EngineX-Hygon/sglang
5
0
Fork 0
Code Issues Pull Requests Actions 7 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

8
python/sglang/srt/models/minicpm3.py
View File

@@ -585,10 +585,12 @@ class MiniCPM3ForCausalLM(nn.Module):
hidden_states = self.model(input_ids, positions, forward_batch, input_embeds)
hidden_states = hidden_states / self.scale_width
if self.config.tie_word_embeddings:
lm_head = self.model.embed_tokens
lm_head_weight = self.model.embed_tokens.weight
else:
lm_head = self.lm_head
return self.logits_processor(input_ids, hidden_states, lm_head, forward_batch)
lm_head_weight = self.lm_head.weight
return self.logits_processor(
input_ids, hidden_states, lm_head_weight, forward_batch
)
def load_weights(self, weights: Iterable[Tuple[str, torch.Tensor]]):
stacked_params_mapping = [