lataon tecasoftai commited on
Commit
e2f676e
·
verified ·
1 Parent(s): eb158eb

Upload 12 files (#2)

Browse files

- Upload 12 files (cf226abfa629bc244163ae994192e49b8751d268)


Co-authored-by: TecaSoftAndTech <tecasoftai@users.noreply.huggingface.co>

eval-results/phonemes_1760931179_HuBERT fine-tuned.json ADDED
The diff for this file is too large to render. See raw diff
 
eval-results/phonemes_1760931179_HuBERT-Base.json ADDED
The diff for this file is too large to render. See raw diff
 
eval-results/phonemes_1760931179_LJSpeech Gruut.json ADDED
The diff for this file is too large to render. See raw diff
 
eval-results/phonemes_1760931179_Timit.json ADDED
The diff for this file is too large to render. See raw diff
 
eval-results/phonemes_1760931179_WavLM.json ADDED
The diff for this file is too large to render. See raw diff
 
eval-results/phonemes_1760931179_Whisper.json ADDED
The diff for this file is too large to render. See raw diff
 
eval-results/results_1760931179_HuBERT-Base.json ADDED
@@ -0,0 +1,29 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "config": {
3
+ "model_name": "HuBERT-Base",
4
+ "model_dtype": "float32",
5
+ "model_sha": ""
6
+ },
7
+ "results": {
8
+ "phoneme_asr": {
9
+ "per": 1.547547540586229,
10
+ "avg_duration": 0.29178321838378907
11
+ },
12
+ "kids_phoneme_md": {
13
+ "per": 19.658098265801193,
14
+ "avg_duration": 0.4278303599357605
15
+ },
16
+ "timit_asr_ipa": {
17
+ "per": 1.7191063939432303,
18
+ "avg_duration": 0.23941755533218384
19
+ },
20
+ "librispeech_asr": {
21
+ "per": 0.5881798687155991,
22
+ "avg_duration": 0.3870742917060852
23
+ },
24
+ "MultiMed": {
25
+ "per": 9.588829127767955,
26
+ "avg_duration": 0.47454899072647094
27
+ }
28
+ }
29
+ }
eval-results/results_1760931179_HuBERT-fine-tuned.json ADDED
@@ -0,0 +1,29 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "config": {
3
+ "model_name": "HuBERT-fine-tuned",
4
+ "model_dtype": "float32",
5
+ "model_sha": ""
6
+ },
7
+ "results": {
8
+ "phoneme_asr": {
9
+ "per": 0.7807717265999417,
10
+ "avg_duration": 0.3225450205802918
11
+ },
12
+ "kids_phoneme_md": {
13
+ "per": 20.704365073778412,
14
+ "avg_duration": 0.5801742506027222
15
+ },
16
+ "timit_asr_ipa": {
17
+ "per": 0.7514373562966179,
18
+ "avg_duration": 0.29422701358795167
19
+ },
20
+ "librispeech_asr": {
21
+ "per": 1.5705384937773277,
22
+ "avg_duration": 0.5730020213127136
23
+ },
24
+ "MultiMed": {
25
+ "per": 9.220691447394458,
26
+ "avg_duration": 0.7558647537231445
27
+ }
28
+ }
29
+ }
eval-results/results_1760931179_LJSpeech-Gruut.json ADDED
@@ -0,0 +1,29 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "config": {
3
+ "model_name": "LJSpeech-Gruut",
4
+ "model_dtype": "float32",
5
+ "model_sha": ""
6
+ },
7
+ "results": {
8
+ "phoneme_asr": {
9
+ "per": 27.27165545359053,
10
+ "avg_duration": 0.17962724924087525
11
+ },
12
+ "kids_phoneme_md": {
13
+ "per": 62.05639578486782,
14
+ "avg_duration": 0.25178531169891355
15
+ },
16
+ "timit_asr_ipa": {
17
+ "per": 26.868515473361054,
18
+ "avg_duration": 0.14754701375961304
19
+ },
20
+ "librispeech_asr": {
21
+ "per": 20.67960537404926,
22
+ "avg_duration": 0.23313429832458496
23
+ },
24
+ "MultiMed": {
25
+ "per": 31.53710463881287,
26
+ "avg_duration": 0.3271867084503174
27
+ }
28
+ }
29
+ }
eval-results/results_1760931179_Timit.json ADDED
@@ -0,0 +1,29 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "config": {
3
+ "model_name": "Timit",
4
+ "model_dtype": "float32",
5
+ "model_sha": ""
6
+ },
7
+ "results": {
8
+ "phoneme_asr": {
9
+ "per": 31.839524020150396,
10
+ "avg_duration": 0.32524588346481326
11
+ },
12
+ "kids_phoneme_md": {
13
+ "per": 48.171072510363295,
14
+ "avg_duration": 0.5857418179512024
15
+ },
16
+ "timit_asr_ipa": {
17
+ "per": 31.957473427990195,
18
+ "avg_duration": 0.2946950888633728
19
+ },
20
+ "librispeech_asr": {
21
+ "per": 29.537610471893803,
22
+ "avg_duration": 0.5887207913398743
23
+ },
24
+ "MultiMed": {
25
+ "per": 37.45253395374299,
26
+ "avg_duration": 0.7519637131690979
27
+ }
28
+ }
29
+ }
eval-results/results_1760931179_WavLM.json ADDED
@@ -0,0 +1,29 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "config": {
3
+ "model_name": "WavLM",
4
+ "model_dtype": "float32",
5
+ "model_sha": ""
6
+ },
7
+ "results": {
8
+ "phoneme_asr": {
9
+ "per": 24.939372174413087,
10
+ "avg_duration": 0.32624215841293336
11
+ },
12
+ "kids_phoneme_md": {
13
+ "per": 64.76552694432064,
14
+ "avg_duration": 0.5985251760482788
15
+ },
16
+ "timit_asr_ipa": {
17
+ "per": 23.20600674022584,
18
+ "avg_duration": 0.2968974304199219
19
+ },
20
+ "librispeech_asr": {
21
+ "per": 32.58195540587739,
22
+ "avg_duration": 0.598731873035431
23
+ },
24
+ "MultiMed": {
25
+ "per": 45.96974612462279,
26
+ "avg_duration": 0.7653125381469726
27
+ }
28
+ }
29
+ }
eval-results/results_1760931179_Whisper.json ADDED
@@ -0,0 +1,29 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "config": {
3
+ "model_name": "Whisper",
4
+ "model_dtype": "float32",
5
+ "model_sha": ""
6
+ },
7
+ "results": {
8
+ "phoneme_asr": {
9
+ "per": 2.7299342309875616,
10
+ "avg_duration": 0.5561290311813355
11
+ },
12
+ "kids_phoneme_md": {
13
+ "per": 24.32141779355694,
14
+ "avg_duration": 0.958488187789917
15
+ },
16
+ "timit_asr_ipa": {
17
+ "per": 2.6108205875808976,
18
+ "avg_duration": 0.43648485660552977
19
+ },
20
+ "librispeech_asr": {
21
+ "per": 1.7348402193405519,
22
+ "avg_duration": 0.9080598521232605
23
+ },
24
+ "MultiMed": {
25
+ "per": 9.315595415190701,
26
+ "avg_duration": 1.1237735629081727
27
+ }
28
+ }
29
+ }