JoaoGuiAlves/thagiPo-llama3.2-1B-math
Full fine-tune vs meta-llama/Llama-3.2-1B. Fine-tune: 100% of the weights changed, across 16 of 16 layers.
Tensors changed: Embeddings 1/1, Norms 17/33, MLP 48/48, Attention 64/64
Full fine-tune vs meta-llama/Llama-3.2-1B. Fine-tune: 100% of the weights changed, across 16 of 16 layers.
Tensors changed: Embeddings 1/1, Norms 17/33, MLP 48/48, Attention 64/64