RedHatAI/Llama-3.1-8B-Instruct-speculator.eagle3
Full fine-tune vs meta-llama/Llama-3.1-8B-Instruct. Fine-tune: 100% of the weights changed, across 1 of 1 layers.
Tensors changed: Embeddings 1/1, Norms 3/3, MLP 3/3, Attention 1/1
Full fine-tune vs meta-llama/Llama-3.1-8B-Instruct. Fine-tune: 100% of the weights changed, across 1 of 1 layers.
Tensors changed: Embeddings 1/1, Norms 3/3, MLP 3/3, Attention 1/1