win10/WhiteRabbitNeo-2.5-Qwen-2.5-Coder-12.3B
Partial fine-tune (fewer layers) vs WhiteRabbitNeo/WhiteRabbitNeo-2.5-Qwen-2.5-Coder-7B. WhiteRabbitNeo/WhiteRabbitNeo-2.5-Qwen-2.5-Coder-7B matches part of win10/WhiteRabbitNeo-2.5-Qwen-2.5-Coder-12.3B (339 of its 579 tensors; the rest of win10/WhiteRabbitNeo-2.5-Qwen-2.5-Coder-12.3B has no counterpart). On the shared part: Partial change: 61% of the weights changed (attention, mlp, norms), in 20 of 28 layers.
Tensors changed: Embeddings 0/1, Norms 40/57, MLP 60/84, Attention 140/196, lm_head 0/1
3 derivatives checked
- Fmuaddib/WhiteRabbitNeo-2.5-Qwen-2.5-Coder-12.3B-mlx-fp16: Same weights, converted
- mradermacher/WhiteRabbitNeo-2.5-Qwen-2.5-Coder-12.3B-i1-GGUF: Quantization
- mradermacher/WhiteRabbitNeo-2.5-Qwen-2.5-Coder-12.3B-GGUF: Quantization