Fine-tune openai/whisper-tiny on processed data from multiple public data sources

all_results.json

@@ -0,0 +1,8 @@
+{
+    "epoch": 14.0424,
+    "total_flos": 9.8544473358336e+17,
+    "train_loss": 0.3740376678466797,
+    "train_runtime": 6491.8649,
+    "train_samples_per_second": 6.162,
+    "train_steps_per_second": 0.77
+}

train_results.json

@@ -0,0 +1,8 @@
+{
+    "epoch": 14.0424,
+    "total_flos": 9.8544473358336e+17,
+    "train_loss": 0.3740376678466797,
+    "train_runtime": 6491.8649,
+    "train_samples_per_second": 6.162,
+    "train_steps_per_second": 0.77
+}

trainer_state.json

@@ -0,0 +1,203 @@
+{
+  "best_global_step": 4000,
+  "best_metric": 78.72470876762722,
+  "best_model_checkpoint": "./whisper-tiny-ml/checkpoint-4000",
+  "epoch": 14.0424,
+  "eval_steps": 500,
+  "global_step": 5000,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 1.0316,
+      "grad_norm": 12.401665687561035,
+      "learning_rate": 9.960000000000001e-06,
+      "loss": 1.4609,
+      "step": 500
+    },
+    {
+      "epoch": 1.0316,
+      "eval_loss": 1.120520830154419,
+      "eval_runtime": 565.3092,
+      "eval_samples_per_second": 0.665,
+      "eval_steps_per_second": 0.333,
+      "eval_wer": 132.1684038422236,
+      "step": 500
+    },
+    {
+      "epoch": 2.0632,
+      "grad_norm": 17.6810302734375,
+      "learning_rate": 8.895555555555556e-06,
+      "loss": 0.9694,
+      "step": 1000
+    },
+    {
+      "epoch": 2.0632,
+      "eval_loss": 0.7588837742805481,
+      "eval_runtime": 519.1847,
+      "eval_samples_per_second": 0.724,
+      "eval_steps_per_second": 0.362,
+      "eval_wer": 100.81749437972614,
+      "step": 1000
+    },
+    {
+      "epoch": 4.0264,
+      "grad_norm": 8.562993049621582,
+      "learning_rate": 7.784444444444445e-06,
+      "loss": 0.4373,
+      "step": 1500
+    },
+    {
+      "epoch": 4.0264,
+      "eval_loss": 0.24429242312908173,
+      "eval_runtime": 448.0414,
+      "eval_samples_per_second": 0.839,
+      "eval_steps_per_second": 0.42,
+      "eval_wer": 90.72143879010832,
+      "step": 1500
+    },
+    {
+      "epoch": 5.058,
+      "grad_norm": 7.884637832641602,
+      "learning_rate": 6.6733333333333335e-06,
+      "loss": 0.2336,
+      "step": 2000
+    },
+    {
+      "epoch": 5.058,
+      "eval_loss": 0.20778511464595795,
+      "eval_runtime": 418.3161,
+      "eval_samples_per_second": 0.899,
+      "eval_steps_per_second": 0.449,
+      "eval_wer": 84.0997343143266,
+      "step": 2000
+    },
+    {
+      "epoch": 7.0212,
+      "grad_norm": 5.552468299865723,
+      "learning_rate": 5.562222222222222e-06,
+      "loss": 0.1689,
+      "step": 2500
+    },
+    {
+      "epoch": 7.0212,
+      "eval_loss": 0.19417865574359894,
+      "eval_runtime": 385.7038,
+      "eval_samples_per_second": 0.975,
+      "eval_steps_per_second": 0.487,
+      "eval_wer": 79.97138769670958,
+      "step": 2500
+    },
+    {
+      "epoch": 8.0528,
+      "grad_norm": 6.440951824188232,
+      "learning_rate": 4.451111111111112e-06,
+      "loss": 0.1322,
+      "step": 3000
+    },
+    {
+      "epoch": 8.0528,
+      "eval_loss": 0.19918522238731384,
+      "eval_runtime": 398.4291,
+      "eval_samples_per_second": 0.944,
+      "eval_steps_per_second": 0.472,
+      "eval_wer": 79.74657674228489,
+      "step": 3000
+    },
+    {
+      "epoch": 10.016,
+      "grad_norm": 11.55614185333252,
+      "learning_rate": 3.3400000000000006e-06,
+      "loss": 0.1063,
+      "step": 3500
+    },
+    {
+      "epoch": 10.016,
+      "eval_loss": 0.20443612337112427,
+      "eval_runtime": 419.5788,
+      "eval_samples_per_second": 0.896,
+      "eval_steps_per_second": 0.448,
+      "eval_wer": 80.42100960555896,
+      "step": 3500
+    },
+    {
+      "epoch": 11.0476,
+      "grad_norm": 4.469738006591797,
+      "learning_rate": 2.228888888888889e-06,
+      "loss": 0.0892,
+      "step": 4000
+    },
+    {
+      "epoch": 11.0476,
+      "eval_loss": 0.2040003538131714,
+      "eval_runtime": 414.9544,
+      "eval_samples_per_second": 0.906,
+      "eval_steps_per_second": 0.453,
+      "eval_wer": 78.72470876762722,
+      "step": 4000
+    },
+    {
+      "epoch": 13.0108,
+      "grad_norm": 3.349318027496338,
+      "learning_rate": 1.117777777777778e-06,
+      "loss": 0.0755,
+      "step": 4500
+    },
+    {
+      "epoch": 13.0108,
+      "eval_loss": 0.21057051420211792,
+      "eval_runtime": 413.9902,
+      "eval_samples_per_second": 0.908,
+      "eval_steps_per_second": 0.454,
+      "eval_wer": 79.99182505620274,
+      "step": 4500
+    },
+    {
+      "epoch": 14.0424,
+      "grad_norm": 7.616532802581787,
+      "learning_rate": 6.666666666666667e-09,
+      "loss": 0.0672,
+      "step": 5000
+    },
+    {
+      "epoch": 14.0424,
+      "eval_loss": 0.20920291543006897,
+      "eval_runtime": 407.6184,
+      "eval_samples_per_second": 0.922,
+      "eval_steps_per_second": 0.461,
+      "eval_wer": 79.88963825873697,
+      "step": 5000
+    },
+    {
+      "epoch": 14.0424,
+      "step": 5000,
+      "total_flos": 9.8544473358336e+17,
+      "train_loss": 0.3740376678466797,
+      "train_runtime": 6491.8649,
+      "train_samples_per_second": 6.162,
+      "train_steps_per_second": 0.77
+    }
+  ],
+  "logging_steps": 500,
+  "max_steps": 5000,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 9223372036854775807,
+  "save_steps": 500,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": true
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 9.8544473358336e+17,
+  "train_batch_size": 2,
+  "trial_name": null,
+  "trial_params": null
+}