HuggingFaceTB/SmolVLM2-2.2B-Instruct
Full fine-tune vs HuggingFaceTB/SmolVLM-Instruct. Fine-tune: 100% of the weights changed, across 27 of 27 layers.
Tensors changed: other 1/1, Norms 159/159, MLP 180/180, Attention 312/312, Embeddings 3/3
13 derivatives checked
- ggml-org/SmolVLM2-2.2B-Instruct-GGUF: Same weights, converted (fewer layers)
- second-state/SmolVLM2-2.2B-Instruct-GGUF: Same weights, converted (fewer layers)
- mradermacher/SmolVLM2-2.2B-Instruct-i1-GGUF: Quantization (fewer layers)
- mlx-community/SmolVLM2-2.2B-Instruct-mlx: Same weights, converted
- mradermacher/SmolVLM2-2.2B-Instruct-GGUF: Same weights, converted (fewer layers)
- gaianet/SmolVLM2-2.2B-Instruct-GGUF: Same weights, converted (fewer layers)
- Chopinab/SmolVLM2-2.2B-Instruct-IQ3_M-GGUF: Quantization (fewer layers)
- lewiswatson/Frame2KG-SmolVLM-2.2b-JSON: LoRA-style (fewer layers)
- Chopinab/SmolVLM2-2.2B-Instruct-Q4_K_M-GGUF: Quantization (fewer layers)
- Chopinab/SmolVLM2-2.2B-Instruct-Q5_K_M-GGUF: Quantization (fewer layers)
- richardyoung/SmolVLM2-2.2B-Instruct-GGUF: Quantization (fewer layers)
- Chopinab/SmolVLM2-2.2B-Instruct-Q8_0-GGUF: Quantization (fewer layers)
- Luigi/edge-fall-vlm-2.2b: LoRA-style
1 other repos hold byte-for-byte identical weight files.