Skip to content

LLaMAMoE fixes #2014

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
ysjprojects wants to merge 3 commits into from
Closed

LLaMAMoE fixes #2014

Changes from 1 commit
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
fc7892b
LLaMAMoE: added L1 norm to probs
ysjprojects Apr 13, 2025
7b5fccd
Merge branch 'main' into llamamoe-fix
Borda May 15, 2025
ae0781c
Merge branch 'main' into llamamoe-fix
Borda May 15, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
4 changes: 3 additions & 1 deletion litgpt/model.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -566,8 +566,10 @@ def forward(self, x: torch.Tensor) -> torch.Tensor:
B, T, C = x.size() # batch size, sequence length, embedding dimensionality (n_embd)
x = x.view(-1, C) # (B*T, C)
router = self.gate(x) # (B*T, n_expert)
router = F.softmax(router, dim=1, dtype=torch.float)
probs, indices = torch.topk(router, self.config.n_expert_per_token) # (B*T, n_expert_per_token)
probs = probs.softmax(dim=1, dtype=torch.float).to(dtype=x.dtype)
probs /= probs.sum(dim=1, keepdim=True)
probs = probs.to(dtype=x.dtype)
masks = indices.unsqueeze(-1) == torch.arange(self.config.n_expert, device=x.device)
masks = masks.permute(2, 0, 1) # (n_expert, B*T, n_expert_per_token)
y = torch.zeros_like(x) # (B*T, C)
Expand Down
Loading