Training checkpoint

README.md

@@ -1,22 +1,24 @@
 ---
-base_model: openai/whisper-small
 library_name: transformers
+language:
+- en
 license: apache-2.0
-metrics:
-- wer
+base_model: openai/whisper-small
 tags:
 - generated_from_trainer
+metrics:
+- wer
 model-index:
-- name: Whisper-squeezeformer-v3
+- name: Whisper-squeezeformer-NSQU-whisper
   results: []
 ---
 
 <!-- This model card has been generated automatically according to the information the Trainer had access to. You
 should probably proofread and complete it, then remove this comment. -->
 
-# Whisper-squeezeformer-v3
+# Whisper-squeezeformer-NSQU-whisper
 
-This model is a fine-tuned version of [openai/whisper-small](https://huggingface.co/openai/whisper-small) on the None dataset.
+This model is a fine-tuned version of [openai/whisper-small](https://huggingface.co/openai/whisper-small) on the LibriSpeech dataset.
 It achieves the following results on the evaluation set:
 - Loss: 0.1511
 - Wer: 6.8035

trainer_state.json

@@ -1,9 +1,9 @@
 {
-  "best_metric": 7.149650030432136,
-  "best_model_checkpoint": "./Whisper-squeezeformer-v3\\checkpoint-20000",
-  "epoch": 8.0,
+  "best_metric": 6.75973828362751,
+  "best_model_checkpoint": "./Whisper-squeezeformer-v3\\checkpoint-27500",
+  "epoch": 12.0,
   "eval_steps": 2500,
-  "global_step": 20000,
+  "global_step": 30000,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
@@ -137,19 +137,83 @@
       "step": 20000
     },
     {
-      "epoch": 8.0,
-      "step": 20000,
-      "total_flos": 1.38658480128e+20,
-      "train_loss": 0.047600341033935546,
-      "train_runtime": 48327.6236,
-      "train_samples_per_second": 8.277,
-      "train_steps_per_second": 0.414
+      "epoch": 9.0,
+      "grad_norm": 4.3119354248046875,
+      "learning_rate": 2.731636363636364e-06,
+      "loss": 0.2168,
+      "step": 22500
+    },
+    {
+      "epoch": 9.0,
+      "eval_loss": 0.15093587338924408,
+      "eval_runtime": 583.1378,
+      "eval_samples_per_second": 4.493,
+      "eval_steps_per_second": 0.562,
+      "eval_wer": 7.050745587340232,
+      "step": 22500
+    },
+    {
+      "epoch": 10.0,
+      "grad_norm": 3.5476253032684326,
+      "learning_rate": 1.822909090909091e-06,
+      "loss": 0.1467,
+      "step": 25000
+    },
+    {
+      "epoch": 10.0,
+      "eval_loss": 0.1494235247373581,
+      "eval_runtime": 591.6025,
+      "eval_samples_per_second": 4.429,
+      "eval_steps_per_second": 0.554,
+      "eval_wer": 6.967057212416311,
+      "step": 25000
+    },
+    {
+      "epoch": 11.0,
+      "grad_norm": 3.3021419048309326,
+      "learning_rate": 9.141818181818182e-07,
+      "loss": 0.1113,
+      "step": 27500
+    },
+    {
+      "epoch": 11.0,
+      "eval_loss": 0.14934100210666656,
+      "eval_runtime": 584.5502,
+      "eval_samples_per_second": 4.482,
+      "eval_steps_per_second": 0.561,
+      "eval_wer": 6.75973828362751,
+      "step": 27500
+    },
+    {
+      "epoch": 12.0,
+      "grad_norm": 2.4910900592803955,
+      "learning_rate": 5.090909090909091e-09,
+      "loss": 0.0914,
+      "step": 30000
+    },
+    {
+      "epoch": 12.0,
+      "eval_loss": 0.15113578736782074,
+      "eval_runtime": 619.6762,
+      "eval_samples_per_second": 4.228,
+      "eval_steps_per_second": 0.529,
+      "eval_wer": 6.80348447961047,
+      "step": 30000
+    },
+    {
+      "epoch": 12.0,
+      "step": 30000,
+      "total_flos": 2.07987720192e+20,
+      "train_loss": 0.04717991434733073,
+      "train_runtime": 48062.4633,
+      "train_samples_per_second": 12.484,
+      "train_steps_per_second": 0.624
     }
   ],
   "logging_steps": 2500,
-  "max_steps": 20000,
+  "max_steps": 30000,
   "num_input_tokens_seen": 0,
-  "num_train_epochs": 8,
+  "num_train_epochs": 12,
   "save_steps": 2500,
   "stateful_callbacks": {
     "TrainerControl": {
@@ -163,7 +227,7 @@
       "attributes": {}
     }
   },
-  "total_flos": 1.38658480128e+20,
+  "total_flos": 2.07987720192e+20,
   "train_batch_size": 20,
   "trial_name": null,
   "trial_params": null