OpenGVLab/InternVL2-1B
Full fine-tune (fewer layers) vs OpenGVLab/InternViT-300M-448px. OpenGVLab/InternViT-300M-448px matches part of OpenGVLab/InternVL2-1B (340 of its 637 tensors; the rest of OpenGVLab/InternVL2-1B has no counterpart). On the shared part: Fine-tune: 100% of the weights changed, across 24 of 24 layers.
Tensors changed: Embeddings 4/4, Attention 96/96, other 48/48, MLP 96/96, Norms 96/96
4 derivatives checked
- mradermacher/InternVL2-1B-i1-GGUF: Quantization (fewer layers)
- mradermacher/InternVL2-1B-GGUF: Same weights, converted (fewer layers)
- 5CD-AI/Vintern-1B-v2: Partial fine-tune
- cstr/internvl2-1b-crispembed-GGUF: Not comparable yet
4 other repos hold byte-for-byte identical weight files.