Training Loss	Epoch	Step	Validation Loss	Wer
0.1908	0.03	1000	0.2235	0.1154
0.1888	0.07	2000	0.2132	0.1131
0.167	0.1	3000	0.2115	0.1133
0.1752	0.14	4000	0.2081	0.1146
0.1656	0.17	5000	0.2002	0.1073
0.1535	0.21	6000	0.1971	0.1086
0.1854	0.24	7000	0.1927	0.1048
0.1722	0.28	8000	0.1889	0.1043
0.166	0.31	9000	0.1850	0.1022
0.1277	0.35	10000	0.1820	0.1032
0.1457	0.38	11000	0.1777	0.0998
0.169	0.42	12000	0.1771	0.0982
0.1612	0.45	13000	0.1724	0.0976
0.1616	0.49	14000	0.1693	0.0956
0.1556	0.52	15000	0.1671	0.0942
0.1448	0.56	16000	0.1646	0.0930
0.117	0.59	17000	0.1613	0.0914
0.1441	0.62	18000	0.1596	0.0899
0.148	0.66	19000	0.1571	0.0895
0.1255	0.69	20000	0.1547	0.0874
0.1479	0.73	21000	0.1525	0.0885
0.1304	0.76	22000	0.1503	0.0861
0.1111	0.8	23000	0.1486	0.0867
0.1337	0.83	24000	0.1472	0.0854
0.1289	0.87	25000	0.1466	0.0855

Transcription:

python

from datasets import load_dataset, Audio
import torch
from transformers import WhisperProcessor, WhisperForConditionalGeneration

# device
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")

# load the model
processor = WhisperProcessor.from_pretrained("clu-ling/whisper-large-v2-spanish")
model = WhisperForConditionalGeneration.from_pretrained("clu-ling/whisper-large-v2-spanish").to(device)
forced_decoder_ids = processor.get_decoder_prompt_ids(language="es", task="transcribe")

# load the dataset
commonvoice_eval = load_dataset("mozilla-foundation/common_voice_11_0", "es", split="validation", streaming=True)
commonvoice_eval = commonvoice_eval.cast_column("audio", Audio(sampling_rate=16000))
sample = next(iter(commonvoice_eval))["audio"]

# features and generate token ids
input_features = processor(sample["array"], sampling_rate=sample["sampling_rate"], return_tensors="pt").input_features
predicted_ids = model.generate(input_features.to(device), forced_decoder_ids=forced_decoder_ids)

# decode
transcription = processor.batch_decode(predicted_ids, skip_special_tokens=True)

print(transcription)

Evaluation:

Evaluates this model on mozilla-foundation/common_voice_11_0 test split.

python

from transformers.models.whisper.english_normalizer import BasicTextNormalizer
from datasets import load_dataset, Audio
import evaluate
import torch
import re
from transformers import WhisperProcessor, WhisperForConditionalGeneration

# device
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")

# metric
wer_metric = evaluate.load("wer")

# model
processor = WhisperProcessor.from_pretrained("clu-ling/whisper-large-v2-spanish")
model = WhisperForConditionalGeneration.from_pretrained("clu-ling/whisper-large-v2-spanish")

# dataset
dataset = load_dataset("mozilla-foundation/common_voice_11_0", "es", split="test", )#cache_dir=args.cache_dir
dataset = dataset.cast_column("audio", Audio(sampling_rate=16000))

#for debuggings: it gets some examples
#dataset = dataset.shard(num_shards=10000, index=0)
#print(dataset)
   
def normalize(batch):
  batch["gold_text"] = whisper_norm(batch['sentence'])
  return batch

def map_wer(batch):
  model.to(device)
  forced_decoder_ids = processor.get_decoder_prompt_ids(language = "es", task = "transcribe")
  inputs = processor(batch["audio"]["array"], sampling_rate=batch["audio"]["sampling_rate"], return_tensors="pt").input_features
  with torch.no_grad():
    generated_ids = model.generate(inputs=inputs.to(device), forced_decoder_ids=forced_decoder_ids)
    transcription = processor.batch_decode(generated_ids, skip_special_tokens=True)[0]
  batch["predicted_text"] = whisper_norm(transcription)
  return batch

# process GOLD text
processed_dataset = dataset.map(normalize)
# get predictions
predicted = processed_dataset.map(map_wer)

# word error rate
wer = wer_metric.compute(references=predicted['gold_text'], predictions=predicted['predicted_text'])
wer = round(100 * wer, 2)
print("WER:", wer)

Framework versions

Transformers 4.26.0.dev0
Pytorch 1.13.1+cu117
Datasets 2.8.1.dev0
Tokenizers 0.13.2