chenchenghao/sglang
1
0
Fork 0
Code Issues Pull Requests Actions 11 Packages Projects Releases Wiki Activity

Revert "[FEAT] Support GGUF format" (#2285)

Browse Source
This commit is contained in:
Lianmin Zheng
2024-11-30 19:03:26 -08:00
committed by GitHub
parent d622851dc9
commit 7e4c6dd8da
39 changed files with 89 additions and 180 deletions
Download Patch File Download Diff File Expand all files Collapse all files

7
python/sglang/srt/models/phi3_small.py
View File

@@ -397,13 +397,10 @@ class Phi3SmallForCausalLM(nn.Module):
def compute_logits(
self,
input_ids: torch.LongTensor,
hidden_states: torch.Tensor,
sampling_metadata,
) -> Optional[torch.Tensor]:
logits = self.logits_processor(
input_ids, self.lm_head, hidden_states, sampling_metadata
)
logits = self.logits_processor(self.lm_head, hidden_states, sampling_metadata)
if self.dummy_token_indices is not None and logits is not None:
logits.index_fill_(-1, self.dummy_token_indices, -torch.inf)
return logits
@@ -425,7 +422,7 @@ class Phi3SmallForCausalLM(nn.Module):
if not get_embedding:
return self.logits_processor(
input_ids, hidden_states, self.lm_head, forward_batch
input_ids, hidden_states, self.lm_head.weight, forward_batch
)
else: