BSC-LT/ALIA-40b-instruct-2606

Full fine-tune vs BSC-LT/ALIA-40b. Fine-tune: 100% of the weights changed, across 48 of 48 layers.

Tensors changed: Embeddings 1/1, MLP 144/144, Attention 192/192, Norms 97/97, lm_head 1/1

3 derivatives checked

• mradermacher/ALIA-40b-instruct-2606-i1-GGUF: Quantization
• mradermacher/ALIA-40b-instruct-2606-GGUF: Quantization
• BSC-LT/ALIA-40b-instruct-2606-GGUF: Quantization

On the Hub · Downloads on Model Pulse