meshllm/Qwen3.8-2.4T-A95B-UD-IQ4_XS-layers
Full fine-tune (fewer layers) vs unsloth/Qwen3.8-2.4T-A95B-GGUF. meshllm/Qwen3.8-2.4T-A95B-UD-IQ4_XS-layers matches part of unsloth/Qwen3.8-2.4T-A95B-GGUF (18 of its 1702 tensors; the rest of unsloth/Qwen3.8-2.4T-A95B-GGUF has no counterpart). On the shared part: meshllm/Qwen3.8-2.4T-A95B-UD-IQ4_XS-layers is a GGUF (IQ4_XS/Q5_K) of different weights. Fine-tune: 100% of the weights changed, across 0 of 0 layers.
Tensors changed: MLP 6/7, Norms 0/3, other 1/6, Attention 2/2