Open aimzieslol opened 5 months ago
👋 @aimzieslol! Thanks for reporting here. Holy stack trace, haha.
Yeah it definitely looks like the loop to hydrate tasks / call it quits for your flow sputtered out. I'll take a look to reproduce thanks 🙇
Same issue here (agent never terminates):
import controlflow as cf
from langchain_openai import ChatOpenAI
model = ChatOpenAI(base_url="http://192.168.170.42:8000/v1", model="microsoft/Phi-3-small-8k-instruct", temperature=0.1)
agent = cf.Agent(model=model)
task = cf.Task(
objective=f"Find interesting AI usages in healthcare.",
result_type=str,
agents=[agent]
)
task.run()
The model is hosted locally using VLLM.
Hey @aimzieslol and @mpmX! Haven't been able to reproduce yet, but hope is #203 might ameliorate this.
@mpmX is this with latest release?
@aaazzam Seems to be working OK. I'll mess around w/it some more. Thanks!
@mpmX is this with latest release? Yes, latest release.
0.8.2
Here's the entire Jupyter notebook:
The output is (forgive the formatting):
Had I not defined a max iterations, it seems like it would have run forever.
Am I missing a step or something?