[Kernel] Triton Configs for Fp8 Block Quantization (#11589)

Signed-off-by: rshaw@neuralmagic.com <rshaw@neuralmagic.com> Signed-off-by: mgoin <michael@neuralmagic.com> Co-authored-by: mgoin <michael@neuralmagic.com> Co-authored-by: simon-mo <xmo@berkeley.edu>

[Kernel] Triton Configs for Fp8 Block Quantization (#11589)
Signed-off-by: rshaw@neuralmagic.com <rshaw@neuralmagic.com> Signed-off-by: mgoin <michael@neuralmagic.com> Co-authored-by: mgoin <michael@neuralmagic.com> Co-authored-by: simon-mo <xmo@berkeley.edu>
9b0c4bab · Robert Shaw · GitHub · 41bf5612 · 9b0c4bab · 9b0c4bab
Unverified Commit 9b0c4bab authored Jan 30, 2025 by Robert Shaw Committed by GitHub Jan 30, 2025
3 changed files
--- a/vllm/model_executor/layers/quantization/utils/configs/N=7168,K=2304,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json
+++ b/vllm/model_executor/layers/quantization/utils/configs/N=7168,K=2304,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json
--- a/vllm/model_executor/layers/quantization/utils/configs/N=7168,K=256,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json
+++ b/vllm/model_executor/layers/quantization/utils/configs/N=7168,K=256,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json
--- a/vllm/model_executor/layers/quantization/utils/fp8_utils.py
+++ b/vllm/model_executor/layers/quantization/utils/fp8_utils.py