EngineX-Hygon/sglang
5
0
Fork 0
Code Issues Pull Requests Actions 7 Projects Releases Wiki Activity

Fix multimodal registry and code sync scripts (#10759)

Browse Source 
Co-authored-by: cctry <shiyang@x.ai>
This commit is contained in:
Lianmin Zheng
2025-09-22 15:36:01 -07:00
committed by GitHub
parent d4041a5eeb
commit 38c00ed7a1
7 changed files with 38 additions and 15 deletions
Download Patch File Download Diff File Expand all files Collapse all files

11
python/sglang/srt/models/torch_native_llama.py
View File

@@ -66,8 +66,8 @@ from sglang.srt.model_executor.forward_batch_info import ForwardBatch
from sglang.srt.model_loader.weight_utils import default_weight_loader
from sglang.srt.utils import add_prefix
tp_size = get_tensor_model_parallel_world_size()
tp_rank = get_tensor_model_parallel_rank()
tp_size: Optional[int] = None
tp_rank: Optional[int] = None
def gate_up_proj_weight_loader(
@@ -341,6 +341,13 @@ class LlamaModel(nn.Module):
quant_config: Optional[QuantizationConfig] = None,
) -> None:
super().__init__()
global tp_size, tp_rank
if tp_size is None:
tp_size = get_tensor_model_parallel_world_size()
if tp_rank is None:
tp_rank = get_tensor_model_parallel_rank()
self.config = config
self.padding_idx = config.pad_token_id
self.vocab_size = config.vocab_size