GuminiResearch/Gumini-1B-Base
Full fine-tune (fewer layers) vs Qwen/Qwen2.5-3B. GuminiResearch/Gumini-1B-Base matches part of Qwen/Qwen2.5-3B (122 of its 434 tensors; the rest of Qwen/Qwen2.5-3B has no counterpart). On the shared part: Fine-tune: 100% of the weights changed, across 10 of 10 layers.
Tensors changed: Embeddings 1/1, Norms 19/21, MLP 30/30, Attention 70/70
3 derivatives checked
- GuminiResearch/Gumini-1B-Base-i1-GGUF: Same weights, converted
- mradermacher/Gumini-1B-Base-inherit-i1-GGUF: Quantization
- mradermacher/Gumini-1B-Base-inherit-GGUF: Same weights, converted