End of training

README.md

@@ -4,11 +4,24 @@ license: apache-2.0
 base_model: openai/whisper-base
 tags:
 - generated_from_trainer
+datasets:
+- razhan/DOLMA-speech
 metrics:
 - wer
 model-index:
 - name: whisper-base-hac
-  results: []
+  results:
+  - task:
+      name: Automatic Speech Recognition
+      type: automatic-speech-recognition
+    dataset:
+      name: razhan/DOLMA-speech hawrami
+      type: razhan/DOLMA-speech
+      args: hawrami
+    metrics:
+    - name: Wer
+      type: wer
+      value: 0.47917770477906113
 ---
 
 <!-- This model card has been generated automatically according to the information the Trainer had access to. You
@@ -16,7 +29,7 @@ should probably proofread and complete it, then remove this comment. -->
 
 # whisper-base-hac
 
-This model is a fine-tuned version of [openai/whisper-base](https://huggingface.co/openai/whisper-base) on an unknown dataset.
+This model is a fine-tuned version of [openai/whisper-base](https://huggingface.co/openai/whisper-base) on the razhan/DOLMA-speech hawrami dataset.
 It achieves the following results on the evaluation set:
 - Loss: 0.3272
 - Wer: 0.4792

all_results.json

@@ -1,17 +1,16 @@
 {
-    "epoch": 4.0,
-    "eval_cer": 0.26081384171739824,
-    "eval_loss": 0.8250513076782227,
-    "eval_model_preparation_time": 0.0031,
-    "eval_runtime": 88.7908,
+    "epoch": 5.0,
+    "eval_cer": 0.10385057579584221,
+    "eval_loss": 0.32716336846351624,
+    "eval_runtime": 86.6976,
     "eval_samples": 1263,
-    "eval_samples_per_second": 14.224,
-    "eval_steps_per_second": 0.056,
-    "eval_wer": 0.9221150789445799,
-    "total_flos": 1.261916109448151e+18,
-    "train_loss": 2.151222774856969,
-    "train_runtime": 1211.2403,
-    "train_samples": 4773,
-    "train_samples_per_second": 15.762,
-    "train_steps_per_second": 0.063
+    "eval_samples_per_second": 14.568,
+    "eval_steps_per_second": 0.058,
+    "eval_wer": 0.47917770477906113,
+    "total_flos": 3.3623422081150484e+18,
+    "train_loss": 0.8413535683243363,
+    "train_runtime": 3448.7065,
+    "train_samples": 10165,
+    "train_samples_per_second": 14.737,
+    "train_steps_per_second": 0.039
 }

eval_results.json

@@ -1,10 +1,10 @@
 {
-    "eval_cer": 0.26081384171739824,
-    "eval_loss": 0.8250513076782227,
-    "eval_model_preparation_time": 0.0031,
-    "eval_runtime": 88.7908,
+    "epoch": 5.0,
+    "eval_cer": 0.10385057579584221,
+    "eval_loss": 0.32716336846351624,
+    "eval_runtime": 86.6976,
     "eval_samples": 1263,
-    "eval_samples_per_second": 14.224,
-    "eval_steps_per_second": 0.056,
-    "eval_wer": 0.9221150789445799
+    "eval_samples_per_second": 14.568,
+    "eval_steps_per_second": 0.058,
+    "eval_wer": 0.47917770477906113
 }

train_results.json

@@ -1,9 +1,9 @@
 {
-    "epoch": 4.0,
-    "total_flos": 1.261916109448151e+18,
-    "train_loss": 2.151222774856969,
-    "train_runtime": 1211.2403,
-    "train_samples": 4773,
-    "train_samples_per_second": 15.762,
-    "train_steps_per_second": 0.063
+    "epoch": 5.0,
+    "total_flos": 3.3623422081150484e+18,
+    "train_loss": 0.8413535683243363,
+    "train_runtime": 3448.7065,
+    "train_samples": 10165,
+    "train_samples_per_second": 14.737,
+    "train_steps_per_second": 0.039
 }

trainer_state.json

@@ -1,116 +1,168 @@
 {
   "best_metric": null,
   "best_model_checkpoint": null,
-  "epoch": 4.0,
+  "epoch": 5.0,
   "eval_steps": 500,
-  "global_step": 76,
+  "global_step": 135,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
   "log_history": [
     {
-      "epoch": 0.5263157894736842,
-      "grad_norm": 64.39277648925781,
-      "learning_rate": 4.0000000000000003e-07,
-      "loss": 3.6464,
+      "epoch": 0.37037037037037035,
+      "grad_norm": 34.00384521484375,
+      "learning_rate": 9.776119402985076e-06,
+      "loss": 3.5666,
       "step": 10
     },
     {
-      "epoch": 1.0,
-      "eval_cer": 0.45834372661511597,
-      "eval_loss": 3.158611297607422,
-      "eval_runtime": 19.3102,
-      "eval_samples_per_second": 10.564,
-      "eval_steps_per_second": 0.052,
-      "eval_wer": 0.9852579852579852,
-      "step": 19
-    },
-    {
-      "epoch": 1.0526315789473684,
-      "grad_norm": 47.9715576171875,
-      "learning_rate": 1.4000000000000001e-06,
-      "loss": 3.405,
+      "epoch": 0.7407407407407407,
+      "grad_norm": 14.095426559448242,
+      "learning_rate": 9.029850746268657e-06,
+      "loss": 1.9468,
       "step": 20
     },
     {
-      "epoch": 1.5789473684210527,
-      "grad_norm": 30.349884033203125,
-      "learning_rate": 2.4000000000000003e-06,
-      "loss": 2.797,
-      "step": 30
+      "epoch": 1.0,
+      "eval_cer": 0.3796859980775393,
+      "eval_loss": 0.9999122619628906,
+      "eval_runtime": 91.0467,
+      "eval_samples_per_second": 13.872,
+      "eval_steps_per_second": 0.055,
+      "eval_wer": 0.9633358058705097,
+      "step": 27
     },
     {
-      "epoch": 2.0,
-      "eval_cer": 0.4428785233225243,
-      "eval_loss": 2.0119194984436035,
-      "eval_runtime": 18.2983,
-      "eval_samples_per_second": 11.149,
-      "eval_steps_per_second": 0.055,
-      "eval_wer": 0.9963144963144963,
-      "step": 38
+      "epoch": 1.1111111111111112,
+      "grad_norm": 6.282424449920654,
+      "learning_rate": 8.283582089552239e-06,
+      "loss": 1.1917,
+      "step": 30
     },
     {
-      "epoch": 2.1052631578947367,
-      "grad_norm": 24.85350227355957,
-      "learning_rate": 3.4000000000000005e-06,
-      "loss": 2.1901,
+      "epoch": 1.4814814814814814,
+      "grad_norm": 4.256287574768066,
+      "learning_rate": 7.537313432835821e-06,
+      "loss": 0.8244,
       "step": 40
     },
     {
-      "epoch": 2.6315789473684212,
-      "grad_norm": 15.238232612609863,
-      "learning_rate": 4.4e-06,
-      "loss": 1.6366,
+      "epoch": 1.8518518518518519,
+      "grad_norm": 2.4645814895629883,
+      "learning_rate": 6.791044776119403e-06,
+      "loss": 0.6132,
       "step": 50
     },
     {
-      "epoch": 3.0,
-      "eval_cer": 0.3800199551010227,
-      "eval_loss": 1.2072163820266724,
-      "eval_runtime": 17.6074,
-      "eval_samples_per_second": 11.586,
+      "epoch": 2.0,
+      "eval_cer": 0.1469174661213412,
+      "eval_loss": 0.468128502368927,
+      "eval_runtime": 87.6022,
+      "eval_samples_per_second": 14.417,
       "eval_steps_per_second": 0.057,
-      "eval_wer": 0.992014742014742,
-      "step": 57
+      "eval_wer": 0.6078202818692381,
+      "step": 54
     },
     {
-      "epoch": 3.1578947368421053,
-      "grad_norm": 8.841925621032715,
-      "learning_rate": 5.400000000000001e-06,
-      "loss": 1.2455,
+      "epoch": 2.2222222222222223,
+      "grad_norm": 2.1577141284942627,
+      "learning_rate": 6.044776119402986e-06,
+      "loss": 0.4936,
       "step": 60
     },
     {
-      "epoch": 3.6842105263157894,
-      "grad_norm": 4.40842342376709,
-      "learning_rate": 6.4000000000000006e-06,
-      "loss": 0.9577,
+      "epoch": 2.5925925925925926,
+      "grad_norm": 2.0963618755340576,
+      "learning_rate": 5.298507462686567e-06,
+      "loss": 0.4327,
       "step": 70
     },
     {
-      "epoch": 4.0,
-      "eval_cer": 0.24893988525816912,
-      "eval_loss": 0.7948786020278931,
-      "eval_runtime": 16.9884,
-      "eval_samples_per_second": 12.008,
-      "eval_steps_per_second": 0.059,
-      "eval_wer": 0.8052825552825553,
-      "step": 76
+      "epoch": 2.962962962962963,
+      "grad_norm": 1.5172204971313477,
+      "learning_rate": 4.5522388059701495e-06,
+      "loss": 0.3976,
+      "step": 80
+    },
+    {
+      "epoch": 3.0,
+      "eval_cer": 0.11282205929472078,
+      "eval_loss": 0.366842657327652,
+      "eval_runtime": 86.2276,
+      "eval_samples_per_second": 14.647,
+      "eval_steps_per_second": 0.058,
+      "eval_wer": 0.5160538306665253,
+      "step": 81
+    },
+    {
+      "epoch": 3.3333333333333335,
+      "grad_norm": 1.841143012046814,
+      "learning_rate": 3.8059701492537314e-06,
+      "loss": 0.3747,
+      "step": 90
+    },
+    {
+      "epoch": 3.7037037037037037,
+      "grad_norm": 1.7737436294555664,
+      "learning_rate": 3.0597014925373137e-06,
+      "loss": 0.3485,
+      "step": 100
     },
     {
       "epoch": 4.0,
-      "step": 76,
-      "total_flos": 1.261916109448151e+18,
-      "train_loss": 2.151222774856969,
-      "train_runtime": 1211.2403,
-      "train_samples_per_second": 15.762,
-      "train_steps_per_second": 0.063
+      "eval_cer": 0.10645155210434061,
+      "eval_loss": 0.33595842123031616,
+      "eval_runtime": 86.4694,
+      "eval_samples_per_second": 14.606,
+      "eval_steps_per_second": 0.058,
+      "eval_wer": 0.488926565645862,
+      "step": 108
+    },
+    {
+      "epoch": 4.074074074074074,
+      "grad_norm": 1.6087193489074707,
+      "learning_rate": 2.3134328358208956e-06,
+      "loss": 0.3447,
+      "step": 110
+    },
+    {
+      "epoch": 4.444444444444445,
+      "grad_norm": 1.753946304321289,
+      "learning_rate": 1.5671641791044779e-06,
+      "loss": 0.338,
+      "step": 120
+    },
+    {
+      "epoch": 4.814814814814815,
+      "grad_norm": 1.5161445140838623,
+      "learning_rate": 8.208955223880598e-07,
+      "loss": 0.3292,
+      "step": 130
+    },
+    {
+      "epoch": 5.0,
+      "eval_cer": 0.10385057579584221,
+      "eval_loss": 0.32716336846351624,
+      "eval_runtime": 85.8417,
+      "eval_samples_per_second": 14.713,
+      "eval_steps_per_second": 0.058,
+      "eval_wer": 0.47917770477906113,
+      "step": 135
+    },
+    {
+      "epoch": 5.0,
+      "step": 135,
+      "total_flos": 3.3623422081150484e+18,
+      "train_loss": 0.8413535683243363,
+      "train_runtime": 3448.7065,
+      "train_samples_per_second": 14.737,
+      "train_steps_per_second": 0.039
     }
   ],
   "logging_steps": 10,
-  "max_steps": 76,
+  "max_steps": 135,
   "num_input_tokens_seen": 0,
-  "num_train_epochs": 4,
+  "num_train_epochs": 5,
   "save_steps": 500,
   "stateful_callbacks": {
     "TrainerControl": {
@@ -124,8 +176,8 @@
       "attributes": {}
     }
   },
-  "total_flos": 1.261916109448151e+18,
-  "train_batch_size": 128,
+  "total_flos": 3.3623422081150484e+18,
+  "train_batch_size": 192,
   "trial_name": null,
   "trial_params": null
 }