dd3434 commited on
Commit
9f999c7
Β·
verified Β·
1 Parent(s): a05bc57

Replace author-specific absolute paths in configs with neutral placeholders

Browse files
alignment.yaml CHANGED
@@ -1,5 +1,5 @@
1
  # Generated 2026-02-02 from:
2
- # /gallery_rijks/jinyoung.park/streamalign/streamASR/hparams/train_chunk_streaming.yaml
3
  # yamllint disable
4
  # ############################################################################
5
  # Model: E2E ASR with Conformer Transducer (Character-level)
@@ -20,7 +20,7 @@ save_folder: results/conformer_transducer_char/3407/save
20
  train_log: results/conformer_transducer_char/3407/train_log.txt
21
 
22
  # Data files
23
- data_folder: /gallery_rijks/jinyoung.park/datasets/LibriTTS
24
 
25
  # Files to hold the manifest data
26
  train_json: !ref <save_folder>/train.json
 
1
  # Generated 2026-02-02 from:
2
+ # /home/streamalign/streamASR/hparams/train_chunk_streaming.yaml
3
  # yamllint disable
4
  # ############################################################################
5
  # Model: E2E ASR with Conformer Transducer (Character-level)
 
20
  train_log: results/conformer_transducer_char/3407/train_log.txt
21
 
22
  # Data files
23
+ data_folder: /home/datasets/LibriTTS
24
 
25
  # Files to hold the manifest data
26
  train_json: !ref <save_folder>/train.json
boundary_classifier/boundary_classifier.yaml CHANGED
@@ -25,14 +25,14 @@ __set_seed: !apply:speechbrain.utils.seed_everything [42]
25
 
26
  # ── ASR model (frozen feature extractor) ────────────────────────────────────
27
  # Path to the YAML used to train the streaming ASR model.
28
- asr_hparams_file: /gallery_rijks/jinyoung.park/streamalign/streamASR/hparams/chunk_streaming_word_fastemit.yaml
29
 
30
  # Path to a checkpoint directory (must contain model.ckpt and normalizer.ckpt).
31
- asr_checkpoint: /gallery_rijks/jinyoung.park/streamalign/streamASR/results/conformer_transducer_char/3419_fastemit/save/CKPT+2026-04-13+02-03-40+00
32
 
33
  # Path to a sentencepiece tokenizer checkpoint (tokenizer.ckpt).
34
  # If null the script searches <asr_checkpoint>/../../pretrained/tokenizer.ckpt.
35
- tokenizer_ckpt: /gallery_rijks/jinyoung.park/streamalign/streamASR/results/conformer_transducer_char/3419_fastemit/pretrained/tokenizer.ckpt
36
 
37
  # ── Streaming configuration (must match the ASR model's inference setting) ──
38
  chunk_size: 4 # encoder output frames per chunk (~40 ms / chunk)
@@ -40,7 +40,7 @@ left_context: 32 # left context in # of chunks
40
 
41
  # ── Dataset ──────────────────────────────────────────────────────────────────
42
  # LibriSpeech root directory (contains train-clean-100/, dev-clean/, …)
43
- data_folder: /gallery_rijks/jinyoung.park/datasets/LibriSpeech
44
 
45
  # Splits used for dataset creation.
46
  train_splits:
@@ -52,7 +52,7 @@ valid_split:
52
  - test-clean
53
 
54
  # Directory where the pre-extracted .pt dataset files are written / read from.
55
- dataset_dir: /gallery_rijks/jinyoung.park/streamalign/streamASR/data/boundary_dataset
56
 
57
  # ── BoundaryClassifier architecture ─────────────────────────────────────────
58
  joint_dim: 640 # must match the ASR model's joint_dim
@@ -74,6 +74,6 @@ num_workers: 4
74
  neg_weight: 3.0
75
 
76
  # ── Output ───────────────────────────────────────────────────────────────────
77
- output_folder: /gallery_rijks/jinyoung.park/streamalign/streamASR/train/results/boundary_classifier_0416
78
  checkpoint_folder: !ref <output_folder>/save
79
  train_log: !ref <output_folder>/train_log.txt
 
25
 
26
  # ── ASR model (frozen feature extractor) ────────────────────────────────────
27
  # Path to the YAML used to train the streaming ASR model.
28
+ asr_hparams_file: /home/streamalign/streamASR/hparams/chunk_streaming_word_fastemit.yaml
29
 
30
  # Path to a checkpoint directory (must contain model.ckpt and normalizer.ckpt).
31
+ asr_checkpoint: /home/streamalign/streamASR/results/conformer_transducer_char/3419_fastemit/save/CKPT+2026-04-13+02-03-40+00
32
 
33
  # Path to a sentencepiece tokenizer checkpoint (tokenizer.ckpt).
34
  # If null the script searches <asr_checkpoint>/../../pretrained/tokenizer.ckpt.
35
+ tokenizer_ckpt: /home/streamalign/streamASR/results/conformer_transducer_char/3419_fastemit/pretrained/tokenizer.ckpt
36
 
37
  # ── Streaming configuration (must match the ASR model's inference setting) ──
38
  chunk_size: 4 # encoder output frames per chunk (~40 ms / chunk)
 
40
 
41
  # ── Dataset ──────────────────────────────────────────────────────────────────
42
  # LibriSpeech root directory (contains train-clean-100/, dev-clean/, …)
43
+ data_folder: /home/datasets/LibriSpeech
44
 
45
  # Splits used for dataset creation.
46
  train_splits:
 
52
  - test-clean
53
 
54
  # Directory where the pre-extracted .pt dataset files are written / read from.
55
+ dataset_dir: /home/streamalign/streamASR/data/boundary_dataset
56
 
57
  # ── BoundaryClassifier architecture ─────────────────────────────────────────
58
  joint_dim: 640 # must match the ASR model's joint_dim
 
74
  neg_weight: 3.0
75
 
76
  # ── Output ───────────────────────────────────────────────────────────────────
77
+ output_folder: /home/streamalign/streamASR/train/results/boundary_classifier_0416
78
  checkpoint_folder: !ref <output_folder>/save
79
  train_log: !ref <output_folder>/train_log.txt
streaming_asr/chunk_streaming_word_fastemit.yaml CHANGED
@@ -21,9 +21,9 @@ pretrain_source: speechbrain/asr-streaming-conformer-librispeech
21
  pretrain_folder: results/conformer_transducer_char/3419_fastemit/pretrained
22
 
23
  # Data files
24
- data_folder: /gallery_rijks/jinyoung.park/datasets/LibriSpeech
25
- emilia_data_folder: /gallery_goya/kangwook.kim/datasets/Emilia-Dataset/Emilia/EN
26
- emilia_train_csv: /gallery_rijks/jinyoung.park/datasets/Emilia/emilia_en_400h.csv
27
 
28
  # CSV files (LibriSpeech, pre-existing from 3409 save)
29
  train_csv:
 
21
  pretrain_folder: results/conformer_transducer_char/3419_fastemit/pretrained
22
 
23
  # Data files
24
+ data_folder: /home/datasets/LibriSpeech
25
+ emilia_data_folder: /home/datasets/Emilia-Dataset/Emilia/EN
26
+ emilia_train_csv: /home/datasets/Emilia/emilia_en_400h.csv
27
 
28
  # CSV files (LibriSpeech, pre-existing from 3409 save)
29
  train_csv: