coder3101/LFM2.5-VL-450M-heretic
Full fine-tune (fewer layers) vs LiquidAI/LFM2.5-350M. LiquidAI/LFM2.5-350M matches part of coder3101/LFM2.5-VL-450M-heretic (148 of its 349 tensors; the rest of coder3101/LFM2.5-VL-450M-heretic has no counterpart). On the shared part: Fine-tune: 100% of the weights changed, across 16 of 16 layers.
Tensors changed: Embeddings 2/2, other 30/30, MLP 48/48, Norms 44/44, Attention 24/24
2 derivatives checked
- FadedRedStar/LFM2.5-VL-450M-heretic-imatrix-GGUF: Quantization (fewer layers)
- FadedRedStar/LFM2.5-VL-450M-heretic-GGUF: Quantization (fewer layers)
1 other repos hold byte-for-byte identical weight files.