BenevolenceMessiah/Qwen2.5-Coder-7B-3x-Instruct-TIES-v1.0
Full fine-tune vs Etherll/Qwen2.5-Coder-7B-Instruct-Ties. Fine-tune: 100% of the weights changed, across 28 of 28 layers.
Tensors changed: Norms 15/57, MLP 84/84, Attention 170/196
2 derivatives checked
- mradermacher/Qwen2.5-Coder-7B-3x-Instruct-TIES-v1.0-i1-GGUF: Quantization
- mradermacher/Qwen2.5-Coder-7B-3x-Instruct-TIES-v1.0-GGUF: Same weights, converted
2 other repos hold byte-for-byte identical weight files.