LiquidAI/LFM2.5-VL-450M
Full fine-tune (fewer layers) vs LiquidAI/LFM2.5-350M. LiquidAI/LFM2.5-350M matches part of LiquidAI/LFM2.5-VL-450M (148 of its 349 tensors; the rest of LiquidAI/LFM2.5-VL-450M 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
18 derivatives checked
- LiquidAI/LFM2.5-VL-450M-GGUF: Same weights, converted (fewer layers)
- mlx-community/LFM2.5-VL-450M-6bit: Quantization
- LiquidAI/LFM2.5-VL-450M-MLX-4bit: Quantization
- NobodyWho/LFM2.5-VL-450M-GGUF: Same weights, converted (fewer layers)
- LiquidAI/LFM2.5-VL-450M-Extract: Full fine-tune
- erikku-sama/lfm2-electronics-vlm-v2: LoRA-style
- Varun12387/LFM2.5-VL-450M-SFT-GGUF: LoRA-style
- LiquidAI/LFM2.5-VL-450M-MLX-8bit: Quantization
- LiquidAI/LFM2.5-VL-450M-MLX-bf16: Different architecture
- 5ch4um1/lfm2.5-vrsbench-lora-450m: Partial fine-tune
- lewiswatson/Frame2KG-LFM-2.5-450m-JSON: Partial fine-tune (fewer layers)
- 5ch4um1/lfm2.5-vrsbench-EUROSAT-terrain-lora-450m: Partial fine-tune
- 5ch4um1/lfm2.5-vrsbench-MADOS-maritime-lora-450m: Partial fine-tune
- webbrain-one/webbrain-vl-2-450M: Partial fine-tune
- LiquidAI/LFM2.5-VL-450M-MLX-6bit: Quantization
- nathanrchn/coffee-leaf-lfm2.5-vl-450m-GGUF: Full fine-tune (fewer layers)
- LiquidAI/LFM2.5-VL-450M-MLX-5bit: Quantization
- mlx-community/LFM2.5-VL-450M-8bit: Quantization
4 other repos hold byte-for-byte identical weight files.