ValiantLabs/Qwen3.6-35B-A3B-Esper3.1
LoRA-style vs Qwen/Qwen3.6-35B-A3B. Looks like a merged LoRA (or a fine-tune with frozen norms): only the weight matrices of down_proj, gate_proj, in_proj_qkv, in_proj_z, out_proj, up_proj changed, in 40 of 40 layers; norms, biases and embeddings are untouched.
Tensors changed: Embeddings 0/5, Attention 90/392, MLP 120/399, Norms 0/247, lm_head 0/1, other 0/1
2 derivatives checked
- mradermacher/Qwen3.6-35B-A3B-Esper3.1-i1-GGUF: Quantization
- mradermacher/Qwen3.6-35B-A3B-Esper3.1-GGUF: Quantization