Merge pull request #340 from mit-han-lab/dev

feat: support PuLID, Double FBCache and TeaCache; better linter

Merge pull request #340 from mit-han-lab/dev
feat: support PuLID, Double FBCache and TeaCache; better linter
37a27712 · Muyang Li · GitHub · c1d6fc84 · 760ab022 · 37a27712
Unverified Commit 37a27712 authored May 01, 2025 by Muyang Li Committed by GitHub May 01, 2025
20 changed files
--- a/src/kernels/zgemm/gemm_w4a4_launch_bf16_fp4.cu
+++ b/src/kernels/zgemm/gemm_w4a4_launch_bf16_fp4.cu
--- a/src/kernels/zgemm/gemm_w4a4_launch_bf16_int4.cu
+++ b/src/kernels/zgemm/gemm_w4a4_launch_bf16_int4.cu
--- a/src/kernels/zgemm/gemm_w4a4_launch_fp16_fp4.cu
+++ b/src/kernels/zgemm/gemm_w4a4_launch_fp16_fp4.cu
--- a/src/kernels/zgemm/gemm_w4a4_launch_fp16_int4.cu
+++ b/src/kernels/zgemm/gemm_w4a4_launch_fp16_int4.cu
--- a/src/kernels/zgemm/gemm_w4a4_launch_fp16_int4_fasteri2f.cu
+++ b/src/kernels/zgemm/gemm_w4a4_launch_fp16_int4_fasteri2f.cu
--- a/src/kernels/zgemm/gemm_w4a4_launch_impl.cuh
+++ b/src/kernels/zgemm/gemm_w4a4_launch_impl.cuh
--- a/src/kernels/zgemm/gemm_w4a4_test.cu
+++ b/src/kernels/zgemm/gemm_w4a4_test.cu
--- a/src/kernels/zgemm/gemm_w8a8.cu
+++ b/src/kernels/zgemm/gemm_w8a8.cu
--- a/src/kernels/zgemm/gemm_w8a8.cuh
+++ b/src/kernels/zgemm/gemm_w8a8.cuh
--- a/src/kernels/zgemm/lora.cuh
+++ b/src/kernels/zgemm/lora.cuh
--- a/src/kernels/zgemm/mma.cuh
+++ b/src/kernels/zgemm/mma.cuh
--- a/src/kernels/zgemm/mma_earlycuda.cuh
+++ b/src/kernels/zgemm/mma_earlycuda.cuh
--- a/src/kernels/zgemm/zgemm.h
+++ b/src/kernels/zgemm/zgemm.h
--- a/src/layernorm.cpp
+++ b/src/layernorm.cpp
--- a/src/layernorm.h
+++ b/src/layernorm.h
--- a/src/pytorch_compat.h
+++ b/src/pytorch_compat.h
--- a/tests/README.md
+++ b/tests/README.md
--- a/tests/data/__init__.py
+++ b/tests/data/__init__.py
--- a/tests/flux/test_flux_cache.py
+++ b/tests/flux/test_flux_cache.py
--- a/tests/flux/test_flux_dev.py
+++ b/tests/flux/test_flux_dev.py