SicariusSicariiStuff/Impish_QWEN_7B-1M
Partial fine-tune vs Qwen/Qwen2.5-7B-Instruct-1M. Partial change: 86% of the weights changed (attention, mlp, norms), in 28 of 28 layers.
Tensors changed: Embeddings 0/1, Norms 6/57, MLP 84/84, Attention 115/196, lm_head 0/1
4 derivatives checked
- BasedAGI/Impish_QWEN_7B-1M-i1-GGUF: Quantization
- mradermacher/Impish_QWEN_7B-1M-i1-GGUF: Quantization
- mradermacher/Impish_QWEN_7B-1M-GGUF: Same weights, converted
- SicariusSicariiStuff/Impish_QWEN_7B-1M_GGUF: Quantization