mradermacher/sft-conta-qwen-7b-grpo-step64-GGUF

Same weights, converted vs LRM-Conta-Detection-Arena/sft-conta-qwen2.5-7b-grpo-step64. Same weights stored in another dtype (BF16 → F16/F32).

Tensors changed: lm_head 0/1, Embeddings 0/1, MLP 0/84, Attention 0/196, Norms 0/57

On the Hub · Downloads on Model Pulse