Skip to content
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

Jump to bottom

replace skip_embed with input_embeds #222

Merged
merged 1 commit into from
Mar 11, 2024
Merged

replace skip_embed with input_embeds #222

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
10 changes: 5 additions & 5 deletions python/sglang/srt/models/llama2.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -227,12 +227,12 @@ def forward(
input_ids: torch.Tensor,
positions: torch.Tensor,
input_metadata: InputMetadata,
skip_embed: bool = False,
input_embeds: torch.Tensor = None,
) -> torch.Tensor:
if not skip_embed:
if input_embeds is None:
hidden_states = self.embed_tokens(input_ids)
else:
hidden_states = input_ids
hidden_states = input_embeds
residual = None
for i in range(len(self.layers)):
layer = self.layers[i]
Expand Down Expand Up @@ -264,9 +264,9 @@ def forward(
input_ids: torch.Tensor,
positions: torch.Tensor,
input_metadata: InputMetadata,
skip_embed: bool = False,
input_embeds: torch.Tensor = None,
) -> torch.Tensor:
hidden_states = self.model(input_ids, positions, input_metadata, skip_embed)
hidden_states = self.model(input_ids, positions, input_metadata, input_embeds)
return self.logits_processor(
input_ids, hidden_states, self.lm_head.weight, input_metadata
)
Expand Down
4 changes: 2 additions & 2 deletions python/sglang/srt/models/llava.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -230,11 +230,11 @@ def forward(
pt += 1

return self.language_model(
input_embeds, positions, input_metadata, skip_embed=True
input_ids, positions, input_metadata, input_embeds=input_embeds
)
elif input_metadata.forward_mode == ForwardMode.DECODE:
return self.language_model(
input_ids, positions, input_metadata, skip_embed=False
input_ids, positions, input_metadata
)

def load_weights(
Expand Down
10 changes: 5 additions & 5 deletions python/sglang/srt/models/mixtral.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -296,12 +296,12 @@ def forward(
input_ids: torch.Tensor,
positions: torch.Tensor,
input_metadata: InputMetadata,
skip_embed: bool = False,
input_embeds: torch.Tensor = None,
) -> torch.Tensor:
if not skip_embed:
if input_embeds is None:
hidden_states = self.embed_tokens(input_ids)
else:
hidden_states = input_ids
hidden_states = input_embeds
residual = None
for i in range(len(self.layers)):
layer = self.layers[i]
Expand Down Expand Up @@ -330,9 +330,9 @@ def forward(
input_ids: torch.Tensor,
positions: torch.Tensor,
input_metadata: InputMetadata,
skip_embed: bool = False,
input_embeds: torch.Tensor = None,
) -> torch.Tensor:
hidden_states = self.model(input_ids, positions, input_metadata, skip_embed)
hidden_states = self.model(input_ids, positions, input_metadata, input_embeds)
return self.logits_processor(
input_ids, hidden_states, self.lm_head.weight, input_metadata
)
Expand Down
10 changes: 5 additions & 5 deletions python/sglang/srt/models/qwen2.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -228,12 +228,12 @@ def forward(
input_ids: torch.Tensor,
positions: torch.Tensor,
input_metadata: InputMetadata,
skip_embed: bool = False,
input_embeds: torch.Tensor = None,
) -> torch.Tensor:
if not skip_embed:
if input_embeds is None:
hidden_states = self.embed_tokens(input_ids)
else:
hidden_states = input_ids
hidden_states = input_embeds
residual = None
for i in range(len(self.layers)):
layer = self.layers[i]
Expand Down Expand Up @@ -265,9 +265,9 @@ def forward(
input_ids: torch.Tensor,
positions: torch.Tensor,
input_metadata: InputMetadata,
skip_embed: bool = False,
input_embeds: torch.Tensor = None,
) -> torch.Tensor:
hidden_states = self.model(input_ids, positions, input_metadata, skip_embed)
hidden_states = self.model(input_ids, positions, input_metadata, input_embeds)
return self.logits_processor(
input_ids, hidden_states, self.lm_head.weight, input_metadata
)
Expand Down