Expert Parallelism for GPT-OSS (#8944)
This commit is contained in:
@@ -2961,3 +2961,8 @@ class ConcurrentCounter:
|
||||
other tasks to run while waiting. When the counter becomes zero, the coroutine resumes.
|
||||
"""
|
||||
self.wait_for(lambda count: count == 0)
|
||||
|
||||
|
||||
@lru_cache(maxsize=1)
|
||||
def is_triton_kernels_available() -> bool:
|
||||
return importlib.util.find_spec("triton_kernels") is not None
|
||||
|
||||
Reference in New Issue
Block a user