nvidia/LocateAnything-3B
Full fine-tune (fewer layers) vs Qwen/Qwen2.5-3B-Instruct. Qwen/Qwen2.5-3B-Instruct matches part of nvidia/LocateAnything-3B (433 of its 770 tensors; the rest of nvidia/LocateAnything-3B has no counterpart). On the shared part: Fine-tune: 100% of the weights changed, across 36 of 36 layers.
Tensors changed: Norms 73/73, MLP 108/108, Attention 252/252
12 derivatives checked
- mudler/locate-anything.cpp-gguf: Not comparable yet
- yuuko-eth/LocateAnything-3B-GGUF: Same weights, converted (fewer layers)
- liodon-ai/LocateAnything-3B-imatrix-GGUF: Quantization (fewer layers)
- sabafallah/LocateAnything-3B-GGUF: Same weights, converted (fewer layers)
- mlx-community/LocateAnything-3B-4bit: Quantization
- ShahzebKhoso/LocateAnything-3B-GGUF: Same weights, converted (fewer layers)
- DhruvalLabs/LocateAnything-3B-GGUF: Quantization (fewer layers)
- RKNNAI/RK182X-VLM-LocateAnything-3B: Quantization
- mlx-community/LocateAnything-3B-8bit: Quantization
- SandLogicTechnologies/LocateAnything-3B-GGUF: Quantization (fewer layers)
- groxaxo/LocateAnything-3B-AutoRound-W4A16: Quantization
- andai-labs/LocateAnything-3B-MLX: Same weights, repacked
33 other repos hold byte-for-byte identical weight files.