altomek/CodeRosa-70B-AB1
Full fine-tune vs codellama/CodeLlama-70b-Python-hf. Fine-tune: 100% of the weights changed, across 80 of 80 layers.
Tensors changed: Norms 161/161, MLP 240/240, Attention 320/320
2 derivatives checked
- mradermacher/CodeRosa-70B-AB1-GGUF: Quantization
- altomek/CodeRosa-70B-AB1-GGUF: Quantization