Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
@@ -1,10 +1,13 @@
|
|
1 |
import gradio as gr
|
2 |
from transformers import AutoTokenizer, AutoModel
|
|
|
3 |
import torch
|
|
|
4 |
from gradio_client import Client
|
5 |
-
from functools import lru_cache
|
6 |
|
7 |
# Cache the model and tokenizer using lru_cache
|
|
|
|
|
8 |
@lru_cache(maxsize=1)
|
9 |
def load_model_and_tokenizer():
|
10 |
model_name = "./all-MiniLM-L6-v2" # Replace with your Space and model path
|
@@ -15,41 +18,85 @@ def load_model_and_tokenizer():
|
|
15 |
# Load the model and tokenizer
|
16 |
tokenizer, model = load_model_and_tokenizer()
|
17 |
|
18 |
-
#
|
19 |
-
|
20 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
21 |
inputs = tokenizer([input_text], padding=True, truncation=True, return_tensors="pt")
|
22 |
-
|
23 |
-
# Run the model
|
24 |
with torch.no_grad():
|
25 |
outputs = model(**inputs)
|
26 |
-
|
27 |
-
|
28 |
-
|
29 |
-
|
30 |
-
|
31 |
-
#
|
32 |
-
|
|
|
|
|
33 |
|
34 |
# Translation client
|
35 |
translation_client = Client("Frenchizer/space_3")
|
36 |
|
37 |
-
def translate_text(input_text):
|
|
|
38 |
return translation_client.predict(input_text)
|
39 |
|
40 |
def process_request(input_text):
|
41 |
-
|
42 |
-
|
43 |
-
|
44 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
45 |
|
46 |
-
# Gradio interface
|
47 |
interface = gr.Interface(
|
48 |
-
fn=
|
49 |
inputs="text",
|
50 |
outputs="text",
|
51 |
title="Frenchizer",
|
52 |
-
description="Translate text from English to French with context detection."
|
53 |
)
|
54 |
|
55 |
-
interface.launch()
|
|
|
1 |
import gradio as gr
|
2 |
from transformers import AutoTokenizer, AutoModel
|
3 |
+
from sklearn.metrics.pairwise import cosine_similarity
|
4 |
import torch
|
5 |
+
import numpy as np
|
6 |
from gradio_client import Client
|
|
|
7 |
|
8 |
# Cache the model and tokenizer using lru_cache
|
9 |
+
from functools import lru_cache
|
10 |
+
|
11 |
@lru_cache(maxsize=1)
|
12 |
def load_model_and_tokenizer():
|
13 |
model_name = "./all-MiniLM-L6-v2" # Replace with your Space and model path
|
|
|
18 |
# Load the model and tokenizer
|
19 |
tokenizer, model = load_model_and_tokenizer()
|
20 |
|
21 |
+
# Precompute label embeddings
|
22 |
+
labels = [
|
23 |
+
"aerospace", "anatomy", "anthropology", "art",
|
24 |
+
"automotive", "blockchain", "biology", "chemistry",
|
25 |
+
"cryptocurrency", "data science", "design", "e-commerce",
|
26 |
+
"education", "engineering", "entertainment", "environment",
|
27 |
+
"fashion", "finance", "food commerce", "general",
|
28 |
+
"gaming", "healthcare", "history", "html",
|
29 |
+
"information technology", "IT", "keywords", "legal",
|
30 |
+
"literature", "machine learning", "marketing", "medicine",
|
31 |
+
"music", "personal development", "philosophy", "physics",
|
32 |
+
"politics", "poetry", "programming", "real estate", "retail",
|
33 |
+
"robotics", "slang", "social media", "speech", "sports",
|
34 |
+
"sustained", "technical", "theater", "tourism", "travel"
|
35 |
+
]
|
36 |
+
|
37 |
+
@lru_cache(maxsize=1)
|
38 |
+
def precompute_label_embeddings():
|
39 |
+
inputs = tokenizer(labels, padding=True, truncation=True, return_tensors="pt")
|
40 |
+
with torch.no_grad():
|
41 |
+
outputs = model(**inputs)
|
42 |
+
return outputs.last_hidden_state.mean(dim=1).numpy() # Mean pooling for embeddings
|
43 |
+
|
44 |
+
label_embeddings = precompute_label_embeddings()
|
45 |
+
|
46 |
+
# Function to detect context
|
47 |
+
def detect_context(input_text, fallback_threshold=0.8, max_results=3):
|
48 |
+
# Encode the input text
|
49 |
inputs = tokenizer([input_text], padding=True, truncation=True, return_tensors="pt")
|
|
|
|
|
50 |
with torch.no_grad():
|
51 |
outputs = model(**inputs)
|
52 |
+
input_embedding = outputs.last_hidden_state.mean(dim=1).numpy() # Mean pooling for embedding
|
53 |
+
|
54 |
+
# Compute cosine similarities
|
55 |
+
similarities = cosine_similarity(input_embedding, label_embeddings)[0]
|
56 |
+
|
57 |
+
# Check for fallback matches
|
58 |
+
fallback_labels = [(labels[i], score) for i, score in enumerate(similarities) if score >= fallback_threshold]
|
59 |
+
fallback_labels = sorted(fallback_labels, key=lambda x: x[1], reverse=True)[:max_results]
|
60 |
+
return fallback_labels
|
61 |
|
62 |
# Translation client
|
63 |
translation_client = Client("Frenchizer/space_3")
|
64 |
|
65 |
+
def translate_text(input_text, context="general"):
|
66 |
+
# Append the context to the input text for the translation client (if needed)
|
67 |
return translation_client.predict(input_text)
|
68 |
|
69 |
def process_request(input_text):
|
70 |
+
# Step 1: Return the general translation immediately
|
71 |
+
general_translation = translate_text(input_text, context="general")
|
72 |
+
|
73 |
+
# Step 2: Detect context in the background
|
74 |
+
context_results = detect_context(input_text)
|
75 |
+
|
76 |
+
# Step 3: Generate additional translations for high-confidence contexts
|
77 |
+
additional_translations = {}
|
78 |
+
for context, score in context_results:
|
79 |
+
if context != "general":
|
80 |
+
additional_translations[context] = translate_text(input_text, context=context)
|
81 |
+
|
82 |
+
# Return the general translation and additional context translations
|
83 |
+
return general_translation, additional_translations
|
84 |
+
|
85 |
+
# Gradio interface with multiple outputs
|
86 |
+
def gradio_interface(input_text):
|
87 |
+
general_translation, additional_translations = process_request(input_text)
|
88 |
+
outputs = f"General Translation: {general_translation}\n\n"
|
89 |
+
for context, translation in additional_translations.items():
|
90 |
+
outputs += f"Context ({context}): {translation}\n\n"
|
91 |
+
return outputs.strip()
|
92 |
|
93 |
+
# Create the Gradio interface
|
94 |
interface = gr.Interface(
|
95 |
+
fn=gradio_interface,
|
96 |
inputs="text",
|
97 |
outputs="text",
|
98 |
title="Frenchizer",
|
99 |
+
description="Translate text from English to French with optimized context detection."
|
100 |
)
|
101 |
|
102 |
+
interface.launch()
|