jalr/tabbyAPI-ollama
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

workaround for tool calling

Browse source
This commit is contained in:
randoentity 2024-11-24 13:40:33 +01:00
parent aa4ccd03d4
commit a52610fb19
1 changed files with 2 additions and 2 deletions
Download patch file Download diff file Expand all files Collapse all files

4
endpoints/OAI/utils/chat_completion.py
View file

@ -454,11 +454,11 @@ async def generate_tool_calls(
if gen["stop_str"] in tool_data.tool_call_start:
if "text" in gen:
# non streaming, all generations will have the text they generated
pre_tool_prompt = await apply_chat_template(data, gen["text"])
pre_tool_prompt, _ = await apply_chat_template(data, gen["text"])
elif current_generations is not None:
# streaming, we wont have text in the generation,
# we'll have to use the current_generations
pre_tool_prompt = await apply_chat_template(data, current_generations)
pre_tool_prompt, _ = await apply_chat_template(data, current_generations)
gen_tasks.append(
asyncio.create_task(