Preserve MTP (Multi Token Prediction) tensor heads for Qwen3.6-35B-A3B-Claude-4.7-Opus-Reasoning-Distilled-oQ8e
#1
by egodoyca - opened
Hi, please ensure the MTP (Multi-Token Prediction) tensor heads are preserved in the (oQe) quants for Qwen3.6-35B-A3B-Claude-4.7-Opus-Reasoning-Distilled-oQ8e:
Base model link: https://huggingface.co/splats/Qwen3.6-35B-A3B-Claude-4.7-Opus-Reasoning-Distilled-oQ8e
Quantization Format: BF16 (source weights)
Preferred Tier: oQ8e
Thank you!
egodoyca changed discussion title from Preserve MTP (Multi Token Prediction) tensor heads for Qwen3.6-27B-oQe to Preserve MTP (Multi Token Prediction) tensor heads for Qwen3.6-35B-A3B-Claude-4.7-Opus-Reasoning-Distilled-oQ8e