sangeet2020 commited on
Commit
c9a2d6d
1 Parent(s): b3e6c6f

formatting

Browse files
Files changed (2) hide show
  1. config.json +1 -2
  2. hyperparams.yaml +33 -26
config.json CHANGED
@@ -1,4 +1,3 @@
1
  {
2
  "speechbrain_interface": "WhisperASR"
3
- }
4
-
 
1
  {
2
  "speechbrain_interface": "WhisperASR"
3
+ }
 
hyperparams.yaml CHANGED
@@ -17,7 +17,7 @@ language: german
17
  normalized_transcripts: true
18
  test_only: false # Set it to True if you only want to do the evaluation
19
 
20
- auto_mix_prec: false
21
  sample_rate: 16000
22
 
23
  # These values are only used for the searchers.
@@ -31,38 +31,45 @@ bos_index: 50258
31
 
32
  # Decoding parameters
33
  min_decode_ratio: 0.0
34
- max_decode_ratio: 1.0
35
  test_beam_size: 8
36
 
37
  # Model parameters
38
- freeze_whisper: false
39
- freeze_encoder_only: false
40
- freeze_encoder: true
41
 
42
- #
43
- # Functions and classes
44
- #
45
- whisper: &id001 !new:speechbrain.lobes.models.huggingface_whisper.HuggingFaceWhisper
46
- source: openai/whisper-large-v2
47
- freeze: false
48
- save_path: openai/whisper-large-v2
49
- encoder_only: false
50
- freeze_encoder: true
51
 
52
 
53
- modules:
54
- whisper: *id001
55
- whisper_opt_class: !name:torch.optim.AdamW
56
- lr: 0.00003
57
- weight_decay: 0.01
 
 
 
58
 
59
  decoder: !new:speechbrain.decoders.seq2seq.S2SWhisperGreedySearch
60
- model: *id001
61
- bos_index: 50363
62
- eos_index: 50257
63
- min_decode_ratio: 0.0
64
- max_decode_ratio: 1.0
65
 
66
- pretrainer: !new:speechbrain.utils.parameter_transfer.Pretrainer
67
- loadables:
 
 
 
 
 
 
 
 
68
  whisper: !ref <whisper>
 
 
 
 
 
 
 
17
  normalized_transcripts: true
18
  test_only: false # Set it to True if you only want to do the evaluation
19
 
20
+ auto_mix_prec: False
21
  sample_rate: 16000
22
 
23
  # These values are only used for the searchers.
 
31
 
32
  # Decoding parameters
33
  min_decode_ratio: 0.0
34
+ max_decode_ratio: 0.1
35
  test_beam_size: 8
36
 
37
  # Model parameters
38
+ freeze_whisper: True
39
+ freeze_encoder: True
 
40
 
 
 
 
 
 
 
 
 
 
41
 
42
 
43
+ whisper: !new:speechbrain.lobes.models.huggingface_whisper.HuggingFaceWhisper
44
+ source: !ref <whisper_hub>
45
+ freeze: !ref <freeze_whisper>
46
+ freeze_encoder: !ref <freeze_encoder>
47
+ save_path: whisper_checkpoints
48
+ encoder_only: False
49
+
50
+
51
 
52
  decoder: !new:speechbrain.decoders.seq2seq.S2SWhisperGreedySearch
53
+ model: !ref <whisper>
54
+ bos_index: !ref <timestamp_index>
55
+ eos_index: !ref <eos_index>
56
+ min_decode_ratio: !ref <min_decode_ratio>
57
+ max_decode_ratio: !ref <max_decode_ratio>
58
 
59
+ # test_beam_searcher: !new:speechbrain.decoders.seq2seq.S2SWhisperBeamSearch
60
+ # module: [!ref <whisper>]
61
+ # bos_index: !ref <timestamp_index>
62
+ # eos_index: !ref <eos_index>
63
+ # min_decode_ratio: !ref <min_decode_ratio>
64
+ # max_decode_ratio: !ref <max_decode_ratio>
65
+ # beam_size: !ref <test_beam_size>
66
+
67
+
68
+ modules:
69
  whisper: !ref <whisper>
70
+ decoder: !ref <decoder>
71
+
72
+
73
+ pretrainer: !new:speechbrain.utils.parameter_transfer.Pretrainer
74
+ loadables:
75
+ whisper: !ref <whisper>