Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
|
@@ -22,11 +22,11 @@ CONFIG = synthid_mixin.DEFAULT_WATERMARKING_CONFIG
|
|
| 22 |
# Function to check for AI-generated content using SynthID and highlight watermark
|
| 23 |
@spaces.GPU
|
| 24 |
def check_plagiarism(text):
|
| 25 |
-
|
| 26 |
# Logits processor for SynthID
|
| 27 |
logits_processor = logits_processing.SynthIDLogitsProcessor(
|
| 28 |
**CONFIG, top_k=40, temperature=0.5
|
| 29 |
)
|
|
|
|
| 30 |
# Tokenize and process the input text
|
| 31 |
inputs = tokenizer(text, return_tensors="pt").to(DEVICE)
|
| 32 |
|
|
@@ -49,9 +49,10 @@ def check_plagiarism(text):
|
|
| 49 |
token_scores = outputs.scores
|
| 50 |
|
| 51 |
# Loop through each generated token and its corresponding score
|
| 52 |
-
for
|
| 53 |
-
|
| 54 |
-
|
|
|
|
| 55 |
|
| 56 |
# If processed score indicates watermark, highlight this token
|
| 57 |
if processed_score.mean().item() > 0.5:
|
|
|
|
| 22 |
# Function to check for AI-generated content using SynthID and highlight watermark
|
| 23 |
@spaces.GPU
|
| 24 |
def check_plagiarism(text):
|
|
|
|
| 25 |
# Logits processor for SynthID
|
| 26 |
logits_processor = logits_processing.SynthIDLogitsProcessor(
|
| 27 |
**CONFIG, top_k=40, temperature=0.5
|
| 28 |
)
|
| 29 |
+
|
| 30 |
# Tokenize and process the input text
|
| 31 |
inputs = tokenizer(text, return_tensors="pt").to(DEVICE)
|
| 32 |
|
|
|
|
| 49 |
token_scores = outputs.scores
|
| 50 |
|
| 51 |
# Loop through each generated token and its corresponding score
|
| 52 |
+
for token_id, score in zip(generated_tokens, token_scores):
|
| 53 |
+
# Apply SynthIDLogitsProcessor to each score by calling it with 'scores=score'
|
| 54 |
+
processed_score = logits_processor(scores=score)
|
| 55 |
+
token_text = tokenizer.decode(token_id.unsqueeze(0)) # Decode token_id for individual token text
|
| 56 |
|
| 57 |
# If processed score indicates watermark, highlight this token
|
| 58 |
if processed_score.mean().item() > 0.5:
|