Spaces:
Runtime error
Runtime error
Zwea Htet
commited on
Commit
•
1230ae3
1
Parent(s):
08d67c4
fixed llama package issue
Browse files- app.py +10 -2
- models/bloom.py +3 -3
app.py
CHANGED
@@ -30,15 +30,23 @@ def validate(token: str):
|
|
30 |
return response
|
31 |
|
32 |
def create_index():
|
33 |
-
index = bloom.initialize_index("")
|
34 |
-
|
35 |
|
36 |
def get_response(vector_index, query_str):
|
37 |
query_engine = vector_index.as_query_engine()
|
38 |
response = query_engine.query(query_str)
|
39 |
return response
|
40 |
|
|
|
41 |
api_key = st.text_input("Enter your OpenAI API key here:", type="password")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
42 |
|
43 |
st.write("---")
|
44 |
input_text = st.text_area("Ask your question")
|
|
|
30 |
return response
|
31 |
|
32 |
def create_index():
|
33 |
+
index = bloom.initialize_index("bloomLlama")
|
34 |
+
return index
|
35 |
|
36 |
def get_response(vector_index, query_str):
|
37 |
query_engine = vector_index.as_query_engine()
|
38 |
response = query_engine.query(query_str)
|
39 |
return response
|
40 |
|
41 |
+
|
42 |
api_key = st.text_input("Enter your OpenAI API key here:", type="password")
|
43 |
+
if api_key:
|
44 |
+
resp = validate(api_key)
|
45 |
+
if ("error" in resp.json()):
|
46 |
+
st.info("Your API Token is incorrect! Try again.")
|
47 |
+
else:
|
48 |
+
os.environ["OPENAI_API_KEY"] = api_key
|
49 |
+
index = create_index()
|
50 |
|
51 |
st.write("---")
|
52 |
input_text = st.text_area("Ask your question")
|
models/bloom.py
CHANGED
@@ -20,12 +20,12 @@ model = AutoModelForCausalLM.from_pretrained(model_name, config='T5Config')
|
|
20 |
|
21 |
# define prompt helper
|
22 |
# set maximum input size
|
23 |
-
|
24 |
# set number of output tokens
|
25 |
num_output = 525
|
26 |
# set maximum chunk overlap
|
27 |
-
|
28 |
-
prompt_helper = PromptHelper(
|
29 |
|
30 |
# define llm
|
31 |
llm_predictor = LLMPredictor(llm=CustomLLM(model, tokenizer))
|
|
|
20 |
|
21 |
# define prompt helper
|
22 |
# set maximum input size
|
23 |
+
context_window = 2048
|
24 |
# set number of output tokens
|
25 |
num_output = 525
|
26 |
# set maximum chunk overlap
|
27 |
+
chunk_overlap_ratio = 0.2
|
28 |
+
prompt_helper = PromptHelper(context_window, num_output, chunk_overlap_ratio)
|
29 |
|
30 |
# define llm
|
31 |
llm_predictor = LLMPredictor(llm=CustomLLM(model, tokenizer))
|