wav2vec2-large-xls-r-300m-urdu-v2

omar47/wav2vec2-large-xls-r-300m-urdu-v2

wav2vec2-large-xls-r-300m-urdu-v2 is a automatic speech recognition model from omar47. 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

2% of all-time downloads

All-time downloads

274

Public

Repo size

11.4 GB

Likes

Public

Hugging Face

Repo makeup

Click a slice to open those files.

.bin1.3 GB · 100%

At a glance

Task: Automatic Speech Recognition
Library: transformers
License: apache-2.0
Model type: wav2vec2
Access: Public
Created: May 7, 2022
Updated: May 14, 2022
SHA: e9346a4e

Model card

From the Hugging Face model README

wav2vec2-large-xls-r-300m-urdu-CV_8_0-and-PRUS_v2

This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the None dataset. It achieves the following results on the evaluation set:

Loss: 1.3541
Wer: 0.6532

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: 0.0003
train_batch_size: 16
eval_batch_size: 8
seed: 42
gradient_accumulation_steps: 2
total_train_batch_size: 32
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: linear
lr_scheduler_warmup_steps: 500
num_epochs: 25
mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
14.8521	0.52	32	20.0617	1.0
9.2152	1.05	64	7.8943	1.0
4.8598	1.57	96	5.1558	1.0
3.866	2.1	128	3.9680	1.0
3.3517	2.62	160	3.4201	1.0
3.2029	3.15	192	3.2355	1.0
3.1509	3.67	224	3.2337	1.0
3.1399	4.2	256	3.1627	1.0
3.0848	4.72	288	3.0550	1.0
2.9806	5.25	320	2.8343	0.9996
2.3814	5.77	352	2.0685	0.9523
1.2936	6.3	384	1.5907	0.8657
0.8656	6.82	416	1.3810	0.8235
0.7014	7.34	448	1.3838	0.7920
0.6015	7.87	480	1.3479	0.8046
0.5341	8.39	512	1.2613	0.7757
0.5031	8.92	544	1.2818	0.7890
0.4349	9.44	576	1.3171	0.7739
0.4198	9.97	608	1.2420	0.7750
0.3593	10.49	640	1.2991	0.7587
0.3252	11.02	672	1.2653	0.7228
0.2715	11.54	704	1.2488	0.7350
0.2733	12.07	736	1.2639	0.7110
0.2338	12.59	768	1.3733	0.7454
0.2403	13.11	800	1.3908	0.7228
0.2106	13.64	832	1.3384	0.7224
0.2041	14.16	864	1.3770	0.7050
0.1814	14.69	896	1.3526	0.6932
0.1742	15.21	928	1.3486	0.6895
0.1658	15.74	960	1.3210	0.6936
0.1455	16.26	992	1.3292	0.6858
0.1399	16.79	1024	1.3521	0.6828
0.1325	17.31	1056	1.3339	0.6876
0.1256	17.84	1088	1.3389	0.6836
0.1219	18.36	1120	1.3496	0.6769
0.1212	18.89	1152	1.3277	0.6776
0.1097	19.41	1184	1.3594	0.6762
0.1129	19.93	1216	1.3448	0.6688
0.1036	20.46	1248	1.3295	0.6710
0.1035	20.98	1280	1.3243	0.6577
0.094	21.51	1312	1.3832	0.6591
0.0912	22.03	1344	1.3857	0.6584
0.0815	22.56	1376	1.3739	0.6547
0.0864	23.08	1408	1.3649	0.6554
0.0772	23.61	1440	1.3791	0.6458
0.0894	24.13	1472	1.3630	0.6488
0.0776	24.66	1504	1.3541	0.6532

Framework versions

Transformers 4.19.1
Pytorch 1.11.0+cu113
Datasets 2.2.1
Tokenizers 0.12.1

Task: Automatic Speech Recognition
Library: transformers
Type: wav2vec2
License: apache-2.0
Created: May 7, 2022
Updated: May 14, 2022