oron-tts / eval

1.13 MB

Ctrl+K

1 contributor

History: 148 commits

btsee

Both prompts on one sentence set, so the columns compare

5deb0c0 verified about 2 months ago

v4
oron-cleaner pipeline at filter policy v4 about 2 months ago
v4_voices
v4 about 2 months ago
common_eval.json

847 Bytes
Both prompts on one sentence set, so the columns compare about 2 months ago
common_eval.py

6.43 kB
Score both prompts on one common sentence set about 2 months ago
cv_eval_sentences.txt

46 kB
Preserve the text holdouts and the n=360/voice results about 2 months ago
cv_v4.sh

2.82 kB
Common Voice v4 clean + voice reselection about 2 months ago
eval_n100.json

21.3 kB
Preserve the text holdouts and the n=360/voice results about 2 months ago
female_choice.json

3.98 kB
v4 prompts: reference clips and the demos they produce about 2 months ago
female_confirm.py

7.93 kB
Confirm from recovered clips about 2 months ago
finalize_v4.sh

1.51 kB
Wire the clip recovery into the finalize run about 2 months ago
finish_v4.sh

3.4 kB
Pod self-reports to the Hub; the SSH proxy is unreliable as a control channel about 2 months ago
fleurs_eval_sentences.txt

50 kB
Preserve the text holdouts and the n=360/voice results about 2 months ago
mbspeech_eval_sentences.txt

30.8 kB
Preserve the text holdouts and the n=360/voice results about 2 months ago
merged_fleurs_eval_sentences.txt

69.9 kB
Preserve the text holdouts and the n=360/voice results about 2 months ago
merged_mbspeech_eval_sentences.txt

62.7 kB
Preserve the text holdouts and the n=360/voice results about 2 months ago
pod_eval.py

11.7 kB
Evaluation driver: n>=100, three seeds, UTMOS with bootstrap intervals about 2 months ago
prompt_eval.json

4.45 kB
prompt eval about 2 months ago
prompt_eval.py

9.52 kB
Rank prompt candidates from the manifest: write_bundle emits only one per gender about 2 months ago
push.py

3.09 kB
Upload voices.json, the file write_bundle actually emits about 2 months ago
read_results.py

2.37 kB
Compact reader for the n>=100 eval results about 2 months ago
rebuild.py

5.8 kB
Wire the clip recovery into the finalize run about 2 months ago
recover_clips.py

3.84 kB
Regenerate the decision's clips by id rather than re-cleaning 704 about 2 months ago
stage_cv_eval_sentences.txt

46 kB
Preserve the text holdouts and the n=360/voice results about 2 months ago
stage_fleurs_eval_sentences.txt

50 kB
Preserve the text holdouts and the n=360/voice results about 2 months ago
v4_bw_summary.json

198 Bytes
v4 run artifacts about 2 months ago
v4_run.sh

7.99 kB
PYTHONPATH for the inline checks about 2 months ago
v4_status.json

142 Bytes
v4 status: 704 clips kept about 2 months ago