[AMD] Enable moe_a2a_backend=mori for DeepSeek-V4 prefill context parallelism (#35611)
This commit is contained in:
@@ -204,10 +204,10 @@ def validate_deepseek_v4_cp(server_args: ServerArgs) -> None:
|
||||
assert (
|
||||
cfg.tp_size <= 8
|
||||
), "Context parallel only supports single machine (tp_size <= 8). Cross-machine CP has precision issues."
|
||||
if cfg.moe_a2a_backend not in ("none", "deepep", "megamoe"):
|
||||
supported_a2a_backends = ("none", "deepep", "megamoe", "mori")
|
||||
if cfg.moe_a2a_backend not in supported_a2a_backends:
|
||||
raise ValueError(
|
||||
"DeepSeekV4 CP supports moe_a2a_backend in "
|
||||
"('none', 'deepep', 'megamoe'), "
|
||||
f"DeepSeekV4 CP supports moe_a2a_backend in {supported_a2a_backends}, "
|
||||
f"got {cfg.moe_a2a_backend!r}."
|
||||
)
|
||||
logger.warning(
|
||||
|
||||
@@ -2304,10 +2304,13 @@ class DeepseekV4DecoderLayer(nn.Module):
|
||||
if moe_a2a_backend.is_none():
|
||||
hidden_states = dsa_cp_gather_hidden_states(hidden_states)
|
||||
else:
|
||||
assert moe_a2a_backend.is_deepep() or moe_a2a_backend.is_megamoe(), (
|
||||
"CP requires DeepEP or megaMoE "
|
||||
"(moe_a2a_backend == deepep or megamoe). "
|
||||
f"Got {moe_a2a_backend.value}."
|
||||
assert (
|
||||
moe_a2a_backend.is_deepep()
|
||||
or moe_a2a_backend.is_megamoe()
|
||||
or moe_a2a_backend.is_mori()
|
||||
), (
|
||||
"CP requires moe_a2a_backend in ('deepep', 'megamoe', 'mori'), "
|
||||
f"got {moe_a2a_backend.value!r}."
|
||||
)
|
||||
elif _use_tp_moe_gather:
|
||||
hidden_states, local_hidden_states = (
|
||||
|
||||
Reference in New Issue
Block a user