apus-ailab/APUS-OpenJev-v1-4B
LoRA-style vs Qwen/Qwen3.5-4B. Looks like a merged LoRA (or a fine-tune with frozen norms): only the weight matrices of down_proj, gate_proj, in_proj_qkv, k_proj, o_proj, out_proj, q_proj, up_proj, v_proj changed, in 32 of 32 layers; norms, biases and embeddings are untouched.
Tensors changed: Embeddings 0/4, MLP 96/196, Attention 80/320, Norms 0/203
4 derivatives checked
- apus-ailab/APUS-OpenJev-v1-4B-GGUF: Same weights, converted
- prithivMLmods/APUS-OpenJev-v1-4B-GGUF: Same weights, converted
- apus-ailab/APUS-OpenJev-v1-4B-MLX-8bit: Quantization
- apus-ailab/APUS-OpenJev-v1-4B-MLX-4bit: Quantization