Spaces:
Running
Running
Update app.py
Browse files
app.py
CHANGED
|
@@ -1,30 +1,46 @@
|
|
| 1 |
import gradio as gr
|
| 2 |
from transformers import AutoTokenizer, AutoModel
|
| 3 |
import torch
|
|
|
|
| 4 |
|
| 5 |
# Load model and tokenizer
|
| 6 |
tokenizer = AutoTokenizer.from_pretrained("Qwen/Qwen3-Embedding-0.6B")
|
| 7 |
model = AutoModel.from_pretrained("Qwen/Qwen3-Embedding-0.6B")
|
| 8 |
|
| 9 |
def get_embedding(text):
|
| 10 |
-
if len(text) > 250:
|
| 11 |
-
return "❌ Error: Input exceeds 250 character limit."
|
| 12 |
-
|
| 13 |
inputs = tokenizer(text, return_tensors="pt", truncation=True)
|
| 14 |
with torch.no_grad():
|
| 15 |
outputs = model(**inputs)
|
| 16 |
-
#
|
| 17 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 18 |
|
| 19 |
-
|
| 20 |
-
return f"✅ Embedding (first 10 values): {embedding[:10]}..."
|
| 21 |
|
| 22 |
demo = gr.Interface(
|
| 23 |
-
fn=
|
| 24 |
-
inputs=
|
|
|
|
|
|
|
|
|
|
| 25 |
outputs="text",
|
| 26 |
-
title="Qwen3 Embedding Demo",
|
| 27 |
-
description="
|
| 28 |
)
|
| 29 |
|
| 30 |
if __name__ == "__main__":
|
|
|
|
| 1 |
import gradio as gr
|
| 2 |
from transformers import AutoTokenizer, AutoModel
|
| 3 |
import torch
|
| 4 |
+
import torch.nn.functional as F
|
| 5 |
|
| 6 |
# Load model and tokenizer
|
| 7 |
tokenizer = AutoTokenizer.from_pretrained("Qwen/Qwen3-Embedding-0.6B")
|
| 8 |
model = AutoModel.from_pretrained("Qwen/Qwen3-Embedding-0.6B")
|
| 9 |
|
| 10 |
def get_embedding(text):
|
|
|
|
|
|
|
|
|
|
| 11 |
inputs = tokenizer(text, return_tensors="pt", truncation=True)
|
| 12 |
with torch.no_grad():
|
| 13 |
outputs = model(**inputs)
|
| 14 |
+
return outputs.last_hidden_state[:, 0, :] # [CLS] token
|
| 15 |
+
|
| 16 |
+
def compare_sentences(reference, comparisons):
|
| 17 |
+
if len(reference) > 250:
|
| 18 |
+
return "❌ Error: Reference exceeds 250 character limit."
|
| 19 |
+
|
| 20 |
+
comparison_list = [s.strip() for s in comparisons.strip().split('\n') if s.strip()]
|
| 21 |
+
if not comparison_list:
|
| 22 |
+
return "❌ Error: No comparison sentences provided."
|
| 23 |
+
|
| 24 |
+
if any(len(s) > 250 for s in comparison_list):
|
| 25 |
+
return "❌ Error: One or more comparison sentences exceed 250 characters."
|
| 26 |
+
|
| 27 |
+
ref_emb = get_embedding(reference)
|
| 28 |
+
comp_embs = torch.cat([get_embedding(s) for s in comparison_list], dim=0)
|
| 29 |
+
|
| 30 |
+
similarities = F.cosine_similarity(ref_emb, comp_embs).tolist()
|
| 31 |
+
results = "\n".join([f"Similarity with: \"{s}\"\n→ {round(score, 4)}" for s, score in zip(comparison_list, similarities)])
|
| 32 |
|
| 33 |
+
return results
|
|
|
|
| 34 |
|
| 35 |
demo = gr.Interface(
|
| 36 |
+
fn=compare_sentences,
|
| 37 |
+
inputs=[
|
| 38 |
+
gr.Textbox(label="Reference Sentence (max 250 characters)", lines=2, placeholder="Type the reference sentence here..."),
|
| 39 |
+
gr.Textbox(label="Comparison Sentences (one per line, each max 250 characters)", lines=8, placeholder="Type comparison sentences here, one per line..."),
|
| 40 |
+
],
|
| 41 |
outputs="text",
|
| 42 |
+
title="Qwen3 Embedding Comparison Demo",
|
| 43 |
+
description="Enter a reference sentence and multiple comparison sentences (one per line). The model computes the cosine similarity between the reference and each comparison."
|
| 44 |
)
|
| 45 |
|
| 46 |
if __name__ == "__main__":
|