diff --git a/python/sglang/srt/layers/torchao_utils.py b/python/sglang/srt/layers/torchao_utils.py index 16eb1f2c5..bc7bde86e 100644 --- a/python/sglang/srt/layers/torchao_utils.py +++ b/python/sglang/srt/layers/torchao_utils.py @@ -3,15 +3,17 @@ Common utilities for torchao. """ import torch -from torchao.quantization import ( - int4_weight_only, - int8_dynamic_activation_int8_weight, - int8_weight_only, - quantize_, -) def torchao_quantize_param_data(param, torchao_config): + # Lazy import to suppress some warnings + from torchao.quantization import ( + int4_weight_only, + int8_dynamic_activation_int8_weight, + int8_weight_only, + quantize_, + ) + dummy_linear = torch.nn.Linear(param.shape[1], param.shape[0], bias=False) dummy_linear.weight = param if "int8wo" in torchao_config: diff --git a/test/killall_sglang.sh b/test/killall_sglang.sh index 0e2cb82a8..c536548d4 100644 --- a/test/killall_sglang.sh +++ b/test/killall_sglang.sh @@ -1 +1 @@ -kill -9 $(ps aux | grep 'sglang' | grep -v 'grep' | awk '{print $2}') +kill -9 $(ps aux | grep 'sglang.launch_server' | grep -v 'grep' | awk '{print $2}')