From abb7bf5a4ac2048aac5bc6cd6c93352bc69b6aad Mon Sep 17 00:00:00 2001 From: Bowen Bao Date: Mon, 7 Oct 2024 21:03:57 +0000 Subject: [PATCH 1/2] Tidy quark quantization script in README.md Remove trailing spaces and fix indentation. --- README.md | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) diff --git a/README.md b/README.md index 056ad75..5ff6209 100644 --- a/README.md +++ b/README.md @@ -96,9 +96,9 @@ Similarly, you can download Meta-Llama-3.1-70B and Meta-Llama-3.1-8B. Run the quantization script in the example folder using the following command line: export MODEL_DIR = [local model checkpoint folder] or meta-llama/Meta-Llama-3.1-405B-Instruct #### single GPU - python3 quantize_quark.py \ + python3 quantize_quark.py \ --model_dir $MODEL_DIR \ - --output_dir Meta-Llama-3.1-405B-Instruct-FP8-KV \ + --output_dir Meta-Llama-3.1-405B-Instruct-FP8-KV \ --quant_scheme w_fp8_a_fp8 \ --kv_cache_dtype fp8 \ --num_calib_data 128 \ @@ -106,9 +106,9 @@ export MODEL_DIR = [local model checkpoint folder] or meta-llama/Meta-Llama-3.1- --no_weight_matrix_merge #### If model size is too large for single GPU, please use multi GPU instead. - python3 quantize_quark.py \ + python3 quantize_quark.py \ --model_dir $MODEL_DIR \ - --output_dir Meta-Llama-3.1-405B-Instruct-FP8-KV \ + --output_dir Meta-Llama-3.1-405B-Instruct-FP8-KV \ --quant_scheme w_fp8_a_fp8 \ --kv_cache_dtype fp8 \ --num_calib_data 128 \ From 91af2a92471ec7a84fb92a00712d783cfc6a6a85 Mon Sep 17 00:00:00 2001 From: Bowen Bao Date: Mon, 7 Oct 2024 21:10:56 +0000 Subject: [PATCH 2/2] Fix PyTorch typo --- README.md | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/README.md b/README.md index 5ff6209..2f4739a 100644 --- a/README.md +++ b/README.md @@ -364,8 +364,8 @@ PyTorch 2.4 Release | Component | Commit/Link | | -- | -- | | Base Docker | rocm/pytorch:rocm6.2_ubuntu22.04_py3.9_pytorch_release_2.2.1 | -| Pytorch Commit |[https://github.com/ROCm/pytorch/commit/c4d355377af3b1c48b37e05f81293c65f25689aa](https://github.com/ROCm/pytorch/commit/c4d355377af3b1c48b37e05f81293c65f25689aa "https://github.com/rocm/pytorch/commit/c4d355377af3b1c48b37e05f81293c65f25689aa") -| Pytoch wheels | [http://rocm-ci.amd.com/view/Release-6.2/job/pytorch-pipeline-manylinux-wheel-builder_rel-6.2/315/execution/node/169/ws/final_pkgs/](http://rocm-ci.amd.com/view/Release-6.2/job/pytorch-pipeline-manylinux-wheel-builder_rel-6.2/315/execution/node/169/ws/final_pkgs/ "http://rocm-ci.amd.com/view/release-6.2/job/pytorch-pipeline-manylinux-wheel-builder_rel-6.2/315/execution/node/169/ws/final_pkgs/") +| PyTorch Commit |[https://github.com/ROCm/pytorch/commit/c4d355377af3b1c48b37e05f81293c65f25689aa](https://github.com/ROCm/pytorch/commit/c4d355377af3b1c48b37e05f81293c65f25689aa "https://github.com/rocm/pytorch/commit/c4d355377af3b1c48b37e05f81293c65f25689aa") +| PyTorch wheels | [http://rocm-ci.amd.com/view/Release-6.2/job/pytorch-pipeline-manylinux-wheel-builder_rel-6.2/315/execution/node/169/ws/final_pkgs/](http://rocm-ci.amd.com/view/Release-6.2/job/pytorch-pipeline-manylinux-wheel-builder_rel-6.2/315/execution/node/169/ws/final_pkgs/ "http://rocm-ci.amd.com/view/release-6.2/job/pytorch-pipeline-manylinux-wheel-builder_rel-6.2/315/execution/node/169/ws/final_pkgs/") | apex | [https://github.com/ROCm/apex/commit/ac13eaffb8a3dd8d574979263aa24bce2a5966a4](https://github.com/ROCm/apex/commit/ac13eaffb8a3dd8d574979263aa24bce2a5966a4 "https://github.com/rocm/apex/commit/ac13eaffb8a3dd8d574979263aa24bce2a5966a4") | torchvision |[https://github.com/pytorch/vision/commit/48b1edffdc6f34b766e2b4bbf23b78bd4df94181](https://github.com/pytorch/vision/commit/48b1edffdc6f34b766e2b4bbf23b78bd4df94181 "https://github.com/pytorch/vision/commit/48b1edffdc6f34b766e2b4bbf23b78bd4df94181") | torchdata |[https://github.com/pytorch/data/commit/5e6f7b7dc5f8c8409a6a140f520a045da8700451](https://github.com/pytorch/data/commit/5e6f7b7dc5f8c8409a6a140f520a045da8700451 "https://github.com/pytorch/data/commit/5e6f7b7dc5f8c8409a6a140f520a045da8700451")