nbeerbower/Dumpling-Qwen2.5-7B-1k-r64-2e-5
LoRA-style vs nbeerbower/EVA-abliterated-Qwen2.5-7B. Looks like a merged LoRA (or a fine-tune with frozen norms): only the weight matrices of down_proj, gate_proj, k_proj, o_proj, q_proj, up_proj, v_proj changed, in 28 of 28 layers; norms, biases and embeddings are untouched.
Tensors changed: lm_head 0/1, Embeddings 0/1, Norms 0/57, MLP 84/84, Attention 112/196
2 derivatives checked
- mradermacher/Dumpling-Qwen2.5-7B-1k-r64-2e-5-i1-GGUF: Quantization
- mradermacher/Dumpling-Qwen2.5-7B-1k-r64-2e-5-GGUF: Same weights, converted