HuggingFaceTB/SmolVLM-Instruct
Full fine-tune (fewer layers) vs HuggingFaceTB/SmolLM2-1.7B-Instruct. HuggingFaceTB/SmolLM2-1.7B-Instruct matches part of HuggingFaceTB/SmolVLM-Instruct (217 of its 657 tensors; the rest of HuggingFaceTB/SmolVLM-Instruct has no counterpart). On the shared part: Fine-tune: 100% of the weights changed, across 24 of 24 layers.
Tensors changed: Norms 49/49, MLP 72/72, Attention 96/96
7 derivatives checked
- HuggingFaceTB/SmolVLM2-2.2B-Instruct: Full fine-tune
- ggml-org/SmolVLM-Instruct-GGUF: Same weights, converted (fewer layers)
- mradermacher/SmolVLM-Instruct-i1-GGUF: Quantization (fewer layers)
- mlx-community/SmolVLM-Instruct-4bit: Quantization
- mradermacher/SmolVLM-Instruct-GGUF: Same weights, converted (fewer layers)
- smdesai/SmolVLM2-2.2B-Instruct-4bit: Quantization
- vidore/ColSmolVLM-base: Same weights, converted
7 other repos hold byte-for-byte identical weight files.