HuggingFaceTB/SmolVLM2-500M-Video-Instruct
Full fine-tune vs HuggingFaceTB/SmolVLM-500M-Instruct. Fine-tune: 100% of the weights changed, across 32 of 32 layers.
Tensors changed: lm_head 1/1, other 1/1, Embeddings 4/4, Norms 101/115, MLP 144/144, Attention 224/224
13 derivatives checked
- lerobot/smolvla_base: Different architecture
- ggml-org/SmolVLM2-500M-Video-Instruct-GGUF: Same weights, converted (fewer layers)
- second-state/SmolVLM2-500M-Video-Instruct-GGUF: Same weights, converted (fewer layers)
- mradermacher/SmolVLM2-500M-Video-Instruct-i1-GGUF: Quantization (fewer layers)
- mradermacher/SmolVLM2-500M-Video-Instruct-GGUF: Same weights, converted (fewer layers)
- jc-builds/smolvlm2-500m-gguf: Quantization (fewer layers)
- ataeff/molequla: Quantization (fewer layers)
- gaianet/SmolVLM2-500M-Video-Instruct-GGUF: Same weights, converted (fewer layers)
- DevQuasar/HuggingFaceTB.SmolVLM2-500M-Video-Instruct-GGUF: Quantization (fewer layers)
- lewiswatson/Frame2KG-SmolVLM-500m-JSON: LoRA-style (fewer layers)
- kangjirm/SmolVLM2-500M-Video-Instruct-Chakra-Q8_0.gguf: Full fine-tune (fewer layers)
- vocaela/Vocaela-500M: Full fine-tune
- vocaela/Vocaela-2-500M-1024R2: Full fine-tune
6 other repos hold byte-for-byte identical weight files.