Spaces:
Sleeping
Sleeping
update app.py
Browse files
app.py
CHANGED
@@ -18,6 +18,14 @@ from transformers.pipelines.audio_utils import ffmpeg_read
|
|
18 |
import torch
|
19 |
from datasets import load_dataset, Dataset, DatasetDict
|
20 |
import spaces
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
21 |
|
22 |
# Constants
|
23 |
MODEL_NAME = "openai/whisper-large-v3-turbo"
|
@@ -56,7 +64,7 @@ pipe = pipeline(
|
|
56 |
model=model,
|
57 |
tokenizer=tokenizer,
|
58 |
feature_extractor=feature_extractor,
|
59 |
-
chunk_length_s=
|
60 |
)
|
61 |
|
62 |
def reset_and_update_dataset(new_data):
|
|
|
18 |
import torch
|
19 |
from datasets import load_dataset, Dataset, DatasetDict
|
20 |
import spaces
|
21 |
+
import subprocess
|
22 |
+
|
23 |
+
# Install Flash Attention if possible
|
24 |
+
subprocess.run(
|
25 |
+
"pip install flash-attn --no-build-isolation",
|
26 |
+
env={"FLASH_ATTENTION_SKIP_CUDA_BUILD": "TRUE"},
|
27 |
+
shell=True,
|
28 |
+
)
|
29 |
|
30 |
# Constants
|
31 |
MODEL_NAME = "openai/whisper-large-v3-turbo"
|
|
|
64 |
model=model,
|
65 |
tokenizer=tokenizer,
|
66 |
feature_extractor=feature_extractor,
|
67 |
+
chunk_length_s=900, # Increased to 15 minutes
|
68 |
)
|
69 |
|
70 |
def reset_and_update_dataset(new_data):
|