Training in progress, step 1000

config.json

@@ -1,5 +1,5 @@
 {
-  "_name_or_path": "distil-whisper/distil-large-v3",
+  "_name_or_path": "distil-whisper/distil-small.en",
   "activation_dropout": 0.0,
   "activation_function": "gelu",
   "apply_spec_augment": false,
@@ -7,21 +7,30 @@
     "WhisperForConditionalGeneration"
   ],
   "attention_dropout": 0.0,
-  "begin_suppress_tokens": null,
+  "begin_suppress_tokens": [
+    220,
+    50256
+  ],
   "bos_token_id": 50257,
   "classifier_proj_size": 256,
-  "d_model": 1280,
-  "decoder_attention_heads": 20,
-  "decoder_ffn_dim": 5120,
+  "d_model": 768,
+  "decoder_attention_heads": 12,
+  "decoder_ffn_dim": 3072,
   "decoder_layerdrop": 0.0,
-  "decoder_layers": 2,
-  "decoder_start_token_id": 50258,
+  "decoder_layers": 4,
+  "decoder_start_token_id": 50257,
   "dropout": 0.0,
-  "encoder_attention_heads": 20,
-  "encoder_ffn_dim": 5120,
+  "encoder_attention_heads": 12,
+  "encoder_ffn_dim": 3072,
   "encoder_layerdrop": 0.0,
-  "encoder_layers": 32,
-  "eos_token_id": 50257,
+  "encoder_layers": 12,
+  "eos_token_id": 50256,
+  "forced_decoder_ids": [
+    [
+      1,
+      50362
+    ]
+  ],
   "init_std": 0.02,
   "is_encoder_decoder": true,
   "mask_feature_length": 10,
@@ -35,13 +44,13 @@
   "max_target_positions": 448,
   "median_filter_width": 7,
   "model_type": "whisper",
-  "num_hidden_layers": 32,
-  "num_mel_bins": 128,
+  "num_hidden_layers": 12,
+  "num_mel_bins": 80,
   "pad_token_id": 50256,
   "scale_embedding": false,
   "torch_dtype": "float32",
   "transformers_version": "4.49.0",
   "use_cache": true,
   "use_weighted_layer_sum": false,
-  "vocab_size": 51866
+  "vocab_size": 51864
 }

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:89d592c91312514e897e129efcd6b0dc3235af58d3d158579a1b22edad4698f8
-size 3025686376
+oid sha256:7a320af996889e395ae0159c591b935ea6d8925572741f716a8e3005a7b35667
+size 664561848

preprocessor_config.json

@@ -1,7 +1,7 @@
 {
   "chunk_length": 30,
   "feature_extractor_type": "WhisperFeatureExtractor",
-  "feature_size": 128,
+  "feature_size": 80,
   "hop_length": 160,
   "n_fft": 400,
   "n_samples": 480000,

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b4244374bec7b8447efdd1c42f1cd99d10661bdfa23eaaf1a5ced2d9efef3c4c
-size 5432
+oid sha256:2a851bdcf386b7cc3887546bfb6dadfbd92e7a2dd1ff500d5c01a0bec40beaa7
+size 5496