End of training

README.md

@@ -16,9 +16,9 @@ should probably proofread and complete it, then remove this comment. -->
 
 This model was trained from scratch on an unknown dataset.
 It achieves the following results on the evaluation set:
-- Loss: 418.4753
-- Wer: 0.7314
-- Cer: 0.2984
+- Loss: 279.1175
+- Wer: 0.5077
+- Cer: 0.1722
 
 ## Model description
 

all_results.json

@@ -0,0 +1,16 @@
+{
+    "epoch": 12.0,
+    "eval_cer": 0.1721829768314973,
+    "eval_loss": 279.11749267578125,
+    "eval_runtime": 24.6629,
+    "eval_samples": 890,
+    "eval_samples_per_second": 36.087,
+    "eval_steps_per_second": 4.541,
+    "eval_wer": 0.5077335858585859,
+    "total_flos": 3.6380056389804335e+19,
+    "train_loss": 929.5830404722528,
+    "train_runtime": 31654.3209,
+    "train_samples": 58040,
+    "train_samples_per_second": 110.013,
+    "train_steps_per_second": 6.877
+}

eval--in domain_results.json

@@ -0,0 +1,10 @@
+{
+    "epoch": 12.0,
+    "eval_cer": 0.1721829768314973,
+    "eval_loss": 279.11749267578125,
+    "eval_runtime": 24.6629,
+    "eval_samples": 890,
+    "eval_samples_per_second": 36.087,
+    "eval_steps_per_second": 4.541,
+    "eval_wer": 0.5077335858585859
+}

train_results.json

@@ -0,0 +1,9 @@
+{
+    "epoch": 12.0,
+    "total_flos": 3.6380056389804335e+19,
+    "train_loss": 929.5830404722528,
+    "train_runtime": 31654.3209,
+    "train_samples": 58040,
+    "train_samples_per_second": 110.013,
+    "train_steps_per_second": 6.877
+}

trainer_state.json

@@ -0,0 +1,255 @@
+{
+  "best_metric": 0.5064709595959596,
+  "best_model_checkpoint": "/scratch/elec/puhe/p/palp3/sami_ASR/base_model_output/wav2vec2-base-sami-cont-pt-22k/outputs/shuff_100_no_idv/wr15/checkpoint-7256",
+  "epoch": 12.0,
+  "eval_steps": 500,
+  "global_step": 43536,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 1.0,
+      "grad_norm": 953.8694458007812,
+      "learning_rate": 5.5341173588141614e-05,
+      "loss": 2898.1974,
+      "step": 3628
+    },
+    {
+      "epoch": 1.0,
+      "eval_cer": 0.18934940473250012,
+      "eval_loss": 315.9604187011719,
+      "eval_runtime": 25.3347,
+      "eval_samples_per_second": 35.13,
+      "eval_steps_per_second": 4.421,
+      "eval_wer": 0.6330492424242424,
+      "step": 3628
+    },
+    {
+      "epoch": 2.0,
+      "grad_norm": 666.8331909179688,
+      "learning_rate": 0.00011088141614602474,
+      "loss": 757.7022,
+      "step": 7256
+    },
+    {
+      "epoch": 2.0,
+      "eval_cer": 0.17208417724645556,
+      "eval_loss": 279.52197265625,
+      "eval_runtime": 25.5129,
+      "eval_samples_per_second": 34.884,
+      "eval_steps_per_second": 4.39,
+      "eval_wer": 0.5064709595959596,
+      "step": 7256
+    },
+    {
+      "epoch": 3.0,
+      "grad_norm": 236.5342559814453,
+      "learning_rate": 0.0001664216587039079,
+      "loss": 662.3705,
+      "step": 10884
+    },
+    {
+      "epoch": 3.0,
+      "eval_cer": 0.17640665909203182,
+      "eval_loss": 301.2926330566406,
+      "eval_runtime": 25.9701,
+      "eval_samples_per_second": 34.27,
+      "eval_steps_per_second": 4.313,
+      "eval_wer": 0.525094696969697,
+      "step": 10884
+    },
+    {
+      "epoch": 4.0,
+      "grad_norm": 595.1166381835938,
+      "learning_rate": 0.000221961901261791,
+      "loss": 647.4321,
+      "step": 14512
+    },
+    {
+      "epoch": 4.0,
+      "eval_cer": 0.2026873487131354,
+      "eval_loss": 309.8589172363281,
+      "eval_runtime": 32.0472,
+      "eval_samples_per_second": 27.772,
+      "eval_steps_per_second": 3.495,
+      "eval_wer": 0.5494002525252525,
+      "step": 14512
+    },
+    {
+      "epoch": 5.0,
+      "grad_norm": 320.42572021484375,
+      "learning_rate": 0.00027747151782432926,
+      "loss": 659.7851,
+      "step": 18140
+    },
+    {
+      "epoch": 5.0,
+      "eval_cer": 0.20700983055871167,
+      "eval_loss": 309.31988525390625,
+      "eval_runtime": 25.5466,
+      "eval_samples_per_second": 34.838,
+      "eval_steps_per_second": 4.384,
+      "eval_wer": 0.5861742424242424,
+      "step": 18140
+    },
+    {
+      "epoch": 6.0,
+      "grad_norm": 1806.6942138671875,
+      "learning_rate": 0.00033302707337988486,
+      "loss": 692.9787,
+      "step": 21768
+    },
+    {
+      "epoch": 6.0,
+      "eval_cer": 0.24119448698315468,
+      "eval_loss": 373.3182678222656,
+      "eval_runtime": 30.9297,
+      "eval_samples_per_second": 28.775,
+      "eval_steps_per_second": 3.621,
+      "eval_wer": 0.6486742424242424,
+      "step": 21768
+    },
+    {
+      "epoch": 7.0,
+      "grad_norm": 305.56292724609375,
+      "learning_rate": 0.00038855200294009556,
+      "loss": 739.3389,
+      "step": 25396
+    },
+    {
+      "epoch": 7.0,
+      "eval_cer": 0.2555945265029887,
+      "eval_loss": 379.1961975097656,
+      "eval_runtime": 24.9788,
+      "eval_samples_per_second": 35.63,
+      "eval_steps_per_second": 4.484,
+      "eval_wer": 0.6764520202020202,
+      "step": 25396
+    },
+    {
+      "epoch": 8.0,
+      "grad_norm": 1148.437744140625,
+      "learning_rate": 0.0004440769325003063,
+      "loss": 786.0181,
+      "step": 29024
+    },
+    {
+      "epoch": 8.0,
+      "eval_cer": 0.2728597539890332,
+      "eval_loss": 414.1454772949219,
+      "eval_runtime": 25.8475,
+      "eval_samples_per_second": 34.433,
+      "eval_steps_per_second": 4.333,
+      "eval_wer": 0.7021780303030303,
+      "step": 29024
+    },
+    {
+      "epoch": 9.0,
+      "grad_norm": 790.4903564453125,
+      "learning_rate": 0.000499601862060517,
+      "loss": 827.0145,
+      "step": 32652
+    },
+    {
+      "epoch": 9.0,
+      "eval_cer": 0.34004347181741834,
+      "eval_loss": 441.2162170410156,
+      "eval_runtime": 27.3805,
+      "eval_samples_per_second": 32.505,
+      "eval_steps_per_second": 4.09,
+      "eval_wer": 0.7870896464646465,
+      "step": 32652
+    },
+    {
+      "epoch": 10.0,
+      "grad_norm": 779.6360473632812,
+      "learning_rate": 0.0004902690403614588,
+      "loss": 849.8929,
+      "step": 36280
+    },
+    {
+      "epoch": 10.0,
+      "eval_cer": 0.2962752556439263,
+      "eval_loss": 422.385498046875,
+      "eval_runtime": 28.0021,
+      "eval_samples_per_second": 31.783,
+      "eval_steps_per_second": 4.0,
+      "eval_wer": 0.7258522727272727,
+      "step": 36280
+    },
+    {
+      "epoch": 11.0,
+      "grad_norm": 399.57818603515625,
+      "learning_rate": 0.0004804651187928314,
+      "loss": 830.547,
+      "step": 39908
+    },
+    {
+      "epoch": 11.0,
+      "eval_cer": 0.3045744207874327,
+      "eval_loss": 426.62811279296875,
+      "eval_runtime": 24.2088,
+      "eval_samples_per_second": 36.763,
+      "eval_steps_per_second": 4.626,
+      "eval_wer": 0.7506313131313131,
+      "step": 39908
+    },
+    {
+      "epoch": 12.0,
+      "grad_norm": 1422.100341796875,
+      "learning_rate": 0.0004706693041053246,
+      "loss": 803.7191,
+      "step": 43536
+    },
+    {
+      "epoch": 12.0,
+      "eval_cer": 0.29839944672232377,
+      "eval_loss": 418.475341796875,
+      "eval_runtime": 26.6689,
+      "eval_samples_per_second": 33.372,
+      "eval_steps_per_second": 4.2,
+      "eval_wer": 0.7313762626262627,
+      "step": 43536
+    },
+    {
+      "epoch": 12.0,
+      "step": 43536,
+      "total_flos": 3.6380056389804335e+19,
+      "train_loss": 929.5830404722528,
+      "train_runtime": 31654.3209,
+      "train_samples_per_second": 110.013,
+      "train_steps_per_second": 6.877
+    }
+  ],
+  "logging_steps": 500,
+  "max_steps": 217680,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 60,
+  "save_steps": 500,
+  "stateful_callbacks": {
+    "EarlyStoppingCallback": {
+      "args": {
+        "early_stopping_patience": 10,
+        "early_stopping_threshold": 0.0
+      },
+      "attributes": {
+        "early_stopping_patience_counter": 10
+      }
+    },
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": true
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.6380056389804335e+19,
+  "train_batch_size": 16,
+  "trial_name": null,
+  "trial_params": null
+}