Training in progress, step 4500, checkpoint

last-checkpoint/config.json

@@ -0,0 +1,118 @@
+{
+  "_name_or_path": "mattbonnell/wav2vec2-base-wonders-phonemes",
+  "activation_dropout": 0.0,
+  "adapter_attn_dim": null,
+  "adapter_kernel_size": 3,
+  "adapter_stride": 2,
+  "add_adapter": false,
+  "apply_spec_augment": true,
+  "architectures": [
+    "Wav2Vec2ForCTC"
+  ],
+  "attention_dropout": 0.1,
+  "bos_token_id": 44,
+  "classifier_proj_size": 256,
+  "codevector_dim": 256,
+  "contrastive_logits_temperature": 0.1,
+  "conv_bias": false,
+  "conv_dim": [
+    512,
+    512,
+    512,
+    512,
+    512,
+    512,
+    512
+  ],
+  "conv_kernel": [
+    10,
+    3,
+    3,
+    3,
+    3,
+    2,
+    2
+  ],
+  "conv_stride": [
+    5,
+    2,
+    2,
+    2,
+    2,
+    2,
+    2
+  ],
+  "ctc_loss_reduction": "mean",
+  "ctc_zero_infinity": false,
+  "diversity_loss_weight": 0.1,
+  "do_stable_layer_norm": false,
+  "eos_token_id": 45,
+  "feat_extract_activation": "gelu",
+  "feat_extract_norm": "group",
+  "feat_proj_dropout": 0.1,
+  "feat_quantizer_dropout": 0.0,
+  "final_dropout": 0.0,
+  "freeze_feat_extract_train": true,
+  "hidden_act": "gelu",
+  "hidden_dropout": 0.1,
+  "hidden_size": 768,
+  "initializer_range": 0.02,
+  "intermediate_size": 3072,
+  "layer_norm_eps": 1e-05,
+  "layerdrop": 0.0,
+  "mask_channel_length": 10,
+  "mask_channel_min_space": 1,
+  "mask_channel_other": 0.0,
+  "mask_channel_prob": 0.0,
+  "mask_channel_selection": "static",
+  "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_min_space": 1,
+  "mask_time_other": 0.0,
+  "mask_time_prob": 0.05,
+  "mask_time_selection": "static",
+  "model_type": "wav2vec2",
+  "no_mask_channel_overlap": false,
+  "no_mask_time_overlap": false,
+  "num_adapter_layers": 3,
+  "num_attention_heads": 12,
+  "num_codevector_groups": 2,
+  "num_codevectors_per_group": 320,
+  "num_conv_pos_embedding_groups": 16,
+  "num_conv_pos_embeddings": 128,
+  "num_feat_extract_layers": 7,
+  "num_hidden_layers": 12,
+  "num_negatives": 100,
+  "output_hidden_size": 768,
+  "pad_token_id": 43,
+  "proj_codevector_dim": 256,
+  "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": "4.44.0",
+  "use_weighted_layer_sum": false,
+  "vocab_size": 44,
+  "xvector_output_dim": 512
+}

last-checkpoint/global_step4500/mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bd96abfdf0f667ccfa2cfd9cb62a3b19397ae74016e18f9a9b36be34b59d777d
+size 197282509

last-checkpoint/global_step4500/zero_pp_rank_0_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0e607167b54a0a603cf83558fcde26bf9a22c462505c6edd944df0cf79d88b06
+size 180416968

last-checkpoint/global_step4500/zero_pp_rank_1_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e6649a324a9936debc9adb68629eebbdc290c8b94427ec79456bf8c4bca70922
+size 180416776

last-checkpoint/global_step4500/zero_pp_rank_2_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:120c2889e8084b7c7a7be2d6f1c1bf57407048e6a061fa2a94eb3fb8db030235
+size 180416776

last-checkpoint/global_step4500/zero_pp_rank_3_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:29339698f7ac0eeffae61567df11c336897bb00b965d0b3d45347b3d1358931f
+size 180416904

last-checkpoint/global_step4500/zero_pp_rank_4_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0b8141306996a8f211eba8ce4a475a76c1def197433b6b264830b1008c76f5f1
+size 180416712

last-checkpoint/global_step4500/zero_pp_rank_5_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:431abe8474245a2298cada3f16c163e6d1eaa928b98ab3c57a5e1d3845974e08
+size 180417096

last-checkpoint/latest

@@ -0,0 +1 @@
+global_step4500

last-checkpoint/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:afa11854fc3f38256f65537c45a0706904c65c965ac824be6c1518136343cf1f
+size 188836816

last-checkpoint/preprocessor_config.json

@@ -0,0 +1,9 @@
+{
+  "do_normalize": true,
+  "feature_extractor_type": "Wav2Vec2FeatureExtractor",
+  "feature_size": 1,
+  "padding_side": "right",
+  "padding_value": 0.0,
+  "return_attention_mask": true,
+  "sampling_rate": 16000
+}

last-checkpoint/rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:618157755d4d53643070c0fe9f2232d72c2856d06d68fa324b284e1a0894f9fa
+size 15536

last-checkpoint/rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c8df10d37b49d0cf833b2d40fff5a44a5d87f7f2d11d390279b5fa4ef5867a77
+size 15472

last-checkpoint/rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a21bf0f1ce65716c86bd9946ac949fe1b906bdb3fa7b3a9fe5bac1470a9c6526
+size 15536

last-checkpoint/rng_state_3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cd64d764026e4b1fe2b2d992bd70febb05e845122f39bb83f89256ad3d638ab1
+size 15536

last-checkpoint/rng_state_4.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f7d18ac7b78cd3763b57e2ecde02ade25812c67485a797876a8b1ed55514af39
+size 15472

last-checkpoint/rng_state_5.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8099d966bd9cbb159a829accfb15cf83a349f505d552231427f6776e4d7ab45b
+size 15472

last-checkpoint/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b6f6921664880dd0bcd6405a752e332045b3848ba423b00290e6c113f562ae1e
+size 1256

last-checkpoint/trainer_state.json

@@ -0,0 +1,681 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 65.21739130434783,
+  "eval_steps": 1500,
+  "global_step": 4500,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.36231884057971014,
+      "grad_norm": 0.7415200471878052,
+      "learning_rate": 9.369565217391306e-06,
+      "loss": 2.8372,
+      "step": 50
+    },
+    {
+      "epoch": 0.7246376811594203,
+      "grad_norm": 0.7866194844245911,
+      "learning_rate": 2.1086956521739132e-05,
+      "loss": 2.7495,
+      "step": 100
+    },
+    {
+      "epoch": 1.0869565217391304,
+      "grad_norm": 1.3562049865722656,
+      "learning_rate": 3.304347826086956e-05,
+      "loss": 2.649,
+      "step": 150
+    },
+    {
+      "epoch": 1.4492753623188406,
+      "grad_norm": 1.167938232421875,
+      "learning_rate": 4.5e-05,
+      "loss": 2.4866,
+      "step": 200
+    },
+    {
+      "epoch": 1.8115942028985508,
+      "grad_norm": 2.102412462234497,
+      "learning_rate": 5.6956521739130437e-05,
+      "loss": 2.2804,
+      "step": 250
+    },
+    {
+      "epoch": 2.1739130434782608,
+      "grad_norm": 1.8163594007492065,
+      "learning_rate": 6.891304347826088e-05,
+      "loss": 2.1049,
+      "step": 300
+    },
+    {
+      "epoch": 2.536231884057971,
+      "grad_norm": 1.6761736869812012,
+      "learning_rate": 8.086956521739131e-05,
+      "loss": 1.9311,
+      "step": 350
+    },
+    {
+      "epoch": 2.898550724637681,
+      "grad_norm": 2.327892303466797,
+      "learning_rate": 9.282608695652174e-05,
+      "loss": 1.7867,
+      "step": 400
+    },
+    {
+      "epoch": 3.260869565217391,
+      "grad_norm": 1.8326770067214966,
+      "learning_rate": 0.0001,
+      "loss": 1.6734,
+      "step": 450
+    },
+    {
+      "epoch": 3.6231884057971016,
+      "grad_norm": 2.092679738998413,
+      "learning_rate": 0.0001,
+      "loss": 1.5518,
+      "step": 500
+    },
+    {
+      "epoch": 3.9855072463768115,
+      "grad_norm": 1.874523401260376,
+      "learning_rate": 0.0001,
+      "loss": 1.5172,
+      "step": 550
+    },
+    {
+      "epoch": 4.3478260869565215,
+      "grad_norm": 2.122560977935791,
+      "learning_rate": 0.0001,
+      "loss": 1.4705,
+      "step": 600
+    },
+    {
+      "epoch": 4.710144927536232,
+      "grad_norm": 1.9091864824295044,
+      "learning_rate": 0.0001,
+      "loss": 1.3828,
+      "step": 650
+    },
+    {
+      "epoch": 5.072463768115942,
+      "grad_norm": 1.9627286195755005,
+      "learning_rate": 0.0001,
+      "loss": 1.3473,
+      "step": 700
+    },
+    {
+      "epoch": 5.434782608695652,
+      "grad_norm": 1.845944881439209,
+      "learning_rate": 0.0001,
+      "loss": 1.3271,
+      "step": 750
+    },
+    {
+      "epoch": 5.797101449275362,
+      "grad_norm": 1.8691362142562866,
+      "learning_rate": 0.0001,
+      "loss": 1.2634,
+      "step": 800
+    },
+    {
+      "epoch": 6.159420289855072,
+      "grad_norm": 2.231165647506714,
+      "learning_rate": 0.0001,
+      "loss": 1.2429,
+      "step": 850
+    },
+    {
+      "epoch": 6.521739130434782,
+      "grad_norm": 1.8022764921188354,
+      "learning_rate": 0.0001,
+      "loss": 1.215,
+      "step": 900
+    },
+    {
+      "epoch": 6.884057971014493,
+      "grad_norm": 1.918607473373413,
+      "learning_rate": 0.0001,
+      "loss": 1.1836,
+      "step": 950
+    },
+    {
+      "epoch": 7.246376811594203,
+      "grad_norm": 1.9824390411376953,
+      "learning_rate": 0.0001,
+      "loss": 1.1721,
+      "step": 1000
+    },
+    {
+      "epoch": 7.608695652173913,
+      "grad_norm": 1.8581913709640503,
+      "learning_rate": 0.0001,
+      "loss": 1.1351,
+      "step": 1050
+    },
+    {
+      "epoch": 7.971014492753623,
+      "grad_norm": 1.8977991342544556,
+      "learning_rate": 0.0001,
+      "loss": 1.1176,
+      "step": 1100
+    },
+    {
+      "epoch": 8.333333333333334,
+      "grad_norm": 2.095898389816284,
+      "learning_rate": 0.0001,
+      "loss": 1.1092,
+      "step": 1150
+    },
+    {
+      "epoch": 8.695652173913043,
+      "grad_norm": 2.253805637359619,
+      "learning_rate": 0.0001,
+      "loss": 1.0644,
+      "step": 1200
+    },
+    {
+      "epoch": 9.057971014492754,
+      "grad_norm": 2.290984869003296,
+      "learning_rate": 0.0001,
+      "loss": 1.0581,
+      "step": 1250
+    },
+    {
+      "epoch": 9.420289855072463,
+      "grad_norm": 1.8521511554718018,
+      "learning_rate": 0.0001,
+      "loss": 1.0616,
+      "step": 1300
+    },
+    {
+      "epoch": 9.782608695652174,
+      "grad_norm": 2.424522876739502,
+      "learning_rate": 0.0001,
+      "loss": 1.0235,
+      "step": 1350
+    },
+    {
+      "epoch": 10.144927536231885,
+      "grad_norm": 1.9674047231674194,
+      "learning_rate": 0.0001,
+      "loss": 1.018,
+      "step": 1400
+    },
+    {
+      "epoch": 10.507246376811594,
+      "grad_norm": 2.169654130935669,
+      "learning_rate": 0.0001,
+      "loss": 0.9963,
+      "step": 1450
+    },
+    {
+      "epoch": 10.869565217391305,
+      "grad_norm": 2.0874452590942383,
+      "learning_rate": 0.0001,
+      "loss": 0.9781,
+      "step": 1500
+    },
+    {
+      "epoch": 11.231884057971014,
+      "grad_norm": 2.0761208534240723,
+      "learning_rate": 0.0001,
+      "loss": 0.9712,
+      "step": 1550
+    },
+    {
+      "epoch": 11.594202898550725,
+      "grad_norm": 2.053912401199341,
+      "learning_rate": 0.0001,
+      "loss": 0.9663,
+      "step": 1600
+    },
+    {
+      "epoch": 11.956521739130435,
+      "grad_norm": 2.502487897872925,
+      "learning_rate": 0.0001,
+      "loss": 0.9438,
+      "step": 1650
+    },
+    {
+      "epoch": 12.318840579710145,
+      "grad_norm": 2.000900983810425,
+      "learning_rate": 0.0001,
+      "loss": 0.9521,
+      "step": 1700
+    },
+    {
+      "epoch": 12.681159420289855,
+      "grad_norm": 2.4034128189086914,
+      "learning_rate": 0.0001,
+      "loss": 0.9084,
+      "step": 1750
+    },
+    {
+      "epoch": 13.043478260869565,
+      "grad_norm": 1.9942097663879395,
+      "learning_rate": 0.0001,
+      "loss": 0.9016,
+      "step": 1800
+    },
+    {
+      "epoch": 13.405797101449275,
+      "grad_norm": 1.8242056369781494,
+      "learning_rate": 0.0001,
+      "loss": 0.9113,
+      "step": 1850
+    },
+    {
+      "epoch": 13.768115942028986,
+      "grad_norm": 2.07265305519104,
+      "learning_rate": 0.0001,
+      "loss": 0.8838,
+      "step": 1900
+    },
+    {
+      "epoch": 14.130434782608695,
+      "grad_norm": 1.9476542472839355,
+      "learning_rate": 0.0001,
+      "loss": 0.8749,
+      "step": 1950
+    },
+    {
+      "epoch": 14.492753623188406,
+      "grad_norm": 1.7369720935821533,
+      "learning_rate": 0.0001,
+      "loss": 0.8767,
+      "step": 2000
+    },
+    {
+      "epoch": 14.855072463768115,
+      "grad_norm": 1.6376968622207642,
+      "learning_rate": 0.0001,
+      "loss": 0.8649,
+      "step": 2050
+    },
+    {
+      "epoch": 15.217391304347826,
+      "grad_norm": 2.408804416656494,
+      "learning_rate": 0.0001,
+      "loss": 0.8546,
+      "step": 2100
+    },
+    {
+      "epoch": 15.579710144927537,
+      "grad_norm": 2.418651819229126,
+      "learning_rate": 0.0001,
+      "loss": 0.8557,
+      "step": 2150
+    },
+    {
+      "epoch": 15.942028985507246,
+      "grad_norm": 1.9724761247634888,
+      "learning_rate": 0.0001,
+      "loss": 0.8418,
+      "step": 2200
+    },
+    {
+      "epoch": 16.304347826086957,
+      "grad_norm": 1.8794658184051514,
+      "learning_rate": 0.0001,
+      "loss": 0.8316,
+      "step": 2250
+    },
+    {
+      "epoch": 16.666666666666668,
+      "grad_norm": 1.8745840787887573,
+      "learning_rate": 0.0001,
+      "loss": 0.8179,
+      "step": 2300
+    },
+    {
+      "epoch": 17.028985507246375,
+      "grad_norm": 2.0710532665252686,
+      "learning_rate": 0.0001,
+      "loss": 0.8078,
+      "step": 2350
+    },
+    {
+      "epoch": 17.391304347826086,
+      "grad_norm": 1.9383975267410278,
+      "learning_rate": 0.0001,
+      "loss": 0.806,
+      "step": 2400
+    },
+    {
+      "epoch": 17.753623188405797,
+      "grad_norm": 2.5431768894195557,
+      "learning_rate": 0.0001,
+      "loss": 0.795,
+      "step": 2450
+    },
+    {
+      "epoch": 18.115942028985508,
+      "grad_norm": 2.2625672817230225,
+      "learning_rate": 0.0001,
+      "loss": 0.7889,
+      "step": 2500
+    },
+    {
+      "epoch": 18.47826086956522,
+      "grad_norm": 1.995342493057251,
+      "learning_rate": 9.98804347826087e-05,
+      "loss": 0.7857,
+      "step": 2550
+    },
+    {
+      "epoch": 18.840579710144926,
+      "grad_norm": 2.2518508434295654,
+      "learning_rate": 9.975323774283072e-05,
+      "loss": 0.774,
+      "step": 2600
+    },
+    {
+      "epoch": 19.202898550724637,
+      "grad_norm": 2.207653522491455,
+      "learning_rate": 9.962604070305273e-05,
+      "loss": 0.762,
+      "step": 2650
+    },
+    {
+      "epoch": 19.565217391304348,
+      "grad_norm": 2.011214017868042,
+      "learning_rate": 9.949884366327475e-05,
+      "loss": 0.7556,
+      "step": 2700
+    },
+    {
+      "epoch": 19.92753623188406,
+      "grad_norm": 1.8435475826263428,
+      "learning_rate": 9.937164662349677e-05,
+      "loss": 0.7505,
+      "step": 2750
+    },
+    {
+      "epoch": 20.28985507246377,
+      "grad_norm": 1.9571833610534668,
+      "learning_rate": 9.924444958371878e-05,
+      "loss": 0.7442,
+      "step": 2800
+    },
+    {
+      "epoch": 20.652173913043477,
+      "grad_norm": 1.830269694328308,
+      "learning_rate": 9.91172525439408e-05,
+      "loss": 0.7288,
+      "step": 2850
+    },
+    {
+      "epoch": 21.014492753623188,
+      "grad_norm": 2.604294538497925,
+      "learning_rate": 9.899005550416282e-05,
+      "loss": 0.719,
+      "step": 2900
+    },
+    {
+      "epoch": 21.3768115942029,
+      "grad_norm": 2.1636900901794434,
+      "learning_rate": 9.886285846438483e-05,
+      "loss": 0.7301,
+      "step": 2950
+    },
+    {
+      "epoch": 21.73913043478261,
+      "grad_norm": 2.368319034576416,
+      "learning_rate": 9.873566142460685e-05,
+      "loss": 0.7184,
+      "step": 3000
+    },
+    {
+      "epoch": 21.73913043478261,
+      "eval_loss": 0.902417004108429,
+      "eval_runtime": 584.856,
+      "eval_samples_per_second": 10.03,
+      "eval_steps_per_second": 0.027,
+      "eval_wer": 0.49515657834042337,
+      "step": 3000
+    },
+    {
+      "epoch": 22.10144927536232,
+      "grad_norm": 2.398688793182373,
+      "learning_rate": 9.860846438482886e-05,
+      "loss": 0.708,
+      "step": 3050
+    },
+    {
+      "epoch": 22.463768115942027,
+      "grad_norm": 2.0567405223846436,
+      "learning_rate": 9.848126734505088e-05,
+      "loss": 0.7062,
+      "step": 3100
+    },
+    {
+      "epoch": 22.82608695652174,
+      "grad_norm": 2.001675844192505,
+      "learning_rate": 9.83540703052729e-05,
+      "loss": 0.7016,
+      "step": 3150
+    },
+    {
+      "epoch": 23.18840579710145,
+      "grad_norm": 2.003246307373047,
+      "learning_rate": 9.822687326549491e-05,
+      "loss": 0.6884,
+      "step": 3200
+    },
+    {
+      "epoch": 23.55072463768116,
+      "grad_norm": 2.49794340133667,
+      "learning_rate": 9.809967622571694e-05,
+      "loss": 0.6846,
+      "step": 3250
+    },
+    {
+      "epoch": 23.91304347826087,
+      "grad_norm": 2.2691776752471924,
+      "learning_rate": 9.797247918593895e-05,
+      "loss": 0.6746,
+      "step": 3300
+    },
+    {
+      "epoch": 24.27536231884058,
+      "grad_norm": 1.9415994882583618,
+      "learning_rate": 9.784528214616096e-05,
+      "loss": 0.6854,
+      "step": 3350
+    },
+    {
+      "epoch": 24.63768115942029,
+      "grad_norm": 2.4430294036865234,
+      "learning_rate": 9.771808510638299e-05,
+      "loss": 0.6747,
+      "step": 3400
+    },
+    {
+      "epoch": 25.0,
+      "grad_norm": 1.759033203125,
+      "learning_rate": 9.7590888066605e-05,
+      "loss": 0.6505,
+      "step": 3450
+    },
+    {
+      "epoch": 25.36231884057971,
+      "grad_norm": 1.9296367168426514,
+      "learning_rate": 9.746369102682701e-05,
+      "loss": 0.6642,
+      "step": 3500
+    },
+    {
+      "epoch": 25.72463768115942,
+      "grad_norm": 1.9138261079788208,
+      "learning_rate": 9.733649398704904e-05,
+      "loss": 0.6469,
+      "step": 3550
+    },
+    {
+      "epoch": 26.08695652173913,
+      "grad_norm": 2.2081732749938965,
+      "learning_rate": 9.720929694727105e-05,
+      "loss": 0.6512,
+      "step": 3600
+    },
+    {
+      "epoch": 26.44927536231884,
+      "grad_norm": 1.6682246923446655,
+      "learning_rate": 9.708209990749306e-05,
+      "loss": 0.6379,
+      "step": 3650
+    },
+    {
+      "epoch": 26.81159420289855,
+      "grad_norm": 2.3980038166046143,
+      "learning_rate": 9.695490286771509e-05,
+      "loss": 0.6365,
+      "step": 3700
+    },
+    {
+      "epoch": 27.17391304347826,
+      "grad_norm": 1.7031484842300415,
+      "learning_rate": 9.683024976873265e-05,
+      "loss": 0.6253,
+      "step": 3750
+    },
+    {
+      "epoch": 27.536231884057973,
+      "grad_norm": 2.0912516117095947,
+      "learning_rate": 9.670305272895468e-05,
+      "loss": 0.6257,
+      "step": 3800
+    },
+    {
+      "epoch": 27.89855072463768,
+      "grad_norm": 2.1551194190979004,
+      "learning_rate": 9.657585568917669e-05,
+      "loss": 0.6266,
+      "step": 3850
+    },
+    {
+      "epoch": 28.26086956521739,
+      "grad_norm": 1.8378657102584839,
+      "learning_rate": 9.64486586493987e-05,
+      "loss": 0.617,
+      "step": 3900
+    },
+    {
+      "epoch": 28.6231884057971,
+      "grad_norm": 1.9624013900756836,
+      "learning_rate": 9.632146160962073e-05,
+      "loss": 0.6194,
+      "step": 3950
+    },
+    {
+      "epoch": 28.985507246376812,
+      "grad_norm": 1.9267085790634155,
+      "learning_rate": 9.619426456984274e-05,
+      "loss": 0.6052,
+      "step": 4000
+    },
+    {
+      "epoch": 58.69565217391305,
+      "grad_norm": 1.3925776481628418,
+      "learning_rate": 9.606706753006477e-05,
+      "loss": 0.5858,
+      "step": 4050
+    },
+    {
+      "epoch": 59.42028985507246,
+      "grad_norm": 1.3453298807144165,
+      "learning_rate": 9.593987049028678e-05,
+      "loss": 0.5715,
+      "step": 4100
+    },
+    {
+      "epoch": 60.14492753623188,
+      "grad_norm": 1.5027530193328857,
+      "learning_rate": 9.581267345050879e-05,
+      "loss": 0.5599,
+      "step": 4150
+    },
+    {
+      "epoch": 60.869565217391305,
+      "grad_norm": 1.3528389930725098,
+      "learning_rate": 9.568547641073081e-05,
+      "loss": 0.5492,
+      "step": 4200
+    },
+    {
+      "epoch": 61.594202898550726,
+      "grad_norm": 1.344739317893982,
+      "learning_rate": 9.555827937095283e-05,
+      "loss": 0.5386,
+      "step": 4250
+    },
+    {
+      "epoch": 62.31884057971015,
+      "grad_norm": 1.4145418405532837,
+      "learning_rate": 9.543108233117484e-05,
+      "loss": 0.5349,
+      "step": 4300
+    },
+    {
+      "epoch": 63.04347826086956,
+      "grad_norm": 1.4277065992355347,
+      "learning_rate": 9.530388529139686e-05,
+      "loss": 0.5177,
+      "step": 4350
+    },
+    {
+      "epoch": 63.768115942028984,
+      "grad_norm": 1.4313771724700928,
+      "learning_rate": 9.517668825161888e-05,
+      "loss": 0.5193,
+      "step": 4400
+    },
+    {
+      "epoch": 64.4927536231884,
+      "grad_norm": 1.3793104887008667,
+      "learning_rate": 9.504949121184089e-05,
+      "loss": 0.5119,
+      "step": 4450
+    },
+    {
+      "epoch": 65.21739130434783,
+      "grad_norm": 1.312393307685852,
+      "learning_rate": 9.492229417206291e-05,
+      "loss": 0.4959,
+      "step": 4500
+    },
+    {
+      "epoch": 65.21739130434783,
+      "eval_loss": 1.0089221000671387,
+      "eval_runtime": 304.2405,
+      "eval_samples_per_second": 19.281,
+      "eval_steps_per_second": 0.026,
+      "eval_wer": 0.4844446722361745,
+      "step": 4500
+    }
+  ],
+  "logging_steps": 50,
+  "max_steps": 20700,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 300,
+  "save_steps": 500,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 5.0623526280854766e+19,
+  "train_batch_size": 64,
+  "trial_name": null,
+  "trial_params": null
+}

last-checkpoint/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1f5b7536589b8d29039b8f1644bdf0cd508f71d65372555bcc932a87903952a7
+size 6264

last-checkpoint/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)