We simply applied the methodology from the paper titled "TransMLA: MLA Is All You Need" to convert the attention mechanism within the Llama-3 model from Grouped Query Attention (GQA) to Multi-Head Latent Attention (MLA), Kinda.

Downloads last month
629
Safetensors
Model size
8B params
Tensor type
BF16
·
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for ForSureTesterSim/Llama-3.1-8B-TransMLA-PreRecovery

Finetuned
(2970)
this model