whisper-medium-pt-3000h

fsicoli/whisper-medium-pt-3000h

whisper-medium-pt-3000h is a automatic speech recognition model from fsicoli. Use it when you need speech turned into text. It is set up for transformers. The card lists the license as apache-2.0.

should probably proofread and complete it, then remove this comment. --

Downloads · 30 days

9% of all-time downloads

All-time downloads

565

Public

Parameters

764M

364 GB on disk

Likes

Public

Hugging Face

Repo makeup

Click a slice to open those files.

.safetensors3.1 GB · 99%

At a glance

Task: Automatic Speech Recognition
Library: transformers
License: apache-2.0
Model type: whisper
Access: Public
Created: May 20, 2024
Updated: Jul 12, 2024
SHA: 673f28a4

Base models

openai/whisper-medium

Datasets

fsicoli/cv17-fleurs-coraa-mls-ted-alcaim-cf-cdc-lapsbm-lapsmail-sydney-lingualibre-voxforge-tatoeba

Model card

From the Hugging Face model README

whisper-medium-pt-3000h

This model is a fine-tuned version of openai/whisper-medium on the fsicoli/cv17-fleurs-coraa-mls-ted-alcaim-cf-cdc-lapsbm-lapsmail-sydney-lingualibre-voxforge-tatoeba default dataset. It achieves the following results on the evaluation set:

Loss: 0.9306
Wer: 0.1101

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 5e-06
train_batch_size: 8
eval_batch_size: 8
seed: 42
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: linear
lr_scheduler_warmup_steps: 10000
num_epochs: 10.0
mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.4423	0.2	20000	0.4723	0.1633
0.4963	0.39	40000	0.4921	0.1547
0.3853	0.59	60000	0.5099	0.1470
0.37	0.79	80000	0.4753	0.1439
0.3615	0.98	100000	0.5074	0.1386
0.2394	1.18	120000	0.4858	0.1341
0.227	1.38	140000	0.5758	0.1323
0.2461	1.57	160000	0.5067	0.1322
0.2078	1.77	180000	0.5087	0.1291
0.2138	1.97	200000	0.5201	0.1273
0.1188	2.16	220000	0.6359	0.1265
0.1009	2.36	240000	0.6229	0.1253
0.1394	2.56	260000	0.5734	0.1231
0.1383	2.75	280000	0.5914	0.1213
0.1332	2.95	300000	0.6174	0.1212
0.0634	3.15	320000	0.6461	0.1190
0.0667	3.34	340000	0.6330	0.1211
0.0546	3.54	360000	0.6927	0.1190
0.1029	3.74	380000	0.6777	0.1184
0.0664	3.93	400000	0.6367	0.1161
0.0665	4.13	420000	0.7467	0.1171
0.0695	4.33	440000	0.7332	0.1164
0.0708	4.52	460000	0.7141	0.1171
0.0695	4.72	480000	0.6869	0.1169
0.0758	4.92	500000	0.7360	0.1153
0.061	5.11	520000	0.7594	0.1161
0.0804	5.31	540000	0.7640	0.1158
0.0963	5.51	560000	0.7848	0.1157
0.0815	5.7	580000	0.7635	0.1145
0.0794	5.9	600000	0.7566	0.1134
0.0907	6.1	620000	0.8152	0.1147
0.0664	6.29	640000	0.8405	0.1123
0.0654	6.49	660000	0.8278	0.1119
0.0652	6.69	680000	0.8267	0.1134
0.1043	6.88	700000	0.8254	0.1122
0.0383	7.08	720000	0.8719	0.1122
0.0461	7.28	740000	0.8640	0.1130
0.0791	7.47	760000	0.8990	0.1122
0.0587	7.67	780000	0.9107	0.1122
0.0578	7.87	800000	0.9060	0.1124
0.0218	8.06	820000	0.8845	0.1111
0.0125	8.26	840000	0.9072	0.1112
0.0172	8.46	860000	0.8899	0.1107
0.0204	8.65	880000	0.9149	0.1108
0.0145	8.85	900000	0.9097	0.1103
0.0146	9.05	920000	0.9084	0.1107
0.0166	9.24	940000	0.9053	0.1103
0.0177	9.44	960000	0.9193	0.1100
0.0157	9.64	980000	0.9212	0.1101
0.0096	9.83	1000000	0.9313	0.1103

Framework versions

Transformers 4.39.0.dev0
Pytorch 2.2.1+cu121
Datasets 2.18.1.dev0
Tokenizers 0.15.0

Task: Automatic Speech Recognition
Library: transformers
Type: whisper
License: apache-2.0
Created: May 20, 2024
Updated: Jul 12, 2024