learning_rate": 0.0002577693940108465,
+      "loss": 2.4982,
+      "step": 1070
+    },
+    {
+      "epoch": 0.6875696323412382,
+      "grad_norm": 4.4740750234847475,
+      "learning_rate": 0.00025706201367601975,
+      "loss": 2.4029,
+      "step": 1080
+    },
+    {
+      "epoch": 0.6939360178258793,
+      "grad_norm": 4.208151138305904,
+      "learning_rate": 0.0002563546333411931,
+      "loss": 2.4836,
+      "step": 1090
+    },
+    {
+      "epoch": 0.7003024033105204,
+      "grad_norm": 5.736418133529195,
+      "learning_rate": 0.0002556472530063664,
+      "loss": 2.6402,
+      "step": 1100
+    },
+    {
+      "epoch": 0.7066687887951616,
+      "grad_norm": 3.6852138552936395,
+      "learning_rate": 0.0002549398726715397,
+      "loss": 2.4941,
+      "step": 1110
+    },
+    {
+      "epoch": 0.7130351742798027,
+      "grad_norm": 4.11339792490071,
+      "learning_rate": 0.000254232492336713,
+      "loss": 2.4641,
+      "step": 1120
+    },
+    {
+      "epoch": 0.7194015597644438,
+      "grad_norm": 3.7487169212790477,
+      "learning_rate": 0.00025352511200188633,
+      "loss": 2.3414,
+      "step": 1130
+    },
+    {
+      "epoch": 0.7257679452490848,
+      "grad_norm": 3.442251781584789,
+      "learning_rate": 0.00025281773166705964,
+      "loss": 2.3932,
+      "step": 1140
+    },
+    {
+      "epoch": 0.7321343307337259,
+      "grad_norm": 3.8470452186807846,
+      "learning_rate": 0.00025211035133223294,
+      "loss": 2.4334,
+      "step": 1150
+    },
+    {
+      "epoch": 0.738500716218367,
+      "grad_norm": 4.506628903442333,
+      "learning_rate": 0.00025140297099740625,
+      "loss": 2.3195,
+      "step": 1160
+    },
+    {
+      "epoch": 0.7448671017030081,
+      "grad_norm": 4.325389792461109,
+      "learning_rate": 0.00025069559066257955,
+      "loss": 2.4027,
+      "step": 1170
+    },
+    {
+      "epoch": 0.7512334871876493,
+      "grad_norm": 3.8669860452410445,
+      "learning_rate": 0.00024998821032775286,
+      "loss": 2.4686,
+      "step": 1180
+    },
+    {
+      "epoch": 0.7575998726722903,
+      "grad_norm": 3.561694549590622,
+      "learning_rate": 0.00024928082999292617,
+      "loss": 2.5395,
+      "step": 1190
+    },
+    {
+      "epoch": 0.7639662581569314,
+      "grad_norm": 4.156976510059248,
+      "learning_rate": 0.00024857344965809947,
+      "loss": 2.4416,
+      "step": 1200
+    },
+    {
+      "epoch": 0.7703326436415725,
+      "grad_norm": 5.616798309964706,
+      "learning_rate": 0.0002478660693232728,
+      "loss": 2.4055,
+      "step": 1210
+    },
+    {
+      "epoch": 0.7766990291262136,
+      "grad_norm": 3.149306910194074,
+      "learning_rate": 0.0002471586889884461,
+      "loss": 2.4488,
+      "step": 1220
+    },
+    {
+      "epoch": 0.7830654146108547,
+      "grad_norm": 3.0254085146949974,
+      "learning_rate": 0.00024645130865361944,
+      "loss": 2.44,
+      "step": 1230
+    },
+    {
+      "epoch": 0.7894318000954957,
+      "grad_norm": 3.3553234269170815,
+      "learning_rate": 0.0002457439283187927,
+      "loss": 2.4689,
+      "step": 1240
+    },
+    {
+      "epoch": 0.7957981855801369,
+      "grad_norm": 3.560968004340506,
+      "learning_rate": 0.000245036547983966,
+      "loss": 2.4328,
+      "step": 1250
+    },
+    {
+      "epoch": 0.802164571064778,
+      "grad_norm": 3.670143268942363,
+      "learning_rate": 0.00024432916764913936,
+      "loss": 2.357,
+      "step": 1260
+    },
+    {
+      "epoch": 0.8085309565494191,
+      "grad_norm": 4.603180595509743,
+      "learning_rate": 0.00024362178731431264,
+      "loss": 2.4336,
+      "step": 1270
+    },
+    {
+      "epoch": 0.8148973420340602,
+      "grad_norm": 3.2173958923931845,
+      "learning_rate": 0.00024291440697948595,
+      "loss": 2.3859,
+      "step": 1280
+    },
+    {
+      "epoch": 0.8212637275187012,
+      "grad_norm": 4.238015371838477,
+      "learning_rate": 0.00024220702664465925,
+      "loss": 2.4234,
+      "step": 1290
+    },
+    {
+      "epoch": 0.8276301130033423,
+      "grad_norm": 3.350821041839247,
+      "learning_rate": 0.00024149964630983258,
+      "loss": 2.34,
+      "step": 1300
+    },
+    {
+      "epoch": 0.8339964984879834,
+      "grad_norm": 3.956822835287818,
+      "learning_rate": 0.00024079226597500586,
+      "loss": 2.409,
+      "step": 1310
+    },
+    {
+      "epoch": 0.8403628839726246,
+      "grad_norm": 3.206745954274851,
+      "learning_rate": 0.00024008488564017917,
+      "loss": 2.4766,
+      "step": 1320
+    },
+    {
+      "epoch": 0.8467292694572657,
+      "grad_norm": 3.16798874537864,
+      "learning_rate": 0.0002393775053053525,
+      "loss": 2.4918,
+      "step": 1330
+    },
+    {
+      "epoch": 0.8530956549419068,
+      "grad_norm": 3.2084119932227613,
+      "learning_rate": 0.0002386701249705258,
+      "loss": 2.3279,
+      "step": 1340
+    },
+    {
+      "epoch": 0.8594620404265478,
+      "grad_norm": 2.6626163554970694,
+      "learning_rate": 0.00023796274463569909,
+      "loss": 2.3762,
+      "step": 1350
+    },
+    {
+      "epoch": 0.8658284259111889,
+      "grad_norm": 3.543568104802714,
+      "learning_rate": 0.00023725536430087242,
+      "loss": 2.4406,
+      "step": 1360
+    },
+    {
+      "epoch": 0.87219481139583,
+      "grad_norm": 3.0950338366546832,
+      "learning_rate": 0.00023654798396604572,
+      "loss": 2.4135,
+      "step": 1370
+    },
+    {
+      "epoch": 0.8785611968804711,
+      "grad_norm": 3.4459117956268583,
+      "learning_rate": 0.00023584060363121903,
+      "loss": 2.3371,
+      "step": 1380
+    },
+    {
+      "epoch": 0.8849275823651122,
+      "grad_norm": 3.3339357175581026,
+      "learning_rate": 0.00023513322329639234,
+      "loss": 2.3568,
+      "step": 1390
+    },
+    {
+      "epoch": 0.8912939678497533,
+      "grad_norm": 2.7770602679788836,
+      "learning_rate": 0.00023442584296156567,
+      "loss": 2.4703,
+      "step": 1400
+    },
+    {
+      "epoch": 0.8976603533343944,
+      "grad_norm": 3.1587359306963925,
+      "learning_rate": 0.00023371846262673895,
+      "loss": 2.3687,
+      "step": 1410
+    },
+    {
+      "epoch": 0.9040267388190355,
+      "grad_norm": 3.6463373041615057,
+      "learning_rate": 0.00023301108229191225,
+      "loss": 2.5037,
+      "step": 1420
+    },
+    {
+      "epoch": 0.9103931243036766,
+      "grad_norm": 3.8799007179484066,
+      "learning_rate": 0.00023230370195708559,
+      "loss": 2.3883,
+      "step": 1430
+    },
+    {
+      "epoch": 0.9167595097883177,
+      "grad_norm": 2.650501985473451,
+      "learning_rate": 0.0002315963216222589,
+      "loss": 2.3818,
+      "step": 1440
+    },
+    {
+      "epoch": 0.9231258952729587,
+      "grad_norm": 3.3575457214628353,
+      "learning_rate": 0.00023088894128743217,
+      "loss": 2.2014,
+      "step": 1450
+    },
+    {
+      "epoch": 0.9294922807575998,
+      "grad_norm": 4.160903578127555,
+      "learning_rate": 0.00023018156095260548,
+      "loss": 2.3586,
+      "step": 1460
+    },
+    {
+      "epoch": 0.935858666242241,
+      "grad_norm": 3.7275531046661805,
+      "learning_rate": 0.0002294741806177788,
+      "loss": 2.4684,
+      "step": 1470
+    },
+    {
+      "epoch": 0.9422250517268821,
+      "grad_norm": 3.5804971221330515,
+      "learning_rate": 0.00022876680028295211,
+      "loss": 2.3984,
+      "step": 1480
+    },
+    {
+      "epoch": 0.9485914372115232,
+      "grad_norm": 2.903402028010888,
+      "learning_rate": 0.00022805941994812542,
+      "loss": 2.3508,
+      "step": 1490
+    },
+    {
+      "epoch": 0.9549578226961642,
+      "grad_norm": 3.2883720153819804,
+      "learning_rate": 0.00022735203961329875,
+      "loss": 2.468,
+      "step": 1500
+    },
+    {
+      "epoch": 0.9613242081808053,
+      "grad_norm": 3.165068282067837,
+      "learning_rate": 0.00022664465927847203,
+      "loss": 2.3387,
+      "step": 1510
+    },
+    {
+      "epoch": 0.9676905936654464,
+      "grad_norm": 2.8318847580527895,
+      "learning_rate": 0.00022593727894364534,
+      "loss": 2.235,
+      "step": 1520
+    },
+    {
+      "epoch": 0.9740569791500875,
+      "grad_norm": 3.0627223359274556,
+      "learning_rate": 0.00022522989860881867,
+      "loss": 2.4135,
+      "step": 1530
+    },
+    {
+      "epoch": 0.9804233646347287,
+      "grad_norm": 4.63667156924246,
+      "learning_rate": 0.00022452251827399198,
+      "loss": 2.3891,
+      "step": 1540
+    },
+    {
+      "epoch": 0.9867897501193698,
+      "grad_norm": 2.7155003830350735,
+      "learning_rate": 0.00022381513793916526,
+      "loss": 2.2814,
+      "step": 1550
+    },
+    {
+      "epoch": 0.9931561356040108,
+      "grad_norm": 3.743790594304019,
+      "learning_rate": 0.00022310775760433856,
+      "loss": 2.1729,
+      "step": 1560
+    },
+    {
+      "epoch": 0.9995225210886519,
+      "grad_norm": 3.0964479659673336,
+      "learning_rate": 0.0002224003772695119,
+      "loss": 2.3365,
+      "step": 1570
+    },
+    {
+      "epoch": 1.005729746936177,
+      "grad_norm": 3.9150118918426307,
+      "learning_rate": 0.0002216929969346852,
+      "loss": 1.9852,
+      "step": 1580
+    },
+    {
+      "epoch": 1.012096132420818,
+      "grad_norm": 3.5380613834183614,
+      "learning_rate": 0.0002209856165998585,
+      "loss": 2.0525,
+      "step": 1590
+    },
+    {
+      "epoch": 1.0184625179054592,
+      "grad_norm": 3.1487817928894613,
+      "learning_rate": 0.00022027823626503184,
+      "loss": 1.9064,
+      "step": 1600
+    },
+    {
+      "epoch": 1.0248289033901004,
+      "grad_norm": 3.266369156342291,
+      "learning_rate": 0.00021957085593020512,
+      "loss": 1.8943,
+      "step": 1610
+    },
+    {
+      "epoch": 1.0311952888747413,
+      "grad_norm": 3.4336556947723307,
+      "learning_rate": 0.00021886347559537842,
+      "loss": 1.7861,
+      "step": 1620
+    },
+    {
+      "epoch": 1.0375616743593825,
+      "grad_norm": 3.832267113304745,
+      "learning_rate": 0.00021815609526055173,
+      "loss": 2.0424,
+      "step": 1630
+    },
+    {
+      "epoch": 1.0439280598440235,
+      "grad_norm": 4.130087016761162,
+      "learning_rate": 0.00021744871492572506,
+      "loss": 1.9139,
+      "step": 1640
+    },
+    {
+      "epoch": 1.0502944453286647,
+      "grad_norm": 2.7335284776977886,
+      "learning_rate": 0.00021674133459089834,
+      "loss": 2.0488,
+      "step": 1650
+    },
+    {
+      "epoch": 1.0566608308133056,
+      "grad_norm": 3.329599622048459,
+      "learning_rate": 0.00021603395425607165,
+      "loss": 1.9137,
+      "step": 1660
+    },
+    {
+      "epoch": 1.0630272162979468,
+      "grad_norm": 3.6314777543543313,
+      "learning_rate": 0.00021532657392124498,
+      "loss": 1.9768,
+      "step": 1670
+    },
+    {
+      "epoch": 1.069393601782588,
+      "grad_norm": 4.218111606262662,
+      "learning_rate": 0.00021461919358641828,
+      "loss": 2.0141,
+      "step": 1680
+    },
+    {
+      "epoch": 1.075759987267229,
+      "grad_norm": 2.895306844768361,
+      "learning_rate": 0.0002139118132515916,
+      "loss": 1.9783,
+      "step": 1690
+    },
+    {
+      "epoch": 1.0821263727518702,
+      "grad_norm": 3.753588342038039,
+      "learning_rate": 0.00021320443291676492,
+      "loss": 2.0227,
+      "step": 1700
+    },
+    {
+      "epoch": 1.0884927582365111,
+      "grad_norm": 3.250518576003523,
+      "learning_rate": 0.0002124970525819382,
+      "loss": 1.8848,
+      "step": 1710
+    },
+    {
+      "epoch": 1.0948591437211523,
+      "grad_norm": 2.980988290019129,
+      "learning_rate": 0.0002117896722471115,
+      "loss": 1.867,
+      "step": 1720
+    },
+    {
+      "epoch": 1.1012255292057933,
+      "grad_norm": 3.3311589155605263,
+      "learning_rate": 0.0002110822919122848,
+      "loss": 1.8766,
+      "step": 1730
+    },
+    {
+      "epoch": 1.1075919146904345,
+      "grad_norm": 3.2929109952092523,
+      "learning_rate": 0.00021037491157745815,
+      "loss": 2.0685,
+      "step": 1740
+    },
+    {
+      "epoch": 1.1139583001750757,
+      "grad_norm": 3.8495434375573243,
+      "learning_rate": 0.00020966753124263143,
+      "loss": 1.8662,
+      "step": 1750
+    },
+    {
+      "epoch": 1.1203246856597167,
+      "grad_norm": 3.223669203555661,
+      "learning_rate": 0.00020896015090780473,
+      "loss": 1.9401,
+      "step": 1760
+    },
+    {
+      "epoch": 1.1266910711443578,
+      "grad_norm": 3.068725890159831,
+      "learning_rate": 0.00020825277057297806,
+      "loss": 1.9732,
+      "step": 1770
+    },
+    {
+      "epoch": 1.1330574566289988,
+      "grad_norm": 3.7652954958158724,
+      "learning_rate": 0.00020754539023815137,
+      "loss": 1.9264,
+      "step": 1780
+    },
+    {
+      "epoch": 1.13942384211364,
+      "grad_norm": 3.283263110696873,
+      "learning_rate": 0.00020683800990332468,
+      "loss": 1.8805,
+      "step": 1790
+    },
+    {
+      "epoch": 1.145790227598281,
+      "grad_norm": 3.578786877708257,
+      "learning_rate": 0.00020613062956849795,
+      "loss": 1.8764,
+      "step": 1800
+    },
+    {
+      "epoch": 1.1521566130829222,
+      "grad_norm": 3.2073175602979402,
+      "learning_rate": 0.0002054232492336713,
+      "loss": 1.9123,
+      "step": 1810
+    },
+    {
+      "epoch": 1.1585229985675634,
+      "grad_norm": 3.960679432169151,
+      "learning_rate": 0.0002047158688988446,
+      "loss": 1.8938,
+      "step": 1820
+    },
+    {
+      "epoch": 1.1648893840522043,
+      "grad_norm": 3.026836467606075,
+      "learning_rate": 0.0002040084885640179,
+      "loss": 1.9291,
+      "step": 1830
+    },
+    {
+      "epoch": 1.1712557695368455,
+      "grad_norm": 3.368777784279614,
+      "learning_rate": 0.00020330110822919123,
+      "loss": 1.9459,
+      "step": 1840
+    },
+    {
+      "epoch": 1.1776221550214865,
+      "grad_norm": 2.843063910159576,
+      "learning_rate": 0.0002025937278943645,
+      "loss": 1.9449,
+      "step": 1850
+    },
+    {
+      "epoch": 1.1839885405061277,
+      "grad_norm": 2.9510127362455876,
+      "learning_rate": 0.00020188634755953782,
+      "loss": 1.9309,
+      "step": 1860
+    },
+    {
+      "epoch": 1.1903549259907686,
+      "grad_norm": 4.545581163313047,
+      "learning_rate": 0.00020117896722471115,
+      "loss": 1.8869,
+      "step": 1870
+    },
+    {
+      "epoch": 1.1967213114754098,
+      "grad_norm": 3.301260365676958,
+      "learning_rate": 0.00020047158688988445,
+      "loss": 1.9613,
+      "step": 1880
+    },
+    {
+      "epoch": 1.203087696960051,
+      "grad_norm": 3.299718394999854,
+      "learning_rate": 0.00019976420655505776,
+      "loss": 2.0072,
+      "step": 1890
+    },
+    {
+      "epoch": 1.209454082444692,
+      "grad_norm": 3.5617690109472444,
+      "learning_rate": 0.00019905682622023104,
+      "loss": 1.9713,
+      "step": 1900
+    },
+    {
+      "epoch": 1.2158204679293332,
+      "grad_norm": 3.1359097580279767,
+      "learning_rate": 0.00019834944588540437,
+      "loss": 1.8154,
+      "step": 1910
+    },
+    {
+      "epoch": 1.2221868534139742,
+      "grad_norm": 3.0984711664236206,
+      "learning_rate": 0.00019764206555057768,
+      "loss": 1.8568,
+      "step": 1920
+    },
+    {
+      "epoch": 1.2285532388986153,
+      "grad_norm": 2.842733095279462,
+      "learning_rate": 0.00019693468521575098,
+      "loss": 1.851,
+      "step": 1930
+    },
+    {
+      "epoch": 1.2349196243832563,
+      "grad_norm": 2.935936669675825,
+      "learning_rate": 0.00019622730488092432,
+      "loss": 1.9604,
+      "step": 1940
+    },
+    {
+      "epoch": 1.2412860098678975,
+      "grad_norm": 3.3716657028276096,
+      "learning_rate": 0.0001955199245460976,
+      "loss": 1.9171,
+      "step": 1950
+    },
+    {
+      "epoch": 1.2476523953525387,
+      "grad_norm": 2.9604032685493133,
+      "learning_rate": 0.0001948125442112709,
+      "loss": 1.8604,
+      "step": 1960
+    },
+    {
+      "epoch": 1.2540187808371797,
+      "grad_norm": 2.7269826434015427,
+      "learning_rate": 0.0001941051638764442,
+      "loss": 1.9076,
+      "step": 1970
+    },
+    {
+      "epoch": 1.2603851663218208,
+      "grad_norm": 4.046544087868519,
+      "learning_rate": 0.00019339778354161754,
+      "loss": 1.8299,
+      "step": 1980
+    },
+    {
+      "epoch": 1.2667515518064618,
+      "grad_norm": 3.12579708206849,
+      "learning_rate": 0.00019269040320679084,
+      "loss": 1.7545,
+      "step": 1990
+    },
+    {
+      "epoch": 1.273117937291103,
+      "grad_norm": 2.7391520471543225,
+      "learning_rate": 0.00019198302287196412,
+      "loss": 1.8973,
+      "step": 2000
+    }
+  ],
+  "logging_steps": 10,
+  "max_steps": 4713,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 3,
+  "save_steps": 1000,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 0.0,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}

training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:12c8365fe88fec2f220ca1fdad64ae61ff90d4f89e41121cd33a79940985af89
+size 8273

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 ZERO_STAGE not 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("Extracting fp32 weights")
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag)
+
+    logger.info("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)