From 02365e8e2774df0319720fdb4ccdc34921537d0f Mon Sep 17 00:00:00 2001 From: extio-de Date: Sun, 20 Oct 2024 21:41:40 +0200 Subject: [PATCH] Readme --- README.md | 14 ++++++++++++++ 1 file changed, 14 insertions(+) diff --git a/README.md b/README.md index 4cf5512..5262471 100644 --- a/README.md +++ b/README.md @@ -1 +1,15 @@ # lm-container + +**Program Name:** Custom ROCm and PyTorch Docker Image + +**Purpose:** The program builds a custom Docker image for running PyTorch on an AMD GPU using ROCm (Radeon Open Compute). The image is designed to support high-performance deep learning workloads, specifically with the LLaMA transformer model. + +**Key Features:** + +* Installs necessary packages and sets environment variables for ROCm and PyTorch +* Clones and builds the PyTorch backend for ROCm (bitsandbytes) and installs it as a package +* Installs additional packages for transformer models (trl, peft, transformers, datasets, and huggingface-hub) +* Clones and builds the LLaMA transformer model (llama.cpp) with ROCm support +* Copies and makes executable two scripts: convert_to_gguf.sh and setup.sh + +**Functionality:** The Docker image can be used to run PyTorch workloads on an AMD GPU, leveraging the performance of ROCm. The setup.sh script, which is run when the container is started, sets up the environment and prepares the image for use.