Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
|
@@ -2,9 +2,7 @@ import gradio as gr
|
|
| 2 |
from langchain.prompts import ChatPromptTemplate
|
| 3 |
from langchain_huggingface import HuggingFaceEndpoint
|
| 4 |
from langchain.schema import HumanMessage, SystemMessage
|
| 5 |
-
from langchain.callbacks
|
| 6 |
-
import os
|
| 7 |
-
import time
|
| 8 |
|
| 9 |
# Custom streaming callback for Gradio
|
| 10 |
class GradioStreamingCallback(BaseCallbackHandler):
|
|
@@ -38,11 +36,8 @@ llm = HuggingFaceEndpoint(
|
|
| 38 |
repo_id="HuggingFaceH4/zephyr-7b-beta",
|
| 39 |
temperature=0.7,
|
| 40 |
top_p=0.9,
|
| 41 |
-
|
| 42 |
-
|
| 43 |
-
"return_full_text": False,
|
| 44 |
-
"repetition_penalty": 1.1,
|
| 45 |
-
},
|
| 46 |
huggingfacehub_api_token=os.getenv("HUGGINGFACEHUB_API_TOKEN")
|
| 47 |
)
|
| 48 |
|
|
|
|
| 2 |
from langchain.prompts import ChatPromptTemplate
|
| 3 |
from langchain_huggingface import HuggingFaceEndpoint
|
| 4 |
from langchain.schema import HumanMessage, SystemMessage
|
| 5 |
+
from langchain.callbacks
|
|
|
|
|
|
|
| 6 |
|
| 7 |
# Custom streaming callback for Gradio
|
| 8 |
class GradioStreamingCallback(BaseCallbackHandler):
|
|
|
|
| 36 |
repo_id="HuggingFaceH4/zephyr-7b-beta",
|
| 37 |
temperature=0.7,
|
| 38 |
top_p=0.9,
|
| 39 |
+
repetition_penalty=1.1,
|
| 40 |
+
model_kwargs={"max_length": 1536},
|
|
|
|
|
|
|
|
|
|
| 41 |
huggingfacehub_api_token=os.getenv("HUGGINGFACEHUB_API_TOKEN")
|
| 42 |
)
|
| 43 |
|