From 04ed74785ceb743b6d3fed724661e10b39e49ced Mon Sep 17 00:00:00 2001 From: GraoMelo <60477737+GraoMelo@users.noreply.github.com> Date: Thu, 2 May 2024 18:17:33 -0300 Subject: [PATCH] Update README.md add requirments JAX CUDA fixed #328 requiriments por install. --- README.md | 9 +++++++++ 1 file changed, 9 insertions(+) diff --git a/README.md b/README.md index f501a07..34e929f 100644 --- a/README.md +++ b/README.md @@ -18,6 +18,15 @@ The script loads the checkpoint and samples from the model on a test input. Due to the large size of the model (314B parameters), a machine with enough GPU memory is required to test the model with the example code. The implementation of the MoE layer in this repository is not efficient. The implementation was chosen to avoid the need for custom kernels to validate the correctness of the model. + +observation: the NVIDIA CUDA dependencies are only for linux: + +``` +jax[cuda12-pip]==0.4.25 -f https://.../jax-releases/jax_cuda_releases.html +``` + +Please check which hardware is compatible with: [Link]([URL](https://docs.nvidia.com/deploy/cuda-compatibility/index.html)) + # Model Specifications Grok-1 is currently designed with the following specifications: