[Perf][DSv4.1] Shard the Engram wkv projection across TP ranks (#58678)
Signed-off-by: Shuolei Wang <shuoleiwang123@gmail.com>
This commit is contained in:
@@ -827,7 +827,7 @@ def test_engram_constructor_honors_offload(monkeypatch, backend, cpu_offload):
|
||||
monkeypatch.setattr(engram_ops, "get_tensor_model_parallel_world_size", lambda: 1)
|
||||
monkeypatch.setattr(engram_ops, "get_tensor_model_parallel_rank", lambda: 0)
|
||||
monkeypatch.setattr(
|
||||
engram_ops, "ReplicatedLinear", lambda *a, **k: torch.nn.Identity()
|
||||
engram_ops, "ColumnParallelLinear", lambda *a, **k: torch.nn.Identity()
|
||||
)
|
||||
stream = torch.cuda.Stream()
|
||||
kwargs = {"prefetch_stream": stream} if backend == "nvidia" else {}
|
||||
|
||||
Reference in New Issue
Block a user