WokeAI/Tankie-LFM2.5-1.2B-SFT-v1

Full fine-tune vs LiquidAI/LFM2.5-1.2B-Instruct. Fine-tune: 100% of the weights changed, across 16 of 16 layers.

Tensors changed: Embeddings 1/2, other 30/30, MLP 48/48, Norms 26/44, Attention 24/24

2 derivatives checked

• mradermacher/Tankie-LFM2.5-1.2B-SFT-v1-i1-GGUF: Quantization
• mradermacher/Tankie-LFM2.5-1.2B-SFT-v1-GGUF: Same weights, converted

On the Hub · Downloads on Model Pulse