TheDrummer/Artemis-31B-v1.2
LoRA-style vs google/gemma-4-31B-it. 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 30 of 60 layers; norms, biases and embeddings are untouched.
Tensors changed: Embeddings 0/4, Norms 0/523, other 0/62, MLP 90/261, Attention 115/338
10 derivatives checked
- bartowski/TheDrummer_Artemis-31B-v1.2-GGUF: Same weights, converted (fewer layers)
- mradermacher/Artemis-31B-v1.2-i1-GGUF: Quantization (fewer layers)
- Abiray/Artemis-31B-v1.2-GGUF: Quantization (fewer layers)
- mradermacher/Artemis-31B-v1.2-GGUF: Quantization (fewer layers)
- FaustianDeal/Artemis-31B-v1.2-BF16-embd-GGUF: Quantization (fewer layers)
- FaustianDeal/Artemis-31B-v1.2-NVFP4-GGUF: Quantization
- ailexleon/Artemis-31B-v1.2-mlx-8Bit: Quantization
- ailexleon/Artemis-31B-v1.2-mlx-4Bit: Quantization
- ailexleon/Artemis-31B-v1.2-mlx-6Bit: Quantization
- DeusImperator/Artemis-31B-v1.2-GGUF-long-ctx: Quantization (fewer layers)