VictorM-Coder commited on
Commit
fe75c23
·
verified ·
1 Parent(s): 25b8beb

Update app.py

Browse files
Files changed (1) hide show
  1. app.py +14 -14
app.py CHANGED
@@ -18,13 +18,18 @@ def get_model():
18
  global tokenizer, model
19
  if model is None:
20
  tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME)
21
- dtype = torch.bfloat16 if (device.type == "cuda" and torch.cuda.is_bf16_supported()) else torch.float32
 
 
 
 
 
22
  model = AutoModelForSequenceClassification.from_pretrained(
23
  MODEL_NAME, torch_dtype=dtype
24
  ).to(device).eval()
25
  return tokenizer, model
26
 
27
- # Threshold set to 20% per your request
28
  THRESHOLD = 0.20
29
 
30
  # -----------------------------
@@ -101,21 +106,16 @@ def analyze(text):
101
  prob_map = {idx: probs[i] for i, idx in enumerate(pure_sents_indices)}
102
 
103
  for i, block in enumerate(blocks):
104
- if block.startswith("\n"):
105
  highlighted_html += block.replace("\n", "<br>")
106
  continue
107
- if block.isspace():
108
- highlighted_html += block
109
- continue
110
 
111
  if i in prob_map:
112
  score = prob_map[i]
113
- if score < 0.35:
114
- color, bg = "#11823b", "rgba(17, 130, 59, 0.15)"
115
- elif score < 0.70:
116
- color, bg = "#b8860b", "rgba(184, 134, 11, 0.15)"
117
- else:
118
- color, bg = "#b80d0d", "rgba(184, 13, 13, 0.15)"
119
 
120
  highlighted_html += (
121
  f"<span style='background:{bg}; padding:2px 4px; border-radius:4px; border-bottom: 2px solid {color};' "
@@ -127,7 +127,7 @@ def analyze(text):
127
 
128
  highlighted_html += "</div>"
129
 
130
- # Logic Updated: > 20% triggers detection label
131
  label = "AI Content Detected" if weighted_avg >= THRESHOLD else "0 or * AI Content Detected"
132
 
133
  df = pd.DataFrame({"Sentence": pure_sents, "AI Confidence": [f"{p:.1%}" for p in probs]})
@@ -157,4 +157,4 @@ with gr.Blocks(theme=gr.themes.Soft()) as demo:
157
  run_btn.click(analyze, inputs=text_input, outputs=[verdict_out, score_out, html_out, table_out])
158
 
159
  if __name__ == "__main__":
160
- demo.launch()
 
18
  global tokenizer, model
19
  if model is None:
20
  tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME)
21
+
22
+ # Hugging Face Build Fix: Force float32 unless a capable GPU is detected
23
+ dtype = torch.float32
24
+ if device.type == "cuda" and torch.cuda.is_bf16_supported():
25
+ dtype = torch.bfloat16
26
+
27
  model = AutoModelForSequenceClassification.from_pretrained(
28
  MODEL_NAME, torch_dtype=dtype
29
  ).to(device).eval()
30
  return tokenizer, model
31
 
32
+ # Only affects the content label as requested
33
  THRESHOLD = 0.20
34
 
35
  # -----------------------------
 
106
  prob_map = {idx: probs[i] for i, idx in enumerate(pure_sents_indices)}
107
 
108
  for i, block in enumerate(blocks):
109
+ if block.startswith("\n") or block.isspace():
110
  highlighted_html += block.replace("\n", "<br>")
111
  continue
 
 
 
112
 
113
  if i in prob_map:
114
  score = prob_map[i]
115
+ # Standard coloring thresholds (Independent of the final label threshold)
116
+ if score < 0.35: color, bg = "#11823b", "rgba(17, 130, 59, 0.15)"
117
+ elif score < 0.70: color, bg = "#b8860b", "rgba(184, 134, 11, 0.15)"
118
+ else: color, bg = "#b80d0d", "rgba(184, 13, 13, 0.15)"
 
 
119
 
120
  highlighted_html += (
121
  f"<span style='background:{bg}; padding:2px 4px; border-radius:4px; border-bottom: 2px solid {color};' "
 
127
 
128
  highlighted_html += "</div>"
129
 
130
+ # CONTENT LABEL LOGIC
131
  label = "AI Content Detected" if weighted_avg >= THRESHOLD else "0 or * AI Content Detected"
132
 
133
  df = pd.DataFrame({"Sentence": pure_sents, "AI Confidence": [f"{p:.1%}" for p in probs]})
 
157
  run_btn.click(analyze, inputs=text_input, outputs=[verdict_out, score_out, html_out, table_out])
158
 
159
  if __name__ == "__main__":
160
+ demo.launch()