mlabonne/BigQwen2.5-Echo-47B-Instruct
Partial fine-tune (fewer layers) vs Qwen/Qwen2.5-32B-Instruct. Qwen/Qwen2.5-32B-Instruct matches part of mlabonne/BigQwen2.5-Echo-47B-Instruct (771 of its 1131 tensors; the rest of mlabonne/BigQwen2.5-Echo-47B-Instruct has no counterpart). On the shared part: Partial change: 82% of the weights changed (attention, mlp, norms), in 55 of 64 layers.
Tensors changed: Embeddings 0/1, Norms 110/129, MLP 165/192, Attention 385/448, lm_head 0/1
3 derivatives checked
- bartowski/BigQwen2.5-Echo-47B-Instruct-GGUF: Same weights, converted
- mradermacher/BigQwen2.5-Echo-47B-Instruct-i1-GGUF: Quantization
- mradermacher/BigQwen2.5-Echo-47B-Instruct-GGUF: Quantization
1 other repos hold byte-for-byte identical weight files.