abhinand/gemma-2b-tamil
LoRA-style vs google/gemma-2b. 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 18 of 18 layers; norms, biases and embeddings are untouched.
Tensors changed: Embeddings 0/1, Norms 0/37, MLP 54/54, Attention 72/72
5 derivatives checked
- abhinand/gemma-2b-it-tamil-v0.1-alpha: LoRA-style
- featherless-ai-quants/abhinand-gemma-2b-tamil-GGUF: Quantization
- abhinand/gemma-2b-it-tamil-v0.1-alpha-GGUF: LoRA-style
- mradermacher/gemma-2b-tamil-GGUF: Same weights, converted
- CryptoYogi/gemma-2b-tamil-base: LoRA-style