vrr/kvcache-ai-ktransformers
2
0
Fork 0
mirror of https://github.com/kvcache-ai/ktransformers.git synced 2025-09-10 23:34:35 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

Update modeling_deepseek_v3.py

Browse source
This commit is contained in:
Atream 2025-04-03 17:13:06 +08:00 • committed by GitHub
parent 016d11e6d4
commit e36ddc36a8
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
1 changed files with 1 additions and 1 deletions
Download patch file Download diff file Expand all files Collapse all files

2
ktransformers/models/modeling_deepseek_v3.py
View file

@ -436,7 +436,7 @@ class MoEGate(nn.Module):
### select top-k experts ### select top-k experts
if self.topk_method == "noaux_tc": if self.topk_method == "noaux_tc":
assert not self.training #assert not self.training
scores_for_choice = scores.view(bsz * seq_len, -1) + self.e_score_correction_bias.unsqueeze(0) scores_for_choice = scores.view(bsz * seq_len, -1) + self.e_score_correction_bias.unsqueeze(0)
group_scores = ( group_scores = (
scores_for_choice.view(bsz * seq_len, self.n_group, -1).topk(2, dim=-1)[0].sum(dim = -1) scores_for_choice.view(bsz * seq_len, self.n_group, -1).topk(2, dim=-1)[0].sum(dim = -1)