Training in progress, step 2000, checkpoint

checkpoint-2000/config.json

@@ -0,0 +1,82 @@
+{
+  "_name_or_path": "kdcyberdude/w2v-multilingual-v1.0",
+  "activation_dropout": 0.0,
+  "adapter_act": "relu",
+  "adapter_kernel_size": 3,
+  "adapter_stride": 2,
+  "add_adapter": true,
+  "apply_spec_augment": true,
+  "architectures": [
+    "Wav2Vec2BertForCTC"
+  ],
+  "attention_dropout": 0.0,
+  "bos_token_id": 1,
+  "classifier_proj_size": 768,
+  "codevector_dim": 768,
+  "conformer_conv_dropout": 0.1,
+  "contrastive_logits_temperature": 0.1,
+  "conv_depthwise_kernel_size": 31,
+  "ctc_loss_reduction": "mean",
+  "ctc_zero_infinity": true,
+  "diversity_loss_weight": 0.1,
+  "eos_token_id": 2,
+  "feat_proj_dropout": 0.0,
+  "feat_quantizer_dropout": 0.0,
+  "feature_projection_input_dim": 160,
+  "final_dropout": 0.1,
+  "hidden_act": "swish",
+  "hidden_dropout": 0.0,
+  "hidden_size": 1024,
+  "initializer_range": 0.02,
+  "intermediate_size": 4096,
+  "layer_norm_eps": 1e-05,
+  "layerdrop": 0.0,
+  "left_max_position_embeddings": 64,
+  "mask_feature_length": 10,
+  "mask_feature_min_masks": 0,
+  "mask_feature_prob": 0.0,
+  "mask_time_length": 10,
+  "mask_time_min_masks": 2,
+  "mask_time_prob": 0.0,
+  "max_source_positions": 5000,
+  "model_type": "wav2vec2-bert",
+  "num_adapter_layers": 1,
+  "num_attention_heads": 16,
+  "num_codevector_groups": 2,
+  "num_codevectors_per_group": 320,
+  "num_hidden_layers": 24,
+  "num_negatives": 100,
+  "output_hidden_size": 1024,
+  "pad_token_id": 221,
+  "position_embeddings_type": "relative_key",
+  "proj_codevector_dim": 768,
+  "right_max_position_embeddings": 8,
+  "rotary_embedding_base": 10000,
+  "tdnn_dilation": [
+    1,
+    2,
+    3,
+    1,
+    1
+  ],
+  "tdnn_dim": [
+    512,
+    512,
+    512,
+    512,
+    1500
+  ],
+  "tdnn_kernel": [
+    5,
+    3,
+    3,
+    1,
+    1
+  ],
+  "torch_dtype": "float16",
+  "transformers_version": "4.48.2",
+  "use_intermediate_ffn_before_adapter": true,
+  "use_weighted_layer_sum": false,
+  "vocab_size": 224,
+  "xvector_output_dim": 512
+}

checkpoint-2000/global_step2000/mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a96c42ed3f36391f1c2538da9e8c979b7b33d41bc32042b9894e02272a8cff50
+size 1228864248

checkpoint-2000/global_step2000/zero_pp_rank_0_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:af72f105e0998837fc3a4da3366ee8df55b8e91287b3696d7e71bf1b192724d1
+size 3685848510

checkpoint-2000/global_step2000/zero_pp_rank_1_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:09e830400d593b82493ba7e4476e5bce2651bc5219133da1ba8ccd69f234199d
+size 3685847294

checkpoint-2000/latest

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

checkpoint-2000/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:aefe673b5c7a0562ed83534c1cc87ba53c50058c71672495a4bd5920db18e033
+size 1228706240

checkpoint-2000/preprocessor_config.json

@@ -0,0 +1,11 @@
+{
+  "feature_extractor_type": "SeamlessM4TFeatureExtractor",
+  "feature_size": 80,
+  "num_mel_bins": 80,
+  "padding_side": "right",
+  "padding_value": 1,
+  "processor_class": "M4TProcessorWithLM",
+  "return_attention_mask": true,
+  "sampling_rate": 16000,
+  "stride": 2
+}

checkpoint-2000/rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b88c07b734f0394fa3460acdd30bd1c1ecdbec8c4fe935e2aa9d3532dcb67c79
+size 14512

checkpoint-2000/rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:71a6b5b9aad93001c2054719d59a5e0b024eae5f275b39ad8ff21c41444c41cb
+size 14512

checkpoint-2000/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:202cbce0eb1350cae2e3983bd312bf00c5e92864b59fa07e72237ea2d4c77f91
+size 1064

checkpoint-2000/trainer_state.json

@@ -0,0 +1,742 @@
+{
+  "best_metric": 0.23469093535410654,
+  "best_model_checkpoint": "./checkpoints/w2v-multilingual-v1.3/checkpoint-2000",
+  "epoch": 0.1647627416520211,
+  "eval_steps": 2000,
+  "global_step": 2000,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.001647627416520211,
+      "grad_norm": 0.2767059803009033,
+      "learning_rate": 4.9431537320810675e-08,
+      "loss": 0.1923,
+      "step": 20
+    },
+    {
+      "epoch": 0.003295254833040422,
+      "grad_norm": 2.006911039352417,
+      "learning_rate": 1.977261492832427e-07,
+      "loss": 0.2947,
+      "step": 40
+    },
+    {
+      "epoch": 0.004942882249560633,
+      "grad_norm": 0.3714193105697632,
+      "learning_rate": 3.624979403526116e-07,
+      "loss": 0.2955,
+      "step": 60
+    },
+    {
+      "epoch": 0.006590509666080844,
+      "grad_norm": 0.3106062412261963,
+      "learning_rate": 5.272697314219806e-07,
+      "loss": 0.2754,
+      "step": 80
+    },
+    {
+      "epoch": 0.008238137082601054,
+      "grad_norm": 3.2078254222869873,
+      "learning_rate": 6.838029329378811e-07,
+      "loss": 0.3177,
+      "step": 100
+    },
+    {
+      "epoch": 0.009885764499121265,
+      "grad_norm": 0.38503074645996094,
+      "learning_rate": 8.485747240072501e-07,
+      "loss": 0.1905,
+      "step": 120
+    },
+    {
+      "epoch": 0.011533391915641476,
+      "grad_norm": 0.15724419057369232,
+      "learning_rate": 1.013346515076619e-06,
+      "loss": 0.2942,
+      "step": 140
+    },
+    {
+      "epoch": 0.013181019332161687,
+      "grad_norm": 0.5840986967086792,
+      "learning_rate": 1.1781183061459877e-06,
+      "loss": 0.2907,
+      "step": 160
+    },
+    {
+      "epoch": 0.014828646748681899,
+      "grad_norm": 0.23122897744178772,
+      "learning_rate": 1.3346515076618883e-06,
+      "loss": 0.2811,
+      "step": 180
+    },
+    {
+      "epoch": 0.016476274165202108,
+      "grad_norm": 2.7248637676239014,
+      "learning_rate": 1.4994232987312573e-06,
+      "loss": 0.3168,
+      "step": 200
+    },
+    {
+      "epoch": 0.01812390158172232,
+      "grad_norm": 1.0846350193023682,
+      "learning_rate": 1.6641950898006263e-06,
+      "loss": 0.1979,
+      "step": 220
+    },
+    {
+      "epoch": 0.01977152899824253,
+      "grad_norm": 1.7905633449554443,
+      "learning_rate": 1.828966880869995e-06,
+      "loss": 0.2656,
+      "step": 240
+    },
+    {
+      "epoch": 0.021419156414762743,
+      "grad_norm": 0.31305554509162903,
+      "learning_rate": 1.993738671939364e-06,
+      "loss": 0.272,
+      "step": 260
+    },
+    {
+      "epoch": 0.023066783831282953,
+      "grad_norm": 0.31258106231689453,
+      "learning_rate": 2.158510463008733e-06,
+      "loss": 0.2693,
+      "step": 280
+    },
+    {
+      "epoch": 0.024714411247803162,
+      "grad_norm": 1.783349871635437,
+      "learning_rate": 2.323282254078102e-06,
+      "loss": 0.3116,
+      "step": 300
+    },
+    {
+      "epoch": 0.026362038664323375,
+      "grad_norm": 0.6936488747596741,
+      "learning_rate": 2.488054045147471e-06,
+      "loss": 0.1931,
+      "step": 320
+    },
+    {
+      "epoch": 0.028009666080843584,
+      "grad_norm": 0.6185577511787415,
+      "learning_rate": 2.65282583621684e-06,
+      "loss": 0.258,
+      "step": 340
+    },
+    {
+      "epoch": 0.029657293497363797,
+      "grad_norm": 0.3592207729816437,
+      "learning_rate": 2.8175976272862085e-06,
+      "loss": 0.2401,
+      "step": 360
+    },
+    {
+      "epoch": 0.03130492091388401,
+      "grad_norm": 1.2324920892715454,
+      "learning_rate": 2.982369418355578e-06,
+      "loss": 0.2371,
+      "step": 380
+    },
+    {
+      "epoch": 0.032952548330404216,
+      "grad_norm": 1.5560214519500732,
+      "learning_rate": 3.147141209424947e-06,
+      "loss": 0.2414,
+      "step": 400
+    },
+    {
+      "epoch": 0.03460017574692443,
+      "grad_norm": 0.8133947253227234,
+      "learning_rate": 3.3119130004943154e-06,
+      "loss": 0.1744,
+      "step": 420
+    },
+    {
+      "epoch": 0.03624780316344464,
+      "grad_norm": 0.8599107265472412,
+      "learning_rate": 3.4766847915636844e-06,
+      "loss": 0.2536,
+      "step": 440
+    },
+    {
+      "epoch": 0.03789543057996485,
+      "grad_norm": 0.3699595034122467,
+      "learning_rate": 3.641456582633054e-06,
+      "loss": 0.2552,
+      "step": 460
+    },
+    {
+      "epoch": 0.03954305799648506,
+      "grad_norm": 1.2955116033554077,
+      "learning_rate": 3.797989784148954e-06,
+      "loss": 0.2532,
+      "step": 480
+    },
+    {
+      "epoch": 0.041190685413005274,
+      "grad_norm": 11.518170356750488,
+      "learning_rate": 3.962761575218322e-06,
+      "loss": 0.2507,
+      "step": 500
+    },
+    {
+      "epoch": 0.042838312829525486,
+      "grad_norm": 0.9779248833656311,
+      "learning_rate": 4.127533366287692e-06,
+      "loss": 0.1733,
+      "step": 520
+    },
+    {
+      "epoch": 0.04448594024604569,
+      "grad_norm": 0.5386682152748108,
+      "learning_rate": 4.29230515735706e-06,
+      "loss": 0.2415,
+      "step": 540
+    },
+    {
+      "epoch": 0.046133567662565905,
+      "grad_norm": 2.7366793155670166,
+      "learning_rate": 4.457076948426429e-06,
+      "loss": 0.2409,
+      "step": 560
+    },
+    {
+      "epoch": 0.04778119507908612,
+      "grad_norm": 1.9912066459655762,
+      "learning_rate": 4.621848739495799e-06,
+      "loss": 0.2134,
+      "step": 580
+    },
+    {
+      "epoch": 0.049428822495606324,
+      "grad_norm": 4.058573246002197,
+      "learning_rate": 4.786620530565167e-06,
+      "loss": 0.2456,
+      "step": 600
+    },
+    {
+      "epoch": 0.05107644991212654,
+      "grad_norm": 0.37456750869750977,
+      "learning_rate": 4.951392321634536e-06,
+      "loss": 0.1589,
+      "step": 620
+    },
+    {
+      "epoch": 0.05272407732864675,
+      "grad_norm": 0.7950440645217896,
+      "learning_rate": 5.116164112703905e-06,
+      "loss": 0.2485,
+      "step": 640
+    },
+    {
+      "epoch": 0.05437170474516696,
+      "grad_norm": 1.884665846824646,
+      "learning_rate": 5.280935903773274e-06,
+      "loss": 0.2197,
+      "step": 660
+    },
+    {
+      "epoch": 0.05601933216168717,
+      "grad_norm": 0.32170844078063965,
+      "learning_rate": 5.445707694842643e-06,
+      "loss": 0.2111,
+      "step": 680
+    },
+    {
+      "epoch": 0.05766695957820738,
+      "grad_norm": 1.8230172395706177,
+      "learning_rate": 5.610479485912012e-06,
+      "loss": 0.2379,
+      "step": 700
+    },
+    {
+      "epoch": 0.059314586994727594,
+      "grad_norm": 1.2472524642944336,
+      "learning_rate": 5.775251276981381e-06,
+      "loss": 0.1684,
+      "step": 720
+    },
+    {
+      "epoch": 0.0609622144112478,
+      "grad_norm": 0.29022061824798584,
+      "learning_rate": 5.94002306805075e-06,
+      "loss": 0.2201,
+      "step": 740
+    },
+    {
+      "epoch": 0.06260984182776802,
+      "grad_norm": 0.49721184372901917,
+      "learning_rate": 6.104794859120118e-06,
+      "loss": 0.2297,
+      "step": 760
+    },
+    {
+      "epoch": 0.06425746924428823,
+      "grad_norm": 0.6133716106414795,
+      "learning_rate": 6.269566650189487e-06,
+      "loss": 0.3173,
+      "step": 780
+    },
+    {
+      "epoch": 0.06590509666080843,
+      "grad_norm": 0.9667792320251465,
+      "learning_rate": 6.434338441258857e-06,
+      "loss": 0.2349,
+      "step": 800
+    },
+    {
+      "epoch": 0.06755272407732865,
+      "grad_norm": 0.3177216053009033,
+      "learning_rate": 6.599110232328226e-06,
+      "loss": 0.1635,
+      "step": 820
+    },
+    {
+      "epoch": 0.06920035149384886,
+      "grad_norm": 0.8457621335983276,
+      "learning_rate": 6.763882023397594e-06,
+      "loss": 0.2777,
+      "step": 840
+    },
+    {
+      "epoch": 0.07084797891036906,
+      "grad_norm": 0.3330087661743164,
+      "learning_rate": 6.928653814466963e-06,
+      "loss": 0.2154,
+      "step": 860
+    },
+    {
+      "epoch": 0.07249560632688928,
+      "grad_norm": 0.5845814943313599,
+      "learning_rate": 7.093425605536333e-06,
+      "loss": 0.2164,
+      "step": 880
+    },
+    {
+      "epoch": 0.07414323374340949,
+      "grad_norm": 2.325303554534912,
+      "learning_rate": 7.258197396605701e-06,
+      "loss": 0.2068,
+      "step": 900
+    },
+    {
+      "epoch": 0.0757908611599297,
+      "grad_norm": 0.21893823146820068,
+      "learning_rate": 7.414730598121602e-06,
+      "loss": 0.1621,
+      "step": 920
+    },
+    {
+      "epoch": 0.07743848857644992,
+      "grad_norm": 0.5854327082633972,
+      "learning_rate": 7.5795023891909705e-06,
+      "loss": 0.2281,
+      "step": 940
+    },
+    {
+      "epoch": 0.07908611599297012,
+      "grad_norm": 0.2406030148267746,
+      "learning_rate": 7.74427418026034e-06,
+      "loss": 0.2342,
+      "step": 960
+    },
+    {
+      "epoch": 0.08073374340949033,
+      "grad_norm": 1.3764126300811768,
+      "learning_rate": 7.90904597132971e-06,
+      "loss": 0.2018,
+      "step": 980
+    },
+    {
+      "epoch": 0.08238137082601055,
+      "grad_norm": 0.9587862491607666,
+      "learning_rate": 8.073817762399077e-06,
+      "loss": 0.2489,
+      "step": 1000
+    },
+    {
+      "epoch": 0.08402899824253075,
+      "grad_norm": 0.9726558923721313,
+      "learning_rate": 8.238589553468447e-06,
+      "loss": 0.154,
+      "step": 1020
+    },
+    {
+      "epoch": 0.08567662565905097,
+      "grad_norm": 1.9828400611877441,
+      "learning_rate": 8.403361344537817e-06,
+      "loss": 0.2336,
+      "step": 1040
+    },
+    {
+      "epoch": 0.08732425307557118,
+      "grad_norm": 0.14761961996555328,
+      "learning_rate": 8.568133135607183e-06,
+      "loss": 0.1949,
+      "step": 1060
+    },
+    {
+      "epoch": 0.08897188049209138,
+      "grad_norm": 0.39077144861221313,
+      "learning_rate": 8.732904926676553e-06,
+      "loss": 0.2067,
+      "step": 1080
+    },
+    {
+      "epoch": 0.0906195079086116,
+      "grad_norm": 2.3257837295532227,
+      "learning_rate": 8.897676717745921e-06,
+      "loss": 0.2184,
+      "step": 1100
+    },
+    {
+      "epoch": 0.09226713532513181,
+      "grad_norm": 1.0966060161590576,
+      "learning_rate": 9.062448508815291e-06,
+      "loss": 0.1655,
+      "step": 1120
+    },
+    {
+      "epoch": 0.09391476274165202,
+      "grad_norm": 0.5648412704467773,
+      "learning_rate": 9.227220299884661e-06,
+      "loss": 0.2097,
+      "step": 1140
+    },
+    {
+      "epoch": 0.09556239015817224,
+      "grad_norm": 0.49816444516181946,
+      "learning_rate": 9.39199209095403e-06,
+      "loss": 0.2052,
+      "step": 1160
+    },
+    {
+      "epoch": 0.09721001757469244,
+      "grad_norm": 1.1164054870605469,
+      "learning_rate": 9.556763882023397e-06,
+      "loss": 0.2298,
+      "step": 1180
+    },
+    {
+      "epoch": 0.09885764499121265,
+      "grad_norm": 0.9453270435333252,
+      "learning_rate": 9.721535673092767e-06,
+      "loss": 0.2157,
+      "step": 1200
+    },
+    {
+      "epoch": 0.10050527240773287,
+      "grad_norm": 0.4588276743888855,
+      "learning_rate": 9.886307464162135e-06,
+      "loss": 0.1433,
+      "step": 1220
+    },
+    {
+      "epoch": 0.10215289982425307,
+      "grad_norm": 0.47535696625709534,
+      "learning_rate": 1.0051079255231505e-05,
+      "loss": 0.2115,
+      "step": 1240
+    },
+    {
+      "epoch": 0.10380052724077328,
+      "grad_norm": 0.585959792137146,
+      "learning_rate": 1.0215851046300875e-05,
+      "loss": 0.2245,
+      "step": 1260
+    },
+    {
+      "epoch": 0.1054481546572935,
+      "grad_norm": 0.24861204624176025,
+      "learning_rate": 1.0380622837370241e-05,
+      "loss": 0.1981,
+      "step": 1280
+    },
+    {
+      "epoch": 0.1070957820738137,
+      "grad_norm": 31.602128982543945,
+      "learning_rate": 1.0545394628439611e-05,
+      "loss": 0.2265,
+      "step": 1300
+    },
+    {
+      "epoch": 0.10874340949033393,
+      "grad_norm": 0.6234269142150879,
+      "learning_rate": 1.0710166419508981e-05,
+      "loss": 0.154,
+      "step": 1320
+    },
+    {
+      "epoch": 0.11039103690685413,
+      "grad_norm": 1.2423540353775024,
+      "learning_rate": 1.087493821057835e-05,
+      "loss": 0.2062,
+      "step": 1340
+    },
+    {
+      "epoch": 0.11203866432337434,
+      "grad_norm": 0.2090279757976532,
+      "learning_rate": 1.1039710001647719e-05,
+      "loss": 0.2214,
+      "step": 1360
+    },
+    {
+      "epoch": 0.11368629173989456,
+      "grad_norm": 0.6145613193511963,
+      "learning_rate": 1.1204481792717087e-05,
+      "loss": 0.1942,
+      "step": 1380
+    },
+    {
+      "epoch": 0.11533391915641476,
+      "grad_norm": 0.9004138708114624,
+      "learning_rate": 1.1369253583786455e-05,
+      "loss": 0.2271,
+      "step": 1400
+    },
+    {
+      "epoch": 0.11698154657293497,
+      "grad_norm": 1.609165906906128,
+      "learning_rate": 1.1534025374855825e-05,
+      "loss": 0.1606,
+      "step": 1420
+    },
+    {
+      "epoch": 0.11862917398945519,
+      "grad_norm": 0.8725568652153015,
+      "learning_rate": 1.1698797165925195e-05,
+      "loss": 0.204,
+      "step": 1440
+    },
+    {
+      "epoch": 0.1202768014059754,
+      "grad_norm": 1.8169455528259277,
+      "learning_rate": 1.1863568956994563e-05,
+      "loss": 0.207,
+      "step": 1460
+    },
+    {
+      "epoch": 0.1219244288224956,
+      "grad_norm": 0.37334388494491577,
+      "learning_rate": 1.2028340748063933e-05,
+      "loss": 0.2033,
+      "step": 1480
+    },
+    {
+      "epoch": 0.12357205623901582,
+      "grad_norm": 15.727474212646484,
+      "learning_rate": 1.2193112539133301e-05,
+      "loss": 0.3298,
+      "step": 1500
+    },
+    {
+      "epoch": 0.12521968365553604,
+      "grad_norm": 4.70903205871582,
+      "learning_rate": 1.2357884330202669e-05,
+      "loss": 0.159,
+      "step": 1520
+    },
+    {
+      "epoch": 0.12686731107205623,
+      "grad_norm": 0.4843326807022095,
+      "learning_rate": 1.2522656121272039e-05,
+      "loss": 0.2141,
+      "step": 1540
+    },
+    {
+      "epoch": 0.12851493848857645,
+      "grad_norm": 0.572084367275238,
+      "learning_rate": 1.2687427912341407e-05,
+      "loss": 0.2182,
+      "step": 1560
+    },
+    {
+      "epoch": 0.13016256590509667,
+      "grad_norm": 0.31078797578811646,
+      "learning_rate": 1.2852199703410777e-05,
+      "loss": 0.1994,
+      "step": 1580
+    },
+    {
+      "epoch": 0.13181019332161686,
+      "grad_norm": 9.014205932617188,
+      "learning_rate": 1.3016971494480145e-05,
+      "loss": 0.2089,
+      "step": 1600
+    },
+    {
+      "epoch": 0.13345782073813708,
+      "grad_norm": 1.0150245428085327,
+      "learning_rate": 1.3181743285549513e-05,
+      "loss": 0.1521,
+      "step": 1620
+    },
+    {
+      "epoch": 0.1351054481546573,
+      "grad_norm": 0.25271451473236084,
+      "learning_rate": 1.3346515076618885e-05,
+      "loss": 0.1996,
+      "step": 1640
+    },
+    {
+      "epoch": 0.1367530755711775,
+      "grad_norm": 0.47118502855300903,
+      "learning_rate": 1.3511286867688253e-05,
+      "loss": 0.2059,
+      "step": 1660
+    },
+    {
+      "epoch": 0.13840070298769772,
+      "grad_norm": 0.5134350657463074,
+      "learning_rate": 1.3676058658757621e-05,
+      "loss": 0.1935,
+      "step": 1680
+    },
+    {
+      "epoch": 0.14004833040421794,
+      "grad_norm": 1.0354816913604736,
+      "learning_rate": 1.384083044982699e-05,
+      "loss": 0.2103,
+      "step": 1700
+    },
+    {
+      "epoch": 0.14169595782073813,
+      "grad_norm": 0.5588876605033875,
+      "learning_rate": 1.4005602240896359e-05,
+      "loss": 0.1598,
+      "step": 1720
+    },
+    {
+      "epoch": 0.14334358523725835,
+      "grad_norm": 0.7309175133705139,
+      "learning_rate": 1.4170374031965727e-05,
+      "loss": 0.2204,
+      "step": 1740
+    },
+    {
+      "epoch": 0.14499121265377857,
+      "grad_norm": 0.6155902743339539,
+      "learning_rate": 1.4335145823035099e-05,
+      "loss": 0.2133,
+      "step": 1760
+    },
+    {
+      "epoch": 0.14663884007029876,
+      "grad_norm": 0.7660940885543823,
+      "learning_rate": 1.4499917614104467e-05,
+      "loss": 0.2065,
+      "step": 1780
+    },
+    {
+      "epoch": 0.14828646748681898,
+      "grad_norm": 1.1954026222229004,
+      "learning_rate": 1.4664689405173835e-05,
+      "loss": 0.2147,
+      "step": 1800
+    },
+    {
+      "epoch": 0.1499340949033392,
+      "grad_norm": 0.4249323606491089,
+      "learning_rate": 1.4829461196243205e-05,
+      "loss": 0.1553,
+      "step": 1820
+    },
+    {
+      "epoch": 0.1515817223198594,
+      "grad_norm": 2.9014129638671875,
+      "learning_rate": 1.4994232987312573e-05,
+      "loss": 0.2208,
+      "step": 1840
+    },
+    {
+      "epoch": 0.1532293497363796,
+      "grad_norm": 1.6474498510360718,
+      "learning_rate": 1.5159004778381941e-05,
+      "loss": 0.209,
+      "step": 1860
+    },
+    {
+      "epoch": 0.15487697715289983,
+      "grad_norm": 0.1585623174905777,
+      "learning_rate": 1.532377656945131e-05,
+      "loss": 0.1873,
+      "step": 1880
+    },
+    {
+      "epoch": 0.15652460456942002,
+      "grad_norm": 1.171941876411438,
+      "learning_rate": 1.548854836052068e-05,
+      "loss": 0.2389,
+      "step": 1900
+    },
+    {
+      "epoch": 0.15817223198594024,
+      "grad_norm": 0.48890382051467896,
+      "learning_rate": 1.5653320151590047e-05,
+      "loss": 0.1679,
+      "step": 1920
+    },
+    {
+      "epoch": 0.15981985940246046,
+      "grad_norm": 0.5568016767501831,
+      "learning_rate": 1.581809194265942e-05,
+      "loss": 0.1968,
+      "step": 1940
+    },
+    {
+      "epoch": 0.16146748681898065,
+      "grad_norm": 0.9775394797325134,
+      "learning_rate": 1.5982863733728787e-05,
+      "loss": 0.2208,
+      "step": 1960
+    },
+    {
+      "epoch": 0.16311511423550087,
+      "grad_norm": 0.60302734375,
+      "learning_rate": 1.6147635524798155e-05,
+      "loss": 0.1929,
+      "step": 1980
+    },
+    {
+      "epoch": 0.1647627416520211,
+      "grad_norm": 1.7513552904129028,
+      "learning_rate": 1.6312407315867526e-05,
+      "loss": 0.2055,
+      "step": 2000
+    },
+    {
+      "epoch": 0.1647627416520211,
+      "eval_loss": 0.5699400305747986,
+      "eval_runtime": 686.8117,
+      "eval_samples_per_second": 31.725,
+      "eval_steps_per_second": 7.932,
+      "eval_wer": 0.23469093535410654,
+      "step": 2000
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 60690,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 5,
+  "save_steps": 2000,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.0783600786383307e+20,
+  "train_batch_size": 24,
+  "trial_name": null,
+  "trial_params": null
+}

checkpoint-2000/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:523a85cdb294d1c4c61a6e49dd5260f441dad9fd7af932d1f950b5017f36ac73
+size 6648

checkpoint-2000/zero_to_fp32.py

@@ -0,0 +1,604 @@
+#!/usr/bin/env python
+
+# Copyright (c) Microsoft Corporation.
+# SPDX-License-Identifier: Apache-2.0
+
+# DeepSpeed Team
+
+# This script extracts fp32 consolidated weights from a zero 1, 2 and 3 DeepSpeed checkpoints. It gets
+# copied into the top level checkpoint dir, so the user can easily do the conversion at any point in
+# the future. Once extracted, the weights don't require DeepSpeed and can be used in any
+# application.
+#
+# example: python zero_to_fp32.py . pytorch_model.bin
+
+import argparse
+import torch
+import glob
+import math
+import os
+import re
+from collections import OrderedDict
+from dataclasses import dataclass
+
+# while this script doesn't use deepspeed to recover data, since the checkpoints are pickled with
+# DeepSpeed data structures it has to be available in the current python environment.
+from deepspeed.utils import logger
+from deepspeed.checkpoint.constants import (DS_VERSION, OPTIMIZER_STATE_DICT, SINGLE_PARTITION_OF_FP32_GROUPS,
+                                            FP32_FLAT_GROUPS, ZERO_STAGE, PARTITION_COUNT, PARAM_SHAPES, BUFFER_NAMES,
+                                            FROZEN_PARAM_SHAPES, FROZEN_PARAM_FRAGMENTS)
+
+
+@dataclass
+class zero_model_state:
+    buffers: dict()
+    param_shapes: dict()
+    shared_params: list
+    ds_version: int
+    frozen_param_shapes: dict()
+    frozen_param_fragments: dict()
+
+
+debug = 0
+
+# load to cpu
+device = torch.device('cpu')
+
+
+def atoi(text):
+    return int(text) if text.isdigit() else text
+
+
+def natural_keys(text):
+    '''
+    alist.sort(key=natural_keys) sorts in human order
+    http://nedbatchelder.com/blog/200712/human_sorting.html
+    (See Toothy's implementation in the comments)
+    '''
+    return [atoi(c) for c in re.split(r'(\d+)', text)]
+
+
+def get_model_state_file(checkpoint_dir, zero_stage):
+    if not os.path.isdir(checkpoint_dir):
+        raise FileNotFoundError(f"Directory '{checkpoint_dir}' doesn't exist")
+
+    # there should be only one file
+    if zero_stage <= 2:
+        file = os.path.join(checkpoint_dir, "mp_rank_00_model_states.pt")
+    elif zero_stage == 3:
+        file = os.path.join(checkpoint_dir, "zero_pp_rank_0_mp_rank_00_model_states.pt")
+
+    if not os.path.exists(file):
+        raise FileNotFoundError(f"can't find model states file at '{file}'")
+
+    return file
+
+
+def get_checkpoint_files(checkpoint_dir, glob_pattern):
+    # XXX: need to test that this simple glob rule works for multi-node setup too
+    ckpt_files = sorted(glob.glob(os.path.join(checkpoint_dir, glob_pattern)), key=natural_keys)
+
+    if len(ckpt_files) == 0:
+        raise FileNotFoundError(f"can't find {glob_pattern} files in directory '{checkpoint_dir}'")
+
+    return ckpt_files
+
+
+def get_optim_files(checkpoint_dir):
+    return get_checkpoint_files(checkpoint_dir, "*_optim_states.pt")
+
+
+def get_model_state_files(checkpoint_dir):
+    return get_checkpoint_files(checkpoint_dir, "*_model_states.pt")
+
+
+def parse_model_states(files):
+    zero_model_states = []
+    for file in files:
+        state_dict = torch.load(file, map_location=device)
+
+        if BUFFER_NAMES not in state_dict:
+            raise ValueError(f"{file} is not a model state checkpoint")
+        buffer_names = state_dict[BUFFER_NAMES]
+        if debug:
+            print("Found buffers:", buffer_names)
+
+        # recover just the buffers while restoring them to fp32 if they were saved in fp16
+        buffers = {k: v.float() for k, v in state_dict["module"].items() if k in buffer_names}
+        param_shapes = state_dict[PARAM_SHAPES]
+
+        # collect parameters that are included in param_shapes
+        param_names = []
+        for s in param_shapes:
+            for name in s.keys():
+                param_names.append(name)
+
+        # update with frozen parameters
+        frozen_param_shapes = state_dict.get(FROZEN_PARAM_SHAPES, None)
+        if frozen_param_shapes is not None:
+            if debug:
+                print(f"Found frozen_param_shapes: {frozen_param_shapes}")
+            param_names += list(frozen_param_shapes.keys())
+
+        # handle shared params
+        shared_params = [[k, v] for k, v in state_dict["shared_params"].items()]
+
+        ds_version = state_dict.get(DS_VERSION, None)
+
+        frozen_param_fragments = state_dict.get(FROZEN_PARAM_FRAGMENTS, None)
+
+        z_model_state = zero_model_state(buffers=buffers,
+                                         param_shapes=param_shapes,
+                                         shared_params=shared_params,
+                                         ds_version=ds_version,
+                                         frozen_param_shapes=frozen_param_shapes,
+                                         frozen_param_fragments=frozen_param_fragments)
+        zero_model_states.append(z_model_state)
+
+    return zero_model_states
+
+
+def parse_optim_states(files, ds_checkpoint_dir):
+
+    total_files = len(files)
+    state_dicts = []
+    for f in files:
+        state_dict = torch.load(f, map_location=device)
+        # immediately discard the potentially huge 2 optimizer states as we only care for fp32 master weights
+        # and also handle the case where it was already removed by another helper script
+        state_dict["optimizer_state_dict"].pop("optimizer_state_dict", None)
+        state_dicts.append(state_dict)
+
+    if not ZERO_STAGE in state_dicts[0][OPTIMIZER_STATE_DICT]:
+        raise ValueError(f"{files[0]} is not a zero checkpoint")
+    zero_stage = state_dicts[0][OPTIMIZER_STATE_DICT][ZERO_STAGE]
+    world_size = state_dicts[0][OPTIMIZER_STATE_DICT][PARTITION_COUNT]
+
+    # For ZeRO-2 each param group can have different partition_count as data parallelism for expert
+    # parameters can be different from data parallelism for non-expert parameters. So we can just
+    # use the max of the partition_count to get the dp world_size.
+
+    if type(world_size) is list:
+        world_size = max(world_size)
+
+    if world_size != total_files:
+        raise ValueError(
+            f"Expected {world_size} of '*_optim_states.pt' under '{ds_checkpoint_dir}' but found {total_files} files. "
+            "Possibly due to an overwrite of an old checkpoint, or a checkpoint didn't get saved by one or more processes."
+        )
+
+    # the groups are named differently in each stage
+    if zero_stage <= 2:
+        fp32_groups_key = SINGLE_PARTITION_OF_FP32_GROUPS
+    elif zero_stage == 3:
+        fp32_groups_key = FP32_FLAT_GROUPS
+    else:
+        raise ValueError(f"unknown zero stage {zero_stage}")
+
+    if zero_stage <= 2:
+        fp32_flat_groups = [state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key] for i in range(len(state_dicts))]
+    elif zero_stage == 3:
+        # if there is more than one param group, there will be multiple flattened tensors - one
+        # flattened tensor per group - for simplicity merge them into a single tensor
+        #
+        # XXX: could make the script more memory efficient for when there are multiple groups - it
+        # will require matching the sub-lists of param_shapes for each param group flattened tensor
+
+        fp32_flat_groups = [
+            torch.cat(state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key], 0) for i in range(len(state_dicts))
+        ]
+
+    return zero_stage, world_size, fp32_flat_groups
+
+
+def _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters):
+    """
+    Returns fp32 state_dict reconstructed from ds checkpoint
+
+    Args:
+        - ``ds_checkpoint_dir``: path to the deepspeed checkpoint folder (where the optimizer files are)
+
+    """
+    print(f"Processing zero checkpoint '{ds_checkpoint_dir}'")
+
+    optim_files = get_optim_files(ds_checkpoint_dir)
+    zero_stage, world_size, fp32_flat_groups = parse_optim_states(optim_files, ds_checkpoint_dir)
+    print(f"Detected checkpoint of type zero stage {zero_stage}, world_size: {world_size}")
+
+    model_files = get_model_state_files(ds_checkpoint_dir)
+
+    zero_model_states = parse_model_states(model_files)
+    print(f'Parsing checkpoint created by deepspeed=={zero_model_states[0].ds_version}')
+
+    if zero_stage <= 2:
+        return _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                                          exclude_frozen_parameters)
+    elif zero_stage == 3:
+        return _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                                          exclude_frozen_parameters)
+
+
+def _zero2_merge_frozen_params(state_dict, zero_model_states):
+    if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
+        return
+
+    frozen_param_shapes = zero_model_states[0].frozen_param_shapes
+    frozen_param_fragments = zero_model_states[0].frozen_param_fragments
+
+    if debug:
+        num_elem = sum(s.numel() for s in frozen_param_shapes.values())
+        print(f'rank 0: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
+
+        wanted_params = len(frozen_param_shapes)
+        wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
+        avail_numel = sum([p.numel() for p in frozen_param_fragments.values()])
+        print(f'Frozen params: Have {avail_numel} numels to process.')
+        print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
+
+    total_params = 0
+    total_numel = 0
+    for name, shape in frozen_param_shapes.items():
+        total_params += 1
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+
+        state_dict[name] = frozen_param_fragments[name]
+
+        if debug:
+            print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
+
+    print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _has_callable(obj, fn):
+    attr = getattr(obj, fn, None)
+    return callable(attr)
+
+
+def _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
+    param_shapes = zero_model_states[0].param_shapes
+
+    # Reconstruction protocol:
+    #
+    # XXX: document this
+
+    if debug:
+        for i in range(world_size):
+            for j in range(len(fp32_flat_groups[0])):
+                print(f"{FP32_FLAT_GROUPS}[{i}][{j}].shape={fp32_flat_groups[i][j].shape}")
+
+    # XXX: memory usage doubles here (zero2)
+    num_param_groups = len(fp32_flat_groups[0])
+    merged_single_partition_of_fp32_groups = []
+    for i in range(num_param_groups):
+        merged_partitions = [sd[i] for sd in fp32_flat_groups]
+        full_single_fp32_vector = torch.cat(merged_partitions, 0)
+        merged_single_partition_of_fp32_groups.append(full_single_fp32_vector)
+    avail_numel = sum(
+        [full_single_fp32_vector.numel() for full_single_fp32_vector in merged_single_partition_of_fp32_groups])
+
+    if debug:
+        wanted_params = sum([len(shapes) for shapes in param_shapes])
+        wanted_numel = sum([sum(shape.numel() for shape in shapes.values()) for shapes in param_shapes])
+        # not asserting if there is a mismatch due to possible padding
+        print(f"Have {avail_numel} numels to process.")
+        print(f"Need {wanted_numel} numels in {wanted_params} params.")
+
+    # params
+    # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
+    # out-of-core computing solution
+    total_numel = 0
+    total_params = 0
+    for shapes, full_single_fp32_vector in zip(param_shapes, merged_single_partition_of_fp32_groups):
+        offset = 0
+        avail_numel = full_single_fp32_vector.numel()
+        for name, shape in shapes.items():
+
+            unpartitioned_numel = shape.numel() if _has_callable(shape, 'numel') else math.prod(shape)
+            total_numel += unpartitioned_numel
+            total_params += 1
+
+            if debug:
+                print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
+            state_dict[name] = full_single_fp32_vector.narrow(0, offset, unpartitioned_numel).view(shape)
+            offset += unpartitioned_numel
+
+        # Z2 started to align to 2*world_size to improve nccl performance. Therefore both offset and
+        # avail_numel can differ by anywhere between 0..2*world_size. Due to two unrelated complex
+        # paddings performed in the code it's almost impossible to predict the exact numbers w/o the
+        # live optimizer object, so we are checking that the numbers are within the right range
+        align_to = 2 * world_size
+
+        def zero2_align(x):
+            return align_to * math.ceil(x / align_to)
+
+        if debug:
+            print(f"original offset={offset}, avail_numel={avail_numel}")
+
+        offset = zero2_align(offset)
+        avail_numel = zero2_align(avail_numel)
+
+        if debug:
+            print(f"aligned  offset={offset}, avail_numel={avail_numel}")
+
+        # Sanity check
+        if offset != avail_numel:
+            raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
+
+    print(f"Reconstructed fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                               exclude_frozen_parameters):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    if not exclude_frozen_parameters:
+        _zero2_merge_frozen_params(state_dict, zero_model_states)
+
+    _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
+
+    # recover shared parameters
+    for pair in zero_model_states[0].shared_params:
+        if pair[1] in state_dict:
+            state_dict[pair[0]] = state_dict[pair[1]]
+
+    return state_dict
+
+
+def zero3_partitioned_param_info(unpartitioned_numel, world_size):
+    remainder = unpartitioned_numel % world_size
+    padding_numel = (world_size - remainder) if remainder else 0
+    partitioned_numel = math.ceil(unpartitioned_numel / world_size)
+    return partitioned_numel, padding_numel
+
+
+def _zero3_merge_frozen_params(state_dict, world_size, zero_model_states):
+    if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
+        return
+
+    if debug:
+        for i in range(world_size):
+            num_elem = sum(s.numel() for s in zero_model_states[i].frozen_param_fragments.values())
+            print(f'rank {i}: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
+
+        frozen_param_shapes = zero_model_states[0].frozen_param_shapes
+        wanted_params = len(frozen_param_shapes)
+        wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
+        avail_numel = sum([p.numel() for p in zero_model_states[0].frozen_param_fragments.values()]) * world_size
+        print(f'Frozen params: Have {avail_numel} numels to process.')
+        print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
+
+    total_params = 0
+    total_numel = 0
+    for name, shape in zero_model_states[0].frozen_param_shapes.items():
+        total_params += 1
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+
+        param_frags = tuple(model_state.frozen_param_fragments[name] for model_state in zero_model_states)
+        state_dict[name] = torch.cat(param_frags, 0).narrow(0, 0, unpartitioned_numel).view(shape)
+
+        partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
+
+        if debug:
+            print(
+                f"Frozen params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
+            )
+
+    print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
+    param_shapes = zero_model_states[0].param_shapes
+    avail_numel = fp32_flat_groups[0].numel() * world_size
+    # Reconstruction protocol: For zero3 we need to zip the partitions together at boundary of each
+    # param, re-consolidating each param, while dealing with padding if any
+
+    # merge list of dicts, preserving order
+    param_shapes = {k: v for d in param_shapes for k, v in d.items()}
+
+    if debug:
+        for i in range(world_size):
+            print(f"{FP32_FLAT_GROUPS}[{i}].shape={fp32_flat_groups[i].shape}")
+
+        wanted_params = len(param_shapes)
+        wanted_numel = sum(shape.numel() for shape in param_shapes.values())
+        # not asserting if there is a mismatch due to possible padding
+        avail_numel = fp32_flat_groups[0].numel() * world_size
+        print(f"Trainable params: Have {avail_numel} numels to process.")
+        print(f"Trainable params: Need {wanted_numel} numels in {wanted_params} params.")
+
+    # params
+    # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
+    # out-of-core computing solution
+    offset = 0
+    total_numel = 0
+    total_params = 0
+    for name, shape in param_shapes.items():
+
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+        total_params += 1
+
+        partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
+
+        if debug:
+            print(
+                f"Trainable params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
+            )
+
+        # XXX: memory usage doubles here
+        state_dict[name] = torch.cat(
+            tuple(fp32_flat_groups[i].narrow(0, offset, partitioned_numel) for i in range(world_size)),
+            0).narrow(0, 0, unpartitioned_numel).view(shape)
+        offset += partitioned_numel
+
+    offset *= world_size
+
+    # Sanity check
+    if offset != avail_numel:
+        raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
+
+    print(f"Reconstructed Trainable fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                               exclude_frozen_parameters):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    if not exclude_frozen_parameters:
+        _zero3_merge_frozen_params(state_dict, world_size, zero_model_states)
+
+    _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
+
+    # recover shared parameters
+    for pair in zero_model_states[0].shared_params:
+        if pair[1] in state_dict:
+            state_dict[pair[0]] = state_dict[pair[1]]
+
+    return state_dict
+
+
+def get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag=None, exclude_frozen_parameters=False):
+    """
+    Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated state_dict that can be loaded with
+    ``load_state_dict()`` and used for training without DeepSpeed or shared with others, for example
+    via a model hub.
+
+    Args:
+        - ``checkpoint_dir``: path to the desired checkpoint folder
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in 'latest' file. e.g., ``global_step14``
+        - ``exclude_frozen_parameters``: exclude frozen parameters
+
+    Returns:
+        - pytorch ``state_dict``
+
+    Note: this approach may not work if your application doesn't have sufficient free CPU memory and
+    you may need to use the offline approach using the ``zero_to_fp32.py`` script that is saved with
+    the checkpoint.
+
+    A typical usage might be ::
+
+        from deepspeed.utils.zero_to_fp32 import get_fp32_state_dict_from_zero_checkpoint
+        # do the training and checkpoint saving
+        state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir) # already on cpu
+        model = model.cpu() # move to cpu
+        model.load_state_dict(state_dict)
+        # submit to model hub or save the model to share with others
+
+    In this example the ``model`` will no longer be usable in the deepspeed context of the same
+    application. i.e. you will need to re-initialize the deepspeed engine, since
+    ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
+
+    If you want it all done for you, use ``load_state_dict_from_zero_checkpoint`` instead.
+
+    """
+    if tag is None:
+        latest_path = os.path.join(checkpoint_dir, 'latest')
+        if os.path.isfile(latest_path):
+            with open(latest_path, 'r') as fd:
+                tag = fd.read().strip()
+        else:
+            raise ValueError(f"Unable to find 'latest' file at {latest_path}")
+
+    ds_checkpoint_dir = os.path.join(checkpoint_dir, tag)
+
+    if not os.path.isdir(ds_checkpoint_dir):
+        raise FileNotFoundError(f"Directory '{ds_checkpoint_dir}' doesn't exist")
+
+    return _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters)
+
+
+def convert_zero_checkpoint_to_fp32_state_dict(checkpoint_dir, output_file, tag=None, exclude_frozen_parameters=False):
+    """
+    Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict`` file that can be
+    loaded with ``torch.load(file)`` + ``load_state_dict()`` and used for training without DeepSpeed.
+
+    Args:
+        - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
+        - ``output_file``: path to the pytorch fp32 state_dict output file (e.g. path/pytorch_model.bin)
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
+        - ``exclude_frozen_parameters``: exclude frozen parameters
+    """
+
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag, exclude_frozen_parameters)
+    print(f"Saving fp32 state dict to {output_file}")
+    torch.save(state_dict, output_file)
+
+
+def load_state_dict_from_zero_checkpoint(model, checkpoint_dir, tag=None):
+    """
+    1. Put the provided model to cpu
+    2. Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict``
+    3. Load it into the provided model
+
+    Args:
+        - ``model``: the model object to update
+        - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
+
+    Returns:
+        - ``model`: modified model
+
+    Make sure you have plenty of CPU memory available before you call this function. If you don't
+    have enough use the ``zero_to_fp32.py`` utility to do the conversion. You will find it
+    conveniently placed for you in the checkpoint folder.
+
+    A typical usage might be ::
+
+        from deepspeed.utils.zero_to_fp32 import load_state_dict_from_zero_checkpoint
+        model = load_state_dict_from_zero_checkpoint(trainer.model, checkpoint_dir)
+        # submit to model hub or save the model to share with others
+
+    Note, that once this was run, the ``model`` will no longer be usable in the deepspeed context
+    of the same application. i.e. you will need to re-initialize the deepspeed engine, since
+    ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
+
+    """
+    logger.info(f"Extracting fp32 weights")
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag)
+
+    logger.info(f"Overwriting model with fp32 weights")
+    model = model.cpu()
+    model.load_state_dict(state_dict, strict=False)
+
+    return model
+
+
+if __name__ == "__main__":
+
+    parser = argparse.ArgumentParser()
+    parser.add_argument("checkpoint_dir",
+                        type=str,
+                        help="path to the desired checkpoint folder, e.g., path/checkpoint-12")
+    parser.add_argument(
+        "output_file",
+        type=str,
+        help="path to the pytorch fp32 state_dict output file (e.g. path/checkpoint-12/pytorch_model.bin)")
+    parser.add_argument("-t",
+                        "--tag",
+                        type=str,
+                        default=None,
+                        help="checkpoint tag used as a unique identifier for checkpoint. e.g., global_step1")
+    parser.add_argument("--exclude_frozen_parameters", action='store_true', help="exclude frozen parameters")
+    parser.add_argument("-d", "--debug", action='store_true', help="enable debug")
+    args = parser.parse_args()
+
+    debug = args.debug
+
+    convert_zero_checkpoint_to_fp32_state_dict(args.checkpoint_dir,
+                                               args.output_file,
+                                               tag=args.tag,
+                                               exclude_frozen_parameters=args.exclude_frozen_parameters)