End of training

README.md

@@ -4,11 +4,24 @@ license: apache-2.0
 base_model: razhan/whisper-base-hawrami
 tags:
 - generated_from_trainer
+datasets:
+- razhan/DOLMA-speech
 metrics:
 - wer
 model-index:
 - name: whisper-base-hawrami-transcription
-  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: 0.40128824476650565
 ---
 
 <!-- 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-hawrami-transcription
 
-This model is a fine-tuned version of [razhan/whisper-base-hawrami](https://huggingface.co/razhan/whisper-base-hawrami) on an unknown dataset.
+This model is a fine-tuned version of [razhan/whisper-base-hawrami](https://huggingface.co/razhan/whisper-base-hawrami) on the razhan/DOLMA-speech hawrami dataset.
 It achieves the following results on the evaluation set:
 - Loss: 0.2612
 - Wer: 0.4013

all_results.json

@@ -1,15 +1,16 @@
 {
     "epoch": 4.0,
-    "eval_loss": 3.5885000228881836,
-    "eval_runtime": 9.5147,
-    "eval_samples": 64,
-    "eval_samples_per_second": 6.726,
-    "eval_steps_per_second": 0.105,
-    "eval_wer": 2.3790849673202614,
-    "total_flos": 3.320831803392e+16,
-    "train_loss": 3.6943881511688232,
-    "train_runtime": 99.8208,
-    "train_samples": 128,
-    "train_samples_per_second": 5.129,
-    "train_steps_per_second": 0.04
+    "eval_cer": 0.08564332792052727,
+    "eval_loss": 0.2611912190914154,
+    "eval_runtime": 165.3715,
+    "eval_samples": 1263,
+    "eval_samples_per_second": 7.637,
+    "eval_steps_per_second": 0.06,
+    "eval_wer": 0.40128824476650565,
+    "total_flos": 2.6372074438656e+18,
+    "train_loss": 0.7337436556816102,
+    "train_runtime": 5396.4455,
+    "train_samples": 10165,
+    "train_samples_per_second": 7.535,
+    "train_steps_per_second": 0.059
 }

eval_results.json

@@ -1,9 +1,10 @@
 {
     "epoch": 4.0,
-    "eval_loss": 3.5885000228881836,
-    "eval_runtime": 9.5147,
-    "eval_samples": 64,
-    "eval_samples_per_second": 6.726,
-    "eval_steps_per_second": 0.105,
-    "eval_wer": 2.3790849673202614
+    "eval_cer": 0.08564332792052727,
+    "eval_loss": 0.2611912190914154,
+    "eval_runtime": 165.3715,
+    "eval_samples": 1263,
+    "eval_samples_per_second": 7.637,
+    "eval_steps_per_second": 0.06,
+    "eval_wer": 0.40128824476650565
 }

train_results.json

@@ -1,9 +1,9 @@
 {
     "epoch": 4.0,
-    "total_flos": 3.320831803392e+16,
-    "train_loss": 3.6943881511688232,
-    "train_runtime": 99.8208,
-    "train_samples": 128,
-    "train_samples_per_second": 5.129,
-    "train_steps_per_second": 0.04
+    "total_flos": 2.6372074438656e+18,
+    "train_loss": 0.7337436556816102,
+    "train_runtime": 5396.4455,
+    "train_samples": 10165,
+    "train_samples_per_second": 7.535,
+    "train_steps_per_second": 0.059
 }

trainer_state.json

@@ -3,59 +3,287 @@
   "best_model_checkpoint": null,
   "epoch": 4.0,
   "eval_steps": 500,
-  "global_step": 4,
+  "global_step": 320,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
   "log_history": [
+    {
+      "epoch": 0.125,
+      "grad_norm": 71.75464630126953,
+      "learning_rate": 4.0000000000000003e-07,
+      "loss": 3.61,
+      "step": 10
+    },
+    {
+      "epoch": 0.25,
+      "grad_norm": 45.94756317138672,
+      "learning_rate": 1.4000000000000001e-06,
+      "loss": 3.3547,
+      "step": 20
+    },
+    {
+      "epoch": 0.375,
+      "grad_norm": 29.31285858154297,
+      "learning_rate": 2.4000000000000003e-06,
+      "loss": 2.7402,
+      "step": 30
+    },
+    {
+      "epoch": 0.5,
+      "grad_norm": 19.44002914428711,
+      "learning_rate": 3.4000000000000005e-06,
+      "loss": 2.12,
+      "step": 40
+    },
+    {
+      "epoch": 0.625,
+      "grad_norm": 13.159948348999023,
+      "learning_rate": 4.4e-06,
+      "loss": 1.5923,
+      "step": 50
+    },
+    {
+      "epoch": 0.75,
+      "grad_norm": 8.778341293334961,
+      "learning_rate": 5.400000000000001e-06,
+      "loss": 1.2181,
+      "step": 60
+    },
+    {
+      "epoch": 0.875,
+      "grad_norm": 5.181447982788086,
+      "learning_rate": 6.4000000000000006e-06,
+      "loss": 0.9039,
+      "step": 70
+    },
+    {
+      "epoch": 1.0,
+      "grad_norm": 5.2995686531066895,
+      "learning_rate": 7.4e-06,
+      "loss": 0.6798,
+      "step": 80
+    },
     {
       "epoch": 1.0,
-      "eval_loss": 3.5885000228881836,
-      "eval_runtime": 9.5223,
-      "eval_samples_per_second": 6.721,
-      "eval_steps_per_second": 0.105,
-      "eval_wer": 2.3790849673202614,
-      "step": 1
+      "eval_cer": 0.1788136402712771,
+      "eval_loss": 0.5512658953666687,
+      "eval_runtime": 167.9142,
+      "eval_samples_per_second": 7.522,
+      "eval_steps_per_second": 0.06,
+      "eval_wer": 0.6854535695115406,
+      "step": 80
+    },
+    {
+      "epoch": 1.125,
+      "grad_norm": 3.869173288345337,
+      "learning_rate": 8.400000000000001e-06,
+      "loss": 0.5485,
+      "step": 90
+    },
+    {
+      "epoch": 1.25,
+      "grad_norm": 3.1130523681640625,
+      "learning_rate": 9.4e-06,
+      "loss": 0.4481,
+      "step": 100
+    },
+    {
+      "epoch": 1.375,
+      "grad_norm": 2.2851545810699463,
+      "learning_rate": 9.81818181818182e-06,
+      "loss": 0.4103,
+      "step": 110
+    },
+    {
+      "epoch": 1.5,
+      "grad_norm": 2.8455662727355957,
+      "learning_rate": 9.363636363636365e-06,
+      "loss": 0.37,
+      "step": 120
+    },
+    {
+      "epoch": 1.625,
+      "grad_norm": 2.575657367706299,
+      "learning_rate": 8.90909090909091e-06,
+      "loss": 0.3513,
+      "step": 130
+    },
+    {
+      "epoch": 1.75,
+      "grad_norm": 2.5928118228912354,
+      "learning_rate": 8.454545454545455e-06,
+      "loss": 0.3504,
+      "step": 140
+    },
+    {
+      "epoch": 1.875,
+      "grad_norm": 2.5617828369140625,
+      "learning_rate": 8.000000000000001e-06,
+      "loss": 0.324,
+      "step": 150
+    },
+    {
+      "epoch": 2.0,
+      "grad_norm": 3.5175580978393555,
+      "learning_rate": 7.545454545454546e-06,
+      "loss": 0.3095,
+      "step": 160
     },
     {
       "epoch": 2.0,
-      "eval_loss": 3.5885000228881836,
-      "eval_runtime": 9.4355,
-      "eval_samples_per_second": 6.783,
-      "eval_steps_per_second": 0.106,
-      "eval_wer": 2.3790849673202614,
-      "step": 2
+      "eval_cer": 0.09716305282261917,
+      "eval_loss": 0.2984148859977722,
+      "eval_runtime": 164.9586,
+      "eval_samples_per_second": 7.656,
+      "eval_steps_per_second": 0.061,
+      "eval_wer": 0.4486312399355878,
+      "step": 160
+    },
+    {
+      "epoch": 2.125,
+      "grad_norm": 2.2015202045440674,
+      "learning_rate": 7.0909090909090916e-06,
+      "loss": 0.2926,
+      "step": 170
+    },
+    {
+      "epoch": 2.25,
+      "grad_norm": 3.0814766883850098,
+      "learning_rate": 6.6363636363636375e-06,
+      "loss": 0.2836,
+      "step": 180
+    },
+    {
+      "epoch": 2.375,
+      "grad_norm": 1.8548959493637085,
+      "learning_rate": 6.181818181818182e-06,
+      "loss": 0.2743,
+      "step": 190
+    },
+    {
+      "epoch": 2.5,
+      "grad_norm": 2.3138978481292725,
+      "learning_rate": 5.727272727272728e-06,
+      "loss": 0.2707,
+      "step": 200
+    },
+    {
+      "epoch": 2.625,
+      "grad_norm": 2.2470438480377197,
+      "learning_rate": 5.272727272727273e-06,
+      "loss": 0.2799,
+      "step": 210
+    },
+    {
+      "epoch": 2.75,
+      "grad_norm": 2.5589115619659424,
+      "learning_rate": 4.818181818181819e-06,
+      "loss": 0.2669,
+      "step": 220
+    },
+    {
+      "epoch": 2.875,
+      "grad_norm": 2.1574714183807373,
+      "learning_rate": 4.363636363636364e-06,
+      "loss": 0.2615,
+      "step": 230
+    },
+    {
+      "epoch": 3.0,
+      "grad_norm": 3.022969961166382,
+      "learning_rate": 3.90909090909091e-06,
+      "loss": 0.2673,
+      "step": 240
     },
     {
       "epoch": 3.0,
-      "eval_loss": 3.5885000228881836,
-      "eval_runtime": 9.2865,
-      "eval_samples_per_second": 6.892,
-      "eval_steps_per_second": 0.108,
-      "eval_wer": 2.3790849673202614,
-      "step": 3
+      "eval_cer": 0.0882032667876588,
+      "eval_loss": 0.2675623595714569,
+      "eval_runtime": 162.9283,
+      "eval_samples_per_second": 7.752,
+      "eval_steps_per_second": 0.061,
+      "eval_wer": 0.4142780461621041,
+      "step": 240
+    },
+    {
+      "epoch": 3.125,
+      "grad_norm": 1.893800973892212,
+      "learning_rate": 3.454545454545455e-06,
+      "loss": 0.2601,
+      "step": 250
+    },
+    {
+      "epoch": 3.25,
+      "grad_norm": 2.2726192474365234,
+      "learning_rate": 3e-06,
+      "loss": 0.2575,
+      "step": 260
+    },
+    {
+      "epoch": 3.375,
+      "grad_norm": 1.9179538488388062,
+      "learning_rate": 2.5454545454545456e-06,
+      "loss": 0.2375,
+      "step": 270
+    },
+    {
+      "epoch": 3.5,
+      "grad_norm": 1.734628677368164,
+      "learning_rate": 2.090909090909091e-06,
+      "loss": 0.238,
+      "step": 280
+    },
+    {
+      "epoch": 3.625,
+      "grad_norm": 1.9225579500198364,
+      "learning_rate": 1.6363636363636365e-06,
+      "loss": 0.237,
+      "step": 290
+    },
+    {
+      "epoch": 3.75,
+      "grad_norm": 1.7918205261230469,
+      "learning_rate": 1.181818181818182e-06,
+      "loss": 0.2364,
+      "step": 300
+    },
+    {
+      "epoch": 3.875,
+      "grad_norm": 2.021205186843872,
+      "learning_rate": 7.272727272727273e-07,
+      "loss": 0.2422,
+      "step": 310
+    },
+    {
+      "epoch": 4.0,
+      "grad_norm": 2.901258707046509,
+      "learning_rate": 2.7272727272727274e-07,
+      "loss": 0.2428,
+      "step": 320
     },
     {
       "epoch": 4.0,
-      "eval_loss": 3.5885000228881836,
-      "eval_runtime": 9.4589,
-      "eval_samples_per_second": 6.766,
-      "eval_steps_per_second": 0.106,
-      "eval_wer": 2.3790849673202614,
-      "step": 4
+      "eval_cer": 0.08564332792052727,
+      "eval_loss": 0.2611912190914154,
+      "eval_runtime": 163.7635,
+      "eval_samples_per_second": 7.712,
+      "eval_steps_per_second": 0.061,
+      "eval_wer": 0.40128824476650565,
+      "step": 320
     },
     {
       "epoch": 4.0,
-      "step": 4,
-      "total_flos": 3.320831803392e+16,
-      "train_loss": 3.6943881511688232,
-      "train_runtime": 99.8208,
-      "train_samples_per_second": 5.129,
-      "train_steps_per_second": 0.04
+      "step": 320,
+      "total_flos": 2.6372074438656e+18,
+      "train_loss": 0.7337436556816102,
+      "train_runtime": 5396.4455,
+      "train_samples_per_second": 7.535,
+      "train_steps_per_second": 0.059
     }
   ],
   "logging_steps": 10,
-  "max_steps": 4,
+  "max_steps": 320,
   "num_input_tokens_seen": 0,
   "num_train_epochs": 4,
   "save_steps": 500,
@@ -71,8 +299,8 @@
       "attributes": {}
     }
   },
-  "total_flos": 3.320831803392e+16,
-  "train_batch_size": 256,
+  "total_flos": 2.6372074438656e+18,
+  "train_batch_size": 128,
   "trial_name": null,
   "trial_params": null
 }