dphn/dolphin-2.9.4-llama3.1-8b
LoRA-style vs meta-llama/Llama-3.1-8B. 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 32 of 32 layers; norms, biases and embeddings are untouched.
Tensors changed: Norms 0/65, MLP 48/96, Attention 64/128
12 derivatives checked
- bartowski/dolphin-2.9.4-llama3.1-8b-GGUF: Quantization
- mradermacher/dolphin-2.9.4-llama3.1-8b-i1-GGUF: Quantization
- yaneivan/dolphin-2.9.4-llama3.1-8b-Q4_K_M-GGUF: Quantization
- mradermacher/dolphin-2.9.4-llama3.1-8b-GGUF: Same weights, converted
- QuantFactory/Lama-DPOlphin-8B-GGUF: Quantization
- solidrust/dolphin-2.9.4-llama3.1-8b-AWQ: Quantization
- NeoChen1024/dolphin-2.9.4-llama3.1-8b-GGUF: Same weights, converted
- MaiDuong/Llama3.1_Uncensored: LoRA-style
- itlwas/dolphin-2.9.4-llama3.1-8b-Q4_K_M-GGUF: Quantization
- mlx-community/dolphin-2.9.4-llama3.1-8b-8bit: Quantization
- paradox0bytes/dolphin-custom-trainedv2: Full fine-tune
- ALIOSAM254/Dolphin-Arabic-Final-F16: LoRA-style
1 other repos hold byte-for-byte identical weight files.