Spaces:
Runtime error
Runtime error
Update run.py
Browse files
run.py
CHANGED
@@ -9,7 +9,7 @@ from huggingface_hub import hf_hub_download, snapshot_download
|
|
9 |
import pyarrow.parquet as pq
|
10 |
import hashlib
|
11 |
import os
|
12 |
-
|
13 |
|
14 |
def load_hf_dataset(dataset_path, auth_token):
|
15 |
dataset = load_dataset(dataset_path, token=auth_token)
|
@@ -18,47 +18,53 @@ def load_hf_dataset(dataset_path, auth_token):
|
|
18 |
|
19 |
def fast_caption(sys_prompt, usr_prompt, temp, top_p, max_tokens, model, key, endpoint, video_src, video_hf, video_hf_auth, parquet_index, video_od, video_od_auth, video_gd, video_gd_auth, frame_format, frame_limit):
|
20 |
progress_info = []
|
21 |
-
|
22 |
-
|
23 |
-
|
24 |
-
|
25 |
-
|
26 |
-
|
27 |
-
if
|
28 |
-
|
29 |
-
|
30 |
-
|
31 |
-
|
32 |
-
|
33 |
-
|
34 |
-
|
35 |
-
|
36 |
-
|
37 |
-
|
38 |
-
|
39 |
-
|
40 |
-
|
41 |
-
|
42 |
-
|
43 |
-
|
44 |
-
|
45 |
-
|
46 |
-
|
47 |
-
|
48 |
-
|
49 |
-
|
50 |
-
|
51 |
-
|
52 |
-
|
53 |
-
|
54 |
-
|
55 |
-
|
56 |
-
|
57 |
-
|
58 |
-
|
59 |
-
|
60 |
-
|
61 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
62 |
|
63 |
with gr.Blocks() as Core:
|
64 |
with gr.Row(variant="panel"):
|
@@ -124,12 +130,12 @@ with gr.Blocks() as Core:
|
|
124 |
video_gd = gr.Text()
|
125 |
video_gd_auth = gr.Text(label="Google Drive Access Token")
|
126 |
caption_button = gr.Button("Caption", variant="primary", size="lg")
|
|
|
127 |
caption_button.click(
|
128 |
fast_caption,
|
129 |
inputs=[sys_prompt, usr_prompt, temp, top_p, max_tokens, model, key, endpoint, video_src, video_hf, video_hf_auth, parquet_index, video_od, video_od_auth, video_gd, video_gd_auth, frame_format, frame_limit],
|
130 |
-
outputs=[
|
131 |
)
|
132 |
|
133 |
if __name__ == "__main__":
|
134 |
-
Core.launch()
|
135 |
-
|
|
|
9 |
import pyarrow.parquet as pq
|
10 |
import hashlib
|
11 |
import os
|
12 |
+
import csv
|
13 |
|
14 |
def load_hf_dataset(dataset_path, auth_token):
|
15 |
dataset = load_dataset(dataset_path, token=auth_token)
|
|
|
18 |
|
19 |
def fast_caption(sys_prompt, usr_prompt, temp, top_p, max_tokens, model, key, endpoint, video_src, video_hf, video_hf_auth, parquet_index, video_od, video_od_auth, video_gd, video_gd_auth, frame_format, frame_limit):
|
20 |
progress_info = []
|
21 |
+
csv_filename = "captions.csv"
|
22 |
+
with open(csv_filename, mode='w', newline='') as csv_file:
|
23 |
+
fieldnames = ['md5', 'caption']
|
24 |
+
writer = csv.DictWriter(csv_file, fieldnames=fieldnames)
|
25 |
+
writer.writeheader()
|
26 |
+
|
27 |
+
if video_src:
|
28 |
+
video = video_src
|
29 |
+
processor = VideoProcessor(frame_format=frame_format, frame_limit=frame_limit)
|
30 |
+
frames = processor._decode(video)
|
31 |
+
base64_list = processor.to_base64_list(frames)
|
32 |
+
debug_image = processor.concatenate(frames)
|
33 |
+
if not key or not endpoint:
|
34 |
+
return "", f"API key or endpoint is missing. Processed {len(frames)} frames.", debug_image
|
35 |
+
api = AzureAPI(key=key, endpoint=endpoint, model=model, temp=temp, top_p=top_p, max_tokens=max_tokens)
|
36 |
+
caption = api.get_caption(sys_prompt, usr_prompt, base64_list)
|
37 |
+
progress_info.append(f"Using model '{model}' with {len(frames)} frames extracted.")
|
38 |
+
writer.writerow({'md5': 'single_video', 'caption': caption})
|
39 |
+
return f"{caption}", "\n".join(progress_info), debug_image
|
40 |
+
elif video_hf and video_hf_auth:
|
41 |
+
current_file_path = os.path.abspath(__file__)
|
42 |
+
current_directory = os.path.dirname(current_file_path)
|
43 |
+
progress_info.append('Begin processing Hugging Face dataset.')
|
44 |
+
temp_parquet_file = hf_hub_download(
|
45 |
+
repo_id=video_hf,
|
46 |
+
filename='data/' + str(parquet_index).zfill(6) + '.parquet',
|
47 |
+
repo_type="dataset",
|
48 |
+
token=video_hf_auth,
|
49 |
+
)
|
50 |
+
parquet_file = pq.ParquetFile(temp_parquet_file)
|
51 |
+
for batch in parquet_file.iter_batches(batch_size=1):
|
52 |
+
df = batch.to_pandas()
|
53 |
+
video = df['video'][0]
|
54 |
+
md5 = hashlib.md5(video).hexdigest()
|
55 |
+
with tempfile.NamedTemporaryFile(dir=current_directory) as temp_file:
|
56 |
+
temp_file.write(video)
|
57 |
+
video_path = temp_file.name
|
58 |
+
processor = VideoProcessor(frame_format=frame_format, frame_limit=frame_limit)
|
59 |
+
frames = processor._decode(video_path)
|
60 |
+
base64_list = processor.to_base64_list(frames)
|
61 |
+
api = AzureAPI(key=key, endpoint=endpoint, model=model, temp=temp, top_p=top_p, max_tokens=max_tokens)
|
62 |
+
caption = api.get_caption(sys_prompt, usr_prompt, base64_list)
|
63 |
+
writer.writerow({'md5': md5, 'caption': caption})
|
64 |
+
progress_info.append(f"Processed video with MD5: {md5}")
|
65 |
+
return csv_filename, "\n".join(progress_info), None
|
66 |
+
else:
|
67 |
+
return "", "No video source selected.", None
|
68 |
|
69 |
with gr.Blocks() as Core:
|
70 |
with gr.Row(variant="panel"):
|
|
|
130 |
video_gd = gr.Text()
|
131 |
video_gd_auth = gr.Text(label="Google Drive Access Token")
|
132 |
caption_button = gr.Button("Caption", variant="primary", size="lg")
|
133 |
+
csv_link = gr.File(label="Download CSV", interactive=False)
|
134 |
caption_button.click(
|
135 |
fast_caption,
|
136 |
inputs=[sys_prompt, usr_prompt, temp, top_p, max_tokens, model, key, endpoint, video_src, video_hf, video_hf_auth, parquet_index, video_od, video_od_auth, video_gd, video_gd_auth, frame_format, frame_limit],
|
137 |
+
outputs=[csv_link, info, frame]
|
138 |
)
|
139 |
|
140 |
if __name__ == "__main__":
|
141 |
+
Core.launch()
|
|