Remove max_new_tokens and add default tool_call_parser, reasoning_parser, enable_thinking values c226526 verified tapu1125 commited on 4 days ago
spinquantless FP8 (no SpinQuant rotation), 256K context - fixes agentic looping 17cacdc verified joerowell commited on 12 days ago
Enable thinking by default, preserve reasoning; drop max_new_tokens cap 610e625 verified joerowell commited on 13 days ago