metadata

language:
  - he
license: apache-2.0
base_model: openai/whisper-medium
tags:
  - hf-asr-leaderboard
  - generated_from_trainer
metrics:
  - wer
model-index:
  - name: he
    results: []

he

This model is a fine-tuned version of openai/whisper-medium on an unknown dataset. It achieves the following results on the evaluation set:

Loss: 0.1756
Wer: 20.1811
Avg Precision Exact: 0.8083
Avg Recall Exact: 0.8102
Avg F1 Exact: 0.8087
Avg Precision Letter Shift: 0.8373
Avg Recall Letter Shift: 0.8394
Avg F1 Letter Shift: 0.8377
Avg Precision Word Level: 0.8427
Avg Recall Word Level: 0.8450
Avg F1 Word Level: 0.8432
Avg Precision Word Shift: 0.9448
Avg Recall Word Shift: 0.9489
Avg F1 Word Shift: 0.9460
Precision Median Exact: 0.9091
Recall Median Exact: 0.9091
F1 Median Exact: 0.9091
Precision Max Exact: 1.0
Recall Max Exact: 1.0
F1 Max Exact: 1.0
Precision Min Exact: 0.0
Recall Min Exact: 0.0
F1 Min Exact: 0.0
Precision Min Letter Shift: 0.0
Recall Min Letter Shift: 0.0
F1 Min Letter Shift: 0.0
Precision Min Word Level: 0.0
Recall Min Word Level: 0.0
F1 Min Word Level: 0.0
Precision Min Word Shift: 0.1429
Recall Min Word Shift: 0.1
F1 Min Word Shift: 0.1176

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: 1e-05
train_batch_size: 8
eval_batch_size: 32
seed: 42
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: linear
lr_scheduler_warmup_steps: 100
training_steps: 40000
mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Avg Precision Exact	Avg Recall Exact	Avg F1 Exact	Avg Precision Letter Shift	Avg Recall Letter Shift	Avg F1 Letter Shift	Avg Precision Word Level	Avg Recall Word Level	Avg F1 Word Level	Avg Precision Word Shift	Avg Recall Word Shift	Avg F1 Word Shift	Precision Median Exact	Recall Median Exact	F1 Median Exact	Precision Max Exact	Recall Max Exact	F1 Max Exact	Precision Min Word Shift	Recall Min Word Shift	F1 Min Word Shift
0.2873	0.16	2000	0.3088	44.3311	0.5598	0.5685	0.5633	0.6026	0.6118	0.6062	0.6138	0.6239	0.6178	0.8019	0.8196	0.8092	0.6154	0.625	0.6207	1.0	1.0	1.0	0.0	0.0	0.0
0.1752	0.32	4000	0.2328	35.1811	0.6557	0.6595	0.6568	0.6946	0.6985	0.6957	0.7041	0.7082	0.7053	0.8676	0.8745	0.8698	0.75	0.75	0.75	1.0	1.0	1.0	0.0	0.0	0.0
0.117	0.48	6000	0.1997	29.4605	0.7124	0.7125	0.7117	0.7514	0.7513	0.7506	0.7604	0.7606	0.7597	0.9031	0.9063	0.9037	0.8182	0.8182	0.8148	1.0	1.0	1.0	0.1111	0.1	0.1053
0.0994	0.64	8000	0.1881	27.5610	0.7359	0.7407	0.7376	0.7708	0.7758	0.7726	0.7783	0.7837	0.7803	0.9117	0.9191	0.9144	0.8333	0.8462	0.8387	1.0	1.0	1.0	0.1429	0.0909	0.1111
0.0664	0.8	10000	0.1837	25.9682	0.7446	0.7529	0.7480	0.7785	0.7873	0.7821	0.7857	0.7944	0.7893	0.9194	0.9277	0.9226	0.8462	0.8571	0.8571	1.0	1.0	1.0	0.1111	0.0909	0.1111
0.0767	0.96	12000	0.1760	24.6157	0.7561	0.7662	0.7604	0.7868	0.7973	0.7913	0.7936	0.8040	0.7980	0.9194	0.9315	0.9245	0.8667	0.875	0.8723	1.0	1.0	1.0	0.0	0.0	0.0
0.0593	1.12	14000	0.1716	23.6511	0.7669	0.7732	0.7694	0.7988	0.8054	0.8014	0.8047	0.8114	0.8073	0.9275	0.9343	0.9300	0.875	0.8889	0.8800	1.0	1.0	1.0	0.1111	0.1111	0.1111
0.0525	1.28	16000	0.1712	23.1264	0.7778	0.7788	0.7777	0.8092	0.8104	0.8092	0.8156	0.8169	0.8156	0.9345	0.9383	0.9356	0.8889	0.8889	0.8889	1.0	1.0	1.0	0.1111	0.1111	0.125
0.0358	1.44	18000	0.1699	22.4538	0.7841	0.7845	0.7837	0.8150	0.8157	0.8147	0.8212	0.8222	0.8211	0.9344	0.9376	0.9351	0.8889	0.8889	0.8889	1.0	1.0	1.0	0.1	0.1	0.1053
0.0323	1.6	20000	0.1713	22.2173	0.7873	0.7926	0.7893	0.8170	0.8224	0.8190	0.8230	0.8286	0.8251	0.9362	0.9424	0.9385	0.9	0.9	0.9	1.0	1.0	1.0	0.1111	0.1	0.1111
0.0248	1.76	22000	0.1683	21.7480	0.7934	0.7945	0.7933	0.8235	0.8248	0.8235	0.8294	0.8310	0.8295	0.9407	0.9433	0.9412	0.9	0.9	0.9	1.0	1.0	1.0	0.1429	0.1	0.1176
0.0209	1.92	24000	0.1696	21.0310	0.7982	0.8000	0.7986	0.8275	0.8292	0.8278	0.8331	0.8351	0.8335	0.9424	0.9461	0.9435	0.9	0.9	0.9	1.0	1.0	1.0	0.1429	0.1	0.1176
0.0202	2.08	26000	0.1713	21.1936	0.7954	0.7988	0.7965	0.8250	0.8285	0.8262	0.8309	0.8346	0.8321	0.9404	0.9460	0.9424	0.9	0.9091	0.9	1.0	1.0	1.0	0.0769	0.0909	0.0833
0.0172	2.24	28000	0.1716	20.7761	0.8013	0.8053	0.8027	0.8304	0.8346	0.8319	0.8359	0.8407	0.8376	0.9404	0.9469	0.9428	0.9091	0.9091	0.9091	1.0	1.0	1.0	0.2	0.1818	0.2000
0.0161	2.4	30000	0.1740	20.6135	0.8052	0.8079	0.8059	0.8351	0.8380	0.8359	0.8408	0.8440	0.8417	0.9439	0.9494	0.9459	0.9091	0.9091	0.9091	1.0	1.0	1.0	0.125	0.1	0.1176
0.01	2.56	32000	0.1743	20.6948	0.8031	0.8048	0.8033	0.8322	0.8339	0.8323	0.8380	0.8399	0.8382	0.9441	0.9480	0.9452	0.9091	0.9091	0.9062	1.0	1.0	1.0	0.1429	0.1	0.1176
0.025	2.72	34000	0.1753	20.6282	0.8033	0.8072	0.8046	0.8327	0.8368	0.8341	0.8383	0.8430	0.8400	0.9419	0.9489	0.9446	0.9091	0.9091	0.9091	1.0	1.0	1.0	0.1429	0.1	0.1176
0.0082	2.88	36000	0.1756	20.3991	0.8060	0.8081	0.8064	0.8354	0.8378	0.8359	0.8406	0.8433	0.8412	0.9436	0.9484	0.9452	0.9091	0.9091	0.9091	1.0	1.0	1.0	0.1429	0.1	0.1176
0.013	3.04	38000	0.1754	20.3030	0.8078	0.8097	0.8082	0.8374	0.8395	0.8378	0.8427	0.8452	0.8433	0.9447	0.9488	0.9459	0.9091	0.9091	0.9091	1.0	1.0	1.0	0.1429	0.1	0.1176
0.0183	3.2	40000	0.1756	20.1811	0.8083	0.8102	0.8087	0.8373	0.8394	0.8377	0.8427	0.8450	0.8432	0.9448	0.9489	0.9460	0.9091	0.9091	0.9091	1.0	1.0	1.0	0.1429	0.1	0.1176

Framework versions

Transformers 4.39.0.dev0
Pytorch 2.2.1+cu121
Datasets 2.16.1
Tokenizers 0.15.0