Summary
When running ml-intern headless from CLI, the run can appear to hang for ~5 minutes (or longer) after a hub_repo_details tool call with no additional progress output.
Environment
- Repo:
huggingface/ml-intern
- Invocation: local CLI (
ml-intern \"...\"), not web backend
- Model shown by CLI:
anthropic/claude-opus-4-6
- Max iterations shown by CLI:
300
- HF token is loaded; Anthropic key is set in env
- Observed on macOS (Darwin)
Reproduction
Run:
ml-intern "Fine-tune a small Whisper model for Arabic speech recognition on a GPU under 5 dollars and compare the fine-tuned model's metrics to the original model."
Observed output reaches:
▸ plan_tool {"todos": [...]}
▸ hub_repo_search {"query": "arabic speech recognition whisper", ...}
▸ hub_repo_details {"repo_ids": ["openai/whisper-small"]}
Then no new logs/events are printed for about 5 minutes. I manually interrupted with Ctrl+C, which produced:
^CWARNING:agent.core.agent_loop:Agent loop cancelled
Goodbye!
Expected behavior
After hub_repo_details, there should either be:
- continued tool/model progress logs, or
- a timeout/error message indicating what call is blocked.
Actual behavior
No output appears for several minutes, so it looks stuck even though the process is still running presumably?
Summary
When running
ml-internheadless from CLI, the run can appear to hang for ~5 minutes (or longer) after ahub_repo_detailstool call with no additional progress output.Environment
huggingface/ml-internml-intern \"...\"), not web backendanthropic/claude-opus-4-6300Reproduction
Run:
ml-intern "Fine-tune a small Whisper model for Arabic speech recognition on a GPU under 5 dollars and compare the fine-tuned model's metrics to the original model."Observed output reaches:
Then no new logs/events are printed for about 5 minutes. I manually interrupted with
Ctrl+C, which produced:Expected behavior
After
hub_repo_details, there should either be:Actual behavior
No output appears for several minutes, so it looks stuck even though the process is still running presumably?