[Fix] Fix DeepSeek V3 MTP on B200 (#13548)
This commit is contained in:
@@ -220,6 +220,7 @@ class DeepseekV3ForCausalLMNextN(DeepseekV3ForCausalLM):
|
|||||||
use_attn_tp_group=get_global_server_args().enable_dp_lm_head,
|
use_attn_tp_group=get_global_server_args().enable_dp_lm_head,
|
||||||
)
|
)
|
||||||
self.logits_processor = LogitsProcessor(config)
|
self.logits_processor = LogitsProcessor(config)
|
||||||
|
self._executed_weight_requant_ue8m0 = False
|
||||||
|
|
||||||
@torch.no_grad()
|
@torch.no_grad()
|
||||||
def forward(
|
def forward(
|
||||||
|
|||||||
Reference in New Issue
Block a user