@tduncan / Nvidia-Optimization-Stack-VLLM Mirror · read-only / commits / 5dfc5abe94
Fork 0

A high-throughput and memory-efficient inference and serving engine for LLMs

Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git

Synced Nov 7, 10:15 UTC

[ROCm] [Release] Change the package from `aiter` to `amd-aiter` (#35198)

Signed-off-by: tjtanaa <tunjian.tan@embeddedllm.com>

TJian committed Mar 3, 2026 at 15:13 UTC 5dfc5abe94fd8a40cd5d93cb5e7b49479ff3657e
2 files changed +3 -3
.buildkite/scripts/annotate-rocm-release.sh
+2 -2
@@ -68,7 +68,7 @@ aws s3 cp s3://${S3_BUCKET}/rocm/${BUILDKITE_COMMIT}/${ROCM_VERSION_PATH}/triton
68 aws s3 cp s3://${S3_BUCKET}/rocm/${BUILDKITE_COMMIT}/${ROCM_VERSION_PATH}/torchvision-*.whl .
69 aws s3 cp s3://${S3_BUCKET}/rocm/${BUILDKITE_COMMIT}/${ROCM_VERSION_PATH}/torchaudio-*.whl .
70 aws s3 cp s3://${S3_BUCKET}/rocm/${BUILDKITE_COMMIT}/${ROCM_VERSION_PATH}/amdsmi-*.whl .
71 -aws s3 cp s3://${S3_BUCKET}/rocm/${BUILDKITE_COMMIT}/${ROCM_VERSION_PATH}/aiter-*.whl .
71 +aws s3 cp s3://${S3_BUCKET}/rocm/${BUILDKITE_COMMIT}/${ROCM_VERSION_PATH}/amd_aiter-*.whl .
72 aws s3 cp s3://${S3_BUCKET}/rocm/${BUILDKITE_COMMIT}/${ROCM_VERSION_PATH}/flash-attn-*.whl .
73 \`\`\`
74
@@ -80,7 +80,7 @@ aws s3 cp s3://${S3_BUCKET}/rocm/${BUILDKITE_COMMIT}/${ROCM_VERSION_PATH}/flash-
80 - **torchvision**: TorchVision for ROCm PyTorch
81 - **torchaudio**: Torchaudio for ROCm PyTorch
82 - **amdsmi**: AMD SMI Python bindings
83 -- **aiter**: Aiter for ROCm
83 +- **amd_aiter**: Aiter for ROCm
84 - **flash-attn**: Flash Attention for ROCm
85
86 ### :warning: Notes
tools/vllm-rocm/pin_rocm_dependencies.py
+1 -1
@@ -64,7 +64,7 @@ def get_custom_wheel_versions(install_dir: str) -> dict[str, str]:
64 ("torchaudio-", "torchaudio"), # Match torchaudio-
65 ("amdsmi-", "amdsmi"), # Match amdsmi-
66 ("flash_attn-", "flash-attn"), # Match flash_attn-
67 - ("aiter-", "aiter"), # Match aiter-
67 + ("amd_aiter-", "amd-aiter"), # Match amd_aiter-
68 ]
69
70 for wheel_file in install_path.glob("*.whl"):