TIGER-Lab/Qwen2.5-32B-Instruct-CFT
Full fine-tune vs Qwen/Qwen2.5-32B-Instruct. Fine-tune: 95% of the weights changed, across 64 of 64 layers.
Tensors changed: Embeddings 0/1, Norms 119/129, MLP 192/192, Attention 431/448, lm_head 0/1
4 derivatives checked
- bartowski/TIGER-Lab_Qwen2.5-32B-Instruct-CFT-GGUF: Quantization
- mradermacher/Qwen2.5-32B-Instruct-CFT-i1-GGUF: Quantization
- mradermacher/Qwen2.5-32B-Instruct-CFT-GGUF: Quantization
- tensorblock/Qwen2.5-32B-Instruct-CFT-GGUF: Quantization
1 other repos hold byte-for-byte identical weight files.