Commit
·
d8378cd
1
Parent(s):
1689179
Fix authentication error and add setup instructions
Browse files- SETUP_INSTRUCTIONS.md +24 -0
- app.py +19 -10
SETUP_INSTRUCTIONS.md
ADDED
@@ -0,0 +1,24 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# Setup Instructions
|
2 |
+
|
3 |
+
## Required Steps:
|
4 |
+
|
5 |
+
1. **Accept the pyannote model conditions**:
|
6 |
+
- Visit https://huggingface.co/pyannote/speaker-diarization-3.1
|
7 |
+
- Click "Agree and access repository" to accept the conditions
|
8 |
+
|
9 |
+
2. **Create a Hugging Face token**:
|
10 |
+
- Go to https://huggingface.co/settings/tokens
|
11 |
+
- Create a new token with "read" permissions
|
12 |
+
- Copy the token
|
13 |
+
|
14 |
+
3. **Add the token to your Space secrets**:
|
15 |
+
- Go to your Space settings: https://huggingface.co/spaces/marcosremar2/speaker-diarization-pyannote/settings
|
16 |
+
- Scroll to "Repository secrets" section
|
17 |
+
- Add a new secret:
|
18 |
+
- Name: `HF_TOKEN`
|
19 |
+
- Value: Your Hugging Face token
|
20 |
+
- Click "Save"
|
21 |
+
|
22 |
+
4. **Restart the Space** after adding the token
|
23 |
+
|
24 |
+
The Space should now work correctly!
|
app.py
CHANGED
@@ -7,21 +7,30 @@ import os
|
|
7 |
|
8 |
# Login to Hugging Face if token is available
|
9 |
hf_token = os.environ.get("HF_TOKEN")
|
10 |
-
if hf_token:
|
11 |
-
|
12 |
|
13 |
-
|
14 |
-
pipeline = Pipeline.from_pretrained(
|
15 |
-
"pyannote/speaker-diarization-3.1",
|
16 |
-
use_auth_token=hf_token
|
17 |
-
)
|
18 |
|
19 |
-
#
|
20 |
-
|
21 |
-
pipeline.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
22 |
|
23 |
def diarize_audio(audio_file):
|
24 |
"""Process audio file and return diarization results"""
|
|
|
|
|
|
|
25 |
try:
|
26 |
# Apply pretrained pipeline
|
27 |
diarization = pipeline(audio_file)
|
|
|
7 |
|
8 |
# Login to Hugging Face if token is available
|
9 |
hf_token = os.environ.get("HF_TOKEN")
|
10 |
+
if not hf_token:
|
11 |
+
raise ValueError("HF_TOKEN environment variable is required. Please set it in the Space settings.")
|
12 |
|
13 |
+
login(token=hf_token)
|
|
|
|
|
|
|
|
|
14 |
|
15 |
+
# Initialize the pipeline
|
16 |
+
try:
|
17 |
+
pipeline = Pipeline.from_pretrained(
|
18 |
+
"pyannote/speaker-diarization-3.1",
|
19 |
+
use_auth_token=hf_token
|
20 |
+
)
|
21 |
+
|
22 |
+
# Send pipeline to GPU if available
|
23 |
+
if torch.cuda.is_available():
|
24 |
+
pipeline.to(torch.device("cuda"))
|
25 |
+
except Exception as e:
|
26 |
+
print(f"Error loading pipeline: {e}")
|
27 |
+
pipeline = None
|
28 |
|
29 |
def diarize_audio(audio_file):
|
30 |
"""Process audio file and return diarization results"""
|
31 |
+
if pipeline is None:
|
32 |
+
return "Pipeline not loaded. Please ensure HF_TOKEN is set and you have access to pyannote/speaker-diarization-3.1"
|
33 |
+
|
34 |
try:
|
35 |
# Apply pretrained pipeline
|
36 |
diarization = pipeline(audio_file)
|