captain-awesome
commited on
Update app.py
Browse files
app.py
CHANGED
@@ -11,13 +11,12 @@ from langchain_community.embeddings import HuggingFaceBgeEmbeddings
|
|
11 |
from langchain_community.llms import CTransformers
|
12 |
from ctransformers import AutoModelForCausalLM
|
13 |
from langchain.llms import HuggingFaceHub
|
14 |
-
from transformers import AutoTokenizer
|
15 |
import os
|
16 |
# from dotenv import load_dotenv
|
17 |
|
18 |
# load_dotenv()
|
19 |
|
20 |
-
os.environ['HUGGINGFACEHUB_API_TOKEN'] = os.getenv("HF_KEY")
|
21 |
|
22 |
|
23 |
def get_vector_store_from_url(url):
|
@@ -85,11 +84,18 @@ def get_response(user_input):
|
|
85 |
# lib="avx2", # for CPU
|
86 |
# )
|
87 |
|
88 |
-
|
89 |
-
llm = HuggingFaceHub(
|
90 |
-
|
91 |
-
|
92 |
-
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
93 |
retriever_chain = get_context_retriever_chain(st.session_state.vector_store,llm)
|
94 |
conversation_rag_chain = get_conversational_rag_chain(retriever_chain,llm)
|
95 |
|
|
|
11 |
from langchain_community.llms import CTransformers
|
12 |
from ctransformers import AutoModelForCausalLM
|
13 |
from langchain.llms import HuggingFaceHub
|
14 |
+
from transformers import AutoModelForCausalLM, AutoTokenizer
|
15 |
import os
|
16 |
# from dotenv import load_dotenv
|
17 |
|
18 |
# load_dotenv()
|
19 |
|
|
|
20 |
|
21 |
|
22 |
def get_vector_store_from_url(url):
|
|
|
84 |
# lib="avx2", # for CPU
|
85 |
# )
|
86 |
|
87 |
+
model_name = "TinyLlama/TinyLlama-1.1B-Chat-v1.0"
|
88 |
+
# llm = HuggingFaceHub(
|
89 |
+
# repo_id=llm_model,
|
90 |
+
# model_kwargs={"temperature": 0.3, "max_new_tokens": 250, "top_k": 3}
|
91 |
+
# )
|
92 |
+
|
93 |
+
model = transformers.AutoModelForCausalLM.from_pretrained(
|
94 |
+
model_name,
|
95 |
+
trust_remote_code=True,
|
96 |
+
torch_dtype=torch.bfloat16,
|
97 |
+
device_map='auto'
|
98 |
+
|
99 |
retriever_chain = get_context_retriever_chain(st.session_state.vector_store,llm)
|
100 |
conversation_rag_chain = get_conversational_rag_chain(retriever_chain,llm)
|
101 |
|