[hotfix] Fix import error: colossal.kernel without triton installed (#4722)

* [hotfix] remove triton kernels from kernel init

* revise bloom/llama kernel imports for infer
This commit is contained in:
Yuanheng Zhao
2023-09-14 18:03:55 +08:00
committed by GitHub
parent c7d6975d29
commit e2c0e7f92a
6 changed files with 28 additions and 30 deletions

View File

@@ -1,14 +1,7 @@
from .cuda_native import FusedScaleMaskSoftmax, LayerNorm, MultiHeadAttention
from .triton import llama_context_attn_fwd, bloom_context_attn_fwd
from .triton import softmax
from .triton import copy_kv_cache_to_dest
__all__ = [
"LayerNorm",
"FusedScaleMaskSoftmax",
"MultiHeadAttention",
"llama_context_attn_fwd",
"bloom_context_attn_fwd",
"softmax",
"copy_kv_cache_to_dest",
]