Spaces:
Runtime error
Runtime error
Update app.py
Browse files
app.py
CHANGED
@@ -10,6 +10,13 @@ MODEL_PATH = "THUDM/cogvlm2-llama3-caption"
|
|
10 |
DEVICE = 'cuda' if torch.cuda.is_available() else 'cpu'
|
11 |
TORCH_TYPE = torch.bfloat16 if torch.cuda.is_available() and torch.cuda.get_device_capability()[0] >= 8 else torch.float16
|
12 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
13 |
# Delay Reasons for Each Manufacturing Step
|
14 |
DELAY_REASONS = {
|
15 |
"Step 1": ["Delay in Bead Insertion", "Lack of raw material"],
|
@@ -100,15 +107,14 @@ def get_step_info(step_number):
|
|
100 |
def load_video(video_data, strategy='chat'):
|
101 |
"""Loads and processes video data into a format suitable for model input."""
|
102 |
bridge.set_bridge('torch')
|
103 |
-
|
104 |
-
|
105 |
if isinstance(video_data, str):
|
106 |
decord_vr = VideoReader(video_data, ctx=cpu(0))
|
107 |
else:
|
108 |
decord_vr = VideoReader(io.BytesIO(video_data), ctx=cpu(0))
|
109 |
|
110 |
total_frames = len(decord_vr)
|
111 |
-
if total_frames <
|
112 |
raise ValueError("Uploaded video is too short for meaningful analysis.")
|
113 |
|
114 |
timestamps = [i[0] for i in decord_vr.get_frame_timestamp(np.arange(total_frames))]
|
@@ -119,7 +125,7 @@ def load_video(video_data, strategy='chat'):
|
|
119 |
closest_num = min(timestamps, key=lambda x: abs(x - second))
|
120 |
index = timestamps.index(closest_num)
|
121 |
frame_id_list.append(index)
|
122 |
-
if len(frame_id_list) >=
|
123 |
break
|
124 |
|
125 |
video_data = decord_vr.get_batch(frame_id_list)
|
@@ -148,7 +154,10 @@ def load_model():
|
|
148 |
|
149 |
def predict(prompt, video_data, temperature, model, tokenizer):
|
150 |
"""Generates predictions based on the video and textual prompt."""
|
151 |
-
|
|
|
|
|
|
|
152 |
|
153 |
inputs = model.build_conversation_input_ids(
|
154 |
tokenizer=tokenizer,
|
@@ -166,12 +175,12 @@ def predict(prompt, video_data, temperature, model, tokenizer):
|
|
166 |
}
|
167 |
|
168 |
gen_kwargs = {
|
169 |
-
"max_new_tokens":
|
170 |
"pad_token_id": tokenizer.pad_token_id,
|
171 |
-
"top_k":
|
172 |
"do_sample": False,
|
173 |
-
"top_p":
|
174 |
-
"temperature":
|
175 |
}
|
176 |
|
177 |
with torch.no_grad():
|
@@ -208,5 +217,5 @@ Potential Delay Reasons:
|
|
208 |
|
209 |
Task: Analyze the provided video to identify the delay reason. Use the following format:
|
210 |
1. **Selected Reason:** [Choose the most likely reason from the list above]
|
211 |
-
2. **Visual Evidence:** [Describe specific visual cues from the
|
212 |
"""
|
|
|
10 |
DEVICE = 'cuda' if torch.cuda.is_available() else 'cpu'
|
11 |
TORCH_TYPE = torch.bfloat16 if torch.cuda.is_available() and torch.cuda.get_device_capability()[0] >= 8 else torch.float16
|
12 |
|
13 |
+
# Configurable constants
|
14 |
+
NUM_FRAMES = 24 # Default number of frames to extract
|
15 |
+
MAX_NEW_TOKENS = 2048
|
16 |
+
TOP_K = 1
|
17 |
+
TOP_P = 0.1
|
18 |
+
DEFAULT_TEMPERATURE = 1.0
|
19 |
+
|
20 |
# Delay Reasons for Each Manufacturing Step
|
21 |
DELAY_REASONS = {
|
22 |
"Step 1": ["Delay in Bead Insertion", "Lack of raw material"],
|
|
|
107 |
def load_video(video_data, strategy='chat'):
|
108 |
"""Loads and processes video data into a format suitable for model input."""
|
109 |
bridge.set_bridge('torch')
|
110 |
+
|
|
|
111 |
if isinstance(video_data, str):
|
112 |
decord_vr = VideoReader(video_data, ctx=cpu(0))
|
113 |
else:
|
114 |
decord_vr = VideoReader(io.BytesIO(video_data), ctx=cpu(0))
|
115 |
|
116 |
total_frames = len(decord_vr)
|
117 |
+
if total_frames < NUM_FRAMES:
|
118 |
raise ValueError("Uploaded video is too short for meaningful analysis.")
|
119 |
|
120 |
timestamps = [i[0] for i in decord_vr.get_frame_timestamp(np.arange(total_frames))]
|
|
|
125 |
closest_num = min(timestamps, key=lambda x: abs(x - second))
|
126 |
index = timestamps.index(closest_num)
|
127 |
frame_id_list.append(index)
|
128 |
+
if len(frame_id_list) >= NUM_FRAMES:
|
129 |
break
|
130 |
|
131 |
video_data = decord_vr.get_batch(frame_id_list)
|
|
|
154 |
|
155 |
def predict(prompt, video_data, temperature, model, tokenizer):
|
156 |
"""Generates predictions based on the video and textual prompt."""
|
157 |
+
try:
|
158 |
+
video = load_video(video_data, strategy='chat')
|
159 |
+
except ValueError as e:
|
160 |
+
return f"Error loading video: {str(e)}"
|
161 |
|
162 |
inputs = model.build_conversation_input_ids(
|
163 |
tokenizer=tokenizer,
|
|
|
175 |
}
|
176 |
|
177 |
gen_kwargs = {
|
178 |
+
"max_new_tokens": MAX_NEW_TOKENS,
|
179 |
"pad_token_id": tokenizer.pad_token_id,
|
180 |
+
"top_k": TOP_K,
|
181 |
"do_sample": False,
|
182 |
+
"top_p": TOP_P,
|
183 |
+
"temperature": temperature or DEFAULT_TEMPERATURE,
|
184 |
}
|
185 |
|
186 |
with torch.no_grad():
|
|
|
217 |
|
218 |
Task: Analyze the provided video to identify the delay reason. Use the following format:
|
219 |
1. **Selected Reason:** [Choose the most likely reason from the list above]
|
220 |
+
2. **Visual Evidence:** [Describe specific visual cues from the video that support your analysis.]
|
221 |
"""
|