NandemoGHS/Anime-Llasa-3B

Full fine-tune vs HKUSTAudio/Llasa-3B. Fine-tune: 100% of the weights changed, across 28 of 28 layers.

Tensors changed: Embeddings 1/1, Norms 27/57, MLP 84/84, Attention 112/112

3 derivatives checked

• mradermacher/Anime-Llasa-3B-i1-GGUF: Quantization
• mradermacher/Anime-Llasa-3B-GGUF: Same weights, converted
• furusu/Anime-Llasa-3B-GGUF-Fix-Special: Quantization

On the Hub · Downloads on Model Pulse