Spaces:
Running
Running
Update app.py
Browse files
app.py
CHANGED
@@ -31,7 +31,6 @@ CHUNK_OVERLAP = 100
|
|
31 |
MAX_CONTEXT_CHUNKS = 3
|
32 |
MODEL_ID = "ibm-granite/granite-vision-3.2-2b"
|
33 |
|
34 |
-
# Device selection
|
35 |
device = "cuda" if torch.cuda.is_available() else "cpu"
|
36 |
|
37 |
# ---------------- Text Helpers ----------------
|
@@ -86,74 +85,59 @@ def extract_docx_text(path):
|
|
86 |
|
87 |
# ---------------- Embedding ----------------
|
88 |
def embed_all():
|
89 |
-
|
90 |
-
|
91 |
-
|
92 |
-
except Exception as e:
|
93 |
-
print("\u274c Failed to load SentenceTransformer:", e)
|
94 |
-
return None, None
|
95 |
|
96 |
try:
|
97 |
-
client = chromadb.PersistentClient(path=CHROMA_PATH)
|
98 |
client.delete_collection(COLLECTION_NAME)
|
99 |
-
|
100 |
-
|
101 |
-
|
102 |
-
return None, None
|
103 |
|
104 |
docs, ids, metas = [], [], []
|
105 |
print("\ud83d\udcc4 Processing manuals...")
|
106 |
|
107 |
-
|
108 |
-
|
109 |
-
|
110 |
-
|
111 |
-
|
112 |
-
|
113 |
-
|
114 |
-
|
115 |
-
|
116 |
-
|
117 |
-
for
|
118 |
-
|
119 |
-
|
120 |
-
|
121 |
-
|
122 |
-
|
123 |
-
|
124 |
-
|
125 |
-
|
126 |
-
|
127 |
-
|
128 |
-
|
129 |
-
|
130 |
-
|
131 |
-
|
132 |
-
|
133 |
-
|
134 |
-
return collection, embedder
|
135 |
-
|
136 |
-
except Exception as e:
|
137 |
-
print("\u274c Error during embedding:", e)
|
138 |
-
return None, None
|
139 |
|
140 |
# ---------------- Model Setup ----------------
|
141 |
def load_model():
|
142 |
-
|
143 |
-
|
144 |
-
|
145 |
-
|
146 |
-
|
147 |
-
|
148 |
-
|
149 |
-
|
150 |
-
|
151 |
-
return pipe, tokenizer
|
152 |
-
except Exception as e:
|
153 |
-
print("\u274c Failed to load model:", e)
|
154 |
-
return None, None
|
155 |
|
156 |
-
# ---------------- QA Logic ----------------
|
157 |
def ask_model(question, context, pipe, tokenizer):
|
158 |
prompt = f"""Use only the following context to answer. If uncertain, say \"I don't know.\"
|
159 |
|
@@ -166,10 +150,12 @@ A:"""
|
|
166 |
output = pipe(prompt, max_new_tokens=512)[0]["generated_text"]
|
167 |
return output.split("A:")[-1].strip()
|
168 |
|
|
|
169 |
def get_answer(question):
|
170 |
if not all([embedder, db, model_pipe, model_tokenizer]):
|
171 |
-
return "
|
172 |
try:
|
|
|
173 |
results = db.query(query_texts=[question], n_results=MAX_CONTEXT_CHUNKS)
|
174 |
context = "\n\n".join(results["documents"][0])
|
175 |
return ask_model(question, context, model_pipe, model_tokenizer)
|
@@ -184,31 +170,20 @@ with gr.Blocks() as demo:
|
|
184 |
question = gr.Textbox(label="Ask your question")
|
185 |
ask = gr.Button("Ask")
|
186 |
answer = gr.Textbox(label="Answer", lines=8)
|
187 |
-
|
188 |
|
189 |
-
|
190 |
-
|
191 |
-
return ans, "" # hide status after success
|
192 |
-
|
193 |
-
ask.click(fn=wrapped_get_answer, inputs=question, outputs=[answer, status])
|
194 |
-
|
195 |
-
# Show status on startup error
|
196 |
-
if not all([embedder, db, model_pipe, model_tokenizer]):
|
197 |
-
status.visible = True
|
198 |
-
status.value = "\u26a0\ufe0f Initialization failed. Check logs or your HF_TOKEN."
|
199 |
|
200 |
-
# Embed + Load Model at Startup
|
201 |
try:
|
202 |
db, embedder = embed_all()
|
203 |
except Exception as e:
|
204 |
print("\u274c Embedding failed:", e)
|
205 |
-
db, embedder = None, None
|
206 |
|
207 |
try:
|
208 |
model_pipe, model_tokenizer = load_model()
|
209 |
except Exception as e:
|
210 |
print("\u274c Model loading failed:", e)
|
211 |
-
model_pipe, model_tokenizer = None, None
|
212 |
|
213 |
if __name__ == "__main__":
|
214 |
demo.launch()
|
|
|
31 |
MAX_CONTEXT_CHUNKS = 3
|
32 |
MODEL_ID = "ibm-granite/granite-vision-3.2-2b"
|
33 |
|
|
|
34 |
device = "cuda" if torch.cuda.is_available() else "cpu"
|
35 |
|
36 |
# ---------------- Text Helpers ----------------
|
|
|
85 |
|
86 |
# ---------------- Embedding ----------------
|
87 |
def embed_all():
|
88 |
+
embedder = SentenceTransformer("all-MiniLM-L6-v2")
|
89 |
+
embedder.eval()
|
90 |
+
client = chromadb.PersistentClient(path=CHROMA_PATH)
|
|
|
|
|
|
|
91 |
|
92 |
try:
|
|
|
93 |
client.delete_collection(COLLECTION_NAME)
|
94 |
+
except:
|
95 |
+
pass
|
96 |
+
collection = client.get_or_create_collection(COLLECTION_NAME)
|
|
|
97 |
|
98 |
docs, ids, metas = [], [], []
|
99 |
print("\ud83d\udcc4 Processing manuals...")
|
100 |
|
101 |
+
for fname in os.listdir(MANUALS_DIR):
|
102 |
+
fpath = os.path.join(MANUALS_DIR, fname)
|
103 |
+
if fname.lower().endswith(".pdf"):
|
104 |
+
pages = extract_pdf_text(fpath)
|
105 |
+
elif fname.lower().endswith(".docx"):
|
106 |
+
pages = extract_docx_text(fpath)
|
107 |
+
else:
|
108 |
+
continue
|
109 |
+
|
110 |
+
for path, page, text in pages:
|
111 |
+
for i, chunk in enumerate(split_chunks(split_sentences(text))):
|
112 |
+
chunk_id = f"{fname}::{page}::{i}"
|
113 |
+
docs.append(chunk)
|
114 |
+
ids.append(chunk_id)
|
115 |
+
metas.append({"source": fname, "page": page})
|
116 |
+
|
117 |
+
if len(docs) >= 16:
|
118 |
+
embs = embedder.encode(docs).tolist()
|
119 |
+
collection.add(documents=docs, ids=ids, metadatas=metas, embeddings=embs)
|
120 |
+
docs, ids, metas = [], [], []
|
121 |
+
|
122 |
+
if docs:
|
123 |
+
embs = embedder.encode(docs).tolist()
|
124 |
+
collection.add(documents=docs, ids=ids, metadatas=metas, embeddings=embs)
|
125 |
+
|
126 |
+
print(f"\u2705 Embedded {len(ids)} chunks.")
|
127 |
+
return collection, embedder
|
|
|
|
|
|
|
|
|
|
|
128 |
|
129 |
# ---------------- Model Setup ----------------
|
130 |
def load_model():
|
131 |
+
tokenizer = AutoTokenizer.from_pretrained(MODEL_ID, token=HF_TOKEN)
|
132 |
+
model = AutoModelForCausalLM.from_pretrained(
|
133 |
+
MODEL_ID,
|
134 |
+
token=HF_TOKEN,
|
135 |
+
device_map="auto" if torch.cuda.is_available() else None,
|
136 |
+
torch_dtype=torch.bfloat16 if torch.cuda.is_available() else torch.float32
|
137 |
+
).to(device)
|
138 |
+
pipe = pipeline("text-generation", model=model, tokenizer=tokenizer, device=0 if torch.cuda.is_available() else -1)
|
139 |
+
return pipe, tokenizer
|
|
|
|
|
|
|
|
|
140 |
|
|
|
141 |
def ask_model(question, context, pipe, tokenizer):
|
142 |
prompt = f"""Use only the following context to answer. If uncertain, say \"I don't know.\"
|
143 |
|
|
|
150 |
output = pipe(prompt, max_new_tokens=512)[0]["generated_text"]
|
151 |
return output.split("A:")[-1].strip()
|
152 |
|
153 |
+
# ---------------- Query ----------------
|
154 |
def get_answer(question):
|
155 |
if not all([embedder, db, model_pipe, model_tokenizer]):
|
156 |
+
return "⚠️ The system is still initializing or failed to load. Please try again later."
|
157 |
try:
|
158 |
+
query_emb = embedder.encode(question, convert_to_tensor=True)
|
159 |
results = db.query(query_texts=[question], n_results=MAX_CONTEXT_CHUNKS)
|
160 |
context = "\n\n".join(results["documents"][0])
|
161 |
return ask_model(question, context, model_pipe, model_tokenizer)
|
|
|
170 |
question = gr.Textbox(label="Ask your question")
|
171 |
ask = gr.Button("Ask")
|
172 |
answer = gr.Textbox(label="Answer", lines=8)
|
173 |
+
ask.click(fn=get_answer, inputs=question, outputs=answer)
|
174 |
|
175 |
+
# ---------------- Startup ----------------
|
176 |
+
embedder = db = model_pipe = model_tokenizer = None
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
177 |
|
|
|
178 |
try:
|
179 |
db, embedder = embed_all()
|
180 |
except Exception as e:
|
181 |
print("\u274c Embedding failed:", e)
|
|
|
182 |
|
183 |
try:
|
184 |
model_pipe, model_tokenizer = load_model()
|
185 |
except Exception as e:
|
186 |
print("\u274c Model loading failed:", e)
|
|
|
187 |
|
188 |
if __name__ == "__main__":
|
189 |
demo.launch()
|