Update app.py
Browse files
app.py
CHANGED
@@ -133,7 +133,6 @@ def split_biomodels(antimony_file_path):
|
|
133 |
return final_items
|
134 |
|
135 |
import chromadb
|
136 |
-
from llama_cpp import Llama
|
137 |
|
138 |
def create_vector_db(final_items):
|
139 |
global db
|
@@ -142,16 +141,19 @@ def create_vector_db(final_items):
|
|
142 |
from chromadb.utils import embedding_functions
|
143 |
embedding_function = embedding_functions.SentenceTransformerEmbeddingFunction(model_name="all-MiniLM-L6-v2")
|
144 |
|
|
|
145 |
db = client.get_or_create_collection(name=collection_name, embedding_function = embedding_function)
|
146 |
|
147 |
-
# Initialize Llama model
|
148 |
-
llm = Llama.from_pretrained(
|
149 |
-
repo_id="xzlinuxmodels/ollama3.1",
|
150 |
-
filename="unsloth.Q6_K.gguf"
|
151 |
-
)
|
152 |
|
153 |
documents = []
|
154 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
155 |
for item in final_items:
|
156 |
prompt = f"""
|
157 |
Summarize the following segment of Antimony in a clear and concise manner:
|
@@ -162,20 +164,14 @@ def create_vector_db(final_items):
|
|
162 |
|
163 |
Here is the antimony segment to summarize: {item}
|
164 |
"""
|
165 |
-
|
166 |
-
|
167 |
-
|
168 |
-
temperature=0.0,
|
169 |
-
top_p=0.1,
|
170 |
-
echo=False,
|
171 |
-
stop=["Q", "\n"]
|
172 |
-
)
|
173 |
-
documents.append(response["choices"][0]["text"].strip())
|
174 |
|
175 |
if final_items:
|
176 |
db.add(
|
177 |
documents=documents,
|
178 |
-
ids=[f"id{i}" for i in range(len(
|
179 |
)
|
180 |
|
181 |
return db
|
@@ -190,12 +186,15 @@ def generate_response(db, query_text, previous_context):
|
|
190 |
return "No results found."
|
191 |
|
192 |
best_recommendation = query_results['documents']
|
193 |
-
|
|
|
|
|
|
|
|
|
194 |
|
195 |
-
|
196 |
-
|
197 |
-
|
198 |
-
)
|
199 |
|
200 |
prompt_template = f"""
|
201 |
Using the context provided below, answer the following question. If the information is insufficient to answer the question, please state that clearly.
|
@@ -210,19 +209,15 @@ def generate_response(db, query_text, previous_context):
|
|
210 |
|
211 |
Question:
|
212 |
{query_text}
|
213 |
-
|
214 |
"""
|
215 |
-
|
216 |
-
|
217 |
-
max_tokens = 100000000,
|
218 |
-
temperature=0.0,
|
219 |
-
top_p=0.1,
|
220 |
-
echo=False,
|
221 |
-
stop = ["Q", "\n"]
|
222 |
-
)
|
223 |
-
final_response = response["choices"][0]["text"].strip()
|
224 |
-
return final_response
|
225 |
|
|
|
|
|
|
|
|
|
|
|
226 |
def streamlit_app():
|
227 |
st.title("BioModels Chat Interface")
|
228 |
|
@@ -235,8 +230,7 @@ def streamlit_app():
|
|
235 |
model_ids = list(models.keys())
|
236 |
selected_models = st.multiselect(
|
237 |
"Select biomodels to analyze",
|
238 |
-
options=model_ids
|
239 |
-
default=[model_ids[0]]
|
240 |
)
|
241 |
|
242 |
if st.button("Analyze Selected Models"):
|
@@ -279,4 +273,4 @@ def streamlit_app():
|
|
279 |
st.write("No models found for the given search query.")
|
280 |
|
281 |
if __name__ == "__main__":
|
282 |
-
streamlit_app()
|
|
|
133 |
return final_items
|
134 |
|
135 |
import chromadb
|
|
|
136 |
|
137 |
def create_vector_db(final_items):
|
138 |
global db
|
|
|
141 |
from chromadb.utils import embedding_functions
|
142 |
embedding_function = embedding_functions.SentenceTransformerEmbeddingFunction(model_name="all-MiniLM-L6-v2")
|
143 |
|
144 |
+
|
145 |
db = client.get_or_create_collection(name=collection_name, embedding_function = embedding_function)
|
146 |
|
|
|
|
|
|
|
|
|
|
|
147 |
|
148 |
documents = []
|
149 |
|
150 |
+
from transformers import AutoModelForCausalLM, AutoTokenizer, BitsAndBytesConfig
|
151 |
+
|
152 |
+
quantization_config = BitsAndBytesConfig(load_in_8bit=True)
|
153 |
+
checkpoint = "HuggingFaceTB/SmolLM-135M"
|
154 |
+
tokenizer = AutoTokenizer.from_pretrained(checkpoint)
|
155 |
+
model = AutoModelForCausalLM.from_pretrained(checkpoint, quantization_config=quantization_config)
|
156 |
+
|
157 |
for item in final_items:
|
158 |
prompt = f"""
|
159 |
Summarize the following segment of Antimony in a clear and concise manner:
|
|
|
164 |
|
165 |
Here is the antimony segment to summarize: {item}
|
166 |
"""
|
167 |
+
inputs = tokenizer.encode(prompt, return_tensors="pt").to("cuda")
|
168 |
+
response = model.generate(inputs)
|
169 |
+
documents.append(tokenizer.decode(response[0]))
|
|
|
|
|
|
|
|
|
|
|
|
|
170 |
|
171 |
if final_items:
|
172 |
db.add(
|
173 |
documents=documents,
|
174 |
+
ids=[f"id{i}" for i in range(len(documents))]
|
175 |
)
|
176 |
|
177 |
return db
|
|
|
186 |
return "No results found."
|
187 |
|
188 |
best_recommendation = query_results['documents']
|
189 |
+
|
190 |
+
import torch
|
191 |
+
from transformers import AutoTokenizer, AutoModelForCausalLM
|
192 |
+
model_path = "nvidia/Mistral-NeMo-Minitron-8B-Base"
|
193 |
+
tokenizer = AutoTokenizer.from_pretrained(model_path)
|
194 |
|
195 |
+
device = 'cuda'
|
196 |
+
dtype = torch.bfloat16
|
197 |
+
model = AutoModelForCausalLM.from_pretrained(model_path, torch_dtype=dtype, device_map=device)
|
|
|
198 |
|
199 |
prompt_template = f"""
|
200 |
Using the context provided below, answer the following question. If the information is insufficient to answer the question, please state that clearly.
|
|
|
209 |
|
210 |
Question:
|
211 |
{query_text}
|
|
|
212 |
"""
|
213 |
+
inputs = tokenizer.encode(prompt_template, return_tensors='pt').to(model.device)
|
214 |
+
outputs = model.generate(inputs, max_length=20000000000000000)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
215 |
|
216 |
+
# Decode and print the output
|
217 |
+
response = tokenizer.decode(outputs[0])
|
218 |
+
print(response)
|
219 |
+
|
220 |
+
|
221 |
def streamlit_app():
|
222 |
st.title("BioModels Chat Interface")
|
223 |
|
|
|
230 |
model_ids = list(models.keys())
|
231 |
selected_models = st.multiselect(
|
232 |
"Select biomodels to analyze",
|
233 |
+
options=model_ids
|
|
|
234 |
)
|
235 |
|
236 |
if st.button("Analyze Selected Models"):
|
|
|
273 |
st.write("No models found for the given search query.")
|
274 |
|
275 |
if __name__ == "__main__":
|
276 |
+
streamlit_app()
|