From 5edbdca6465e1f70df4c2d7e2879414a1db4c297 Mon Sep 17 00:00:00 2001 From: TianyiZhao1437 Date: Wed, 12 Nov 2025 12:43:55 +0000 Subject: [PATCH 1/2] fix(backend): fix monkey patch taking no effect. --- src/parallax/sglang/monkey_patch.py | 64 +++++++++++++++++------------ 1 file changed, 37 insertions(+), 27 deletions(-) diff --git a/src/parallax/sglang/monkey_patch.py b/src/parallax/sglang/monkey_patch.py index 4b2d45c9..299d4026 100644 --- a/src/parallax/sglang/monkey_patch.py +++ b/src/parallax/sglang/monkey_patch.py @@ -1,35 +1,45 @@ -from parallax.sglang.monkey_patch_utils.glm4_moe_model import ( - apply_glm4_moe_monkey_patch, -) -from parallax.sglang.monkey_patch_utils.gpt_oss_model import apply_gpt_oss_monkey_patch -from parallax.sglang.monkey_patch_utils.minimax_m2_model import ( - apply_minimax_m2_monkey_patch, -) -from parallax.sglang.monkey_patch_utils.model_parallel import ( - apply_model_parallel_monkey_patch, -) -from parallax.sglang.monkey_patch_utils.qwen3_next_config import ( - apply_qwen3_next_config_monkey_patch, -) -from parallax.sglang.monkey_patch_utils.qwen3_next_model import ( - apply_qwen3_next_monkey_patch, -) -from parallax.sglang.monkey_patch_utils.triton_backend import ( - apply_triton_backend_init_monkey_patch, -) -from parallax.sglang.monkey_patch_utils.weight_loader_filter import ( - apply_weight_loader_filter_patch, -) - - -## Here is some patch func for sglang -## Hopefully, when sglang support pipeline parallelism natively, we can remove these patches +""" +Here is some patch func for sglang +Hopefully, when sglang support pipeline parallelism natively, we can remove these patches +""" + def apply_parallax_sglang_monkey_patch(): + from parallax.sglang.monkey_patch_utils.model_parallel import ( + apply_model_parallel_monkey_patch, + ) apply_model_parallel_monkey_patch() + + from parallax.sglang.monkey_patch_utils.triton_backend import ( + apply_triton_backend_init_monkey_patch, + ) apply_triton_backend_init_monkey_patch() + + from parallax.sglang.monkey_patch_utils.weight_loader_filter import ( + apply_weight_loader_filter_patch, + ) apply_weight_loader_filter_patch() - apply_qwen3_next_monkey_patch() + + from parallax.sglang.monkey_patch_utils.qwen3_next_config import ( + apply_qwen3_next_config_monkey_patch, + ) apply_qwen3_next_config_monkey_patch() + + from parallax.sglang.monkey_patch_utils.qwen3_next_model import ( + apply_qwen3_next_monkey_patch, + ) + apply_qwen3_next_monkey_patch() + + from parallax.sglang.monkey_patch_utils.gpt_oss_model import ( + apply_gpt_oss_monkey_patch, + ) apply_gpt_oss_monkey_patch() + + from parallax.sglang.monkey_patch_utils.minimax_m2_model import ( + apply_minimax_m2_monkey_patch, + ) apply_minimax_m2_monkey_patch() + + from parallax.sglang.monkey_patch_utils.glm4_moe_model import ( + apply_glm4_moe_monkey_patch, + ) apply_glm4_moe_monkey_patch() From 2730588c170d9ed818b80a9d0e4c8d5627af08cf Mon Sep 17 00:00:00 2001 From: TianyiZhao1437 Date: Wed, 12 Nov 2025 20:55:20 +0800 Subject: [PATCH 2/2] fix pre-commit --- src/parallax/sglang/monkey_patch.py | 9 +++++++++ 1 file changed, 9 insertions(+) diff --git a/src/parallax/sglang/monkey_patch.py b/src/parallax/sglang/monkey_patch.py index 299d4026..a8b9da70 100644 --- a/src/parallax/sglang/monkey_patch.py +++ b/src/parallax/sglang/monkey_patch.py @@ -3,43 +3,52 @@ Hopefully, when sglang support pipeline parallelism natively, we can remove these patches """ + def apply_parallax_sglang_monkey_patch(): from parallax.sglang.monkey_patch_utils.model_parallel import ( apply_model_parallel_monkey_patch, ) + apply_model_parallel_monkey_patch() from parallax.sglang.monkey_patch_utils.triton_backend import ( apply_triton_backend_init_monkey_patch, ) + apply_triton_backend_init_monkey_patch() from parallax.sglang.monkey_patch_utils.weight_loader_filter import ( apply_weight_loader_filter_patch, ) + apply_weight_loader_filter_patch() from parallax.sglang.monkey_patch_utils.qwen3_next_config import ( apply_qwen3_next_config_monkey_patch, ) + apply_qwen3_next_config_monkey_patch() from parallax.sglang.monkey_patch_utils.qwen3_next_model import ( apply_qwen3_next_monkey_patch, ) + apply_qwen3_next_monkey_patch() from parallax.sglang.monkey_patch_utils.gpt_oss_model import ( apply_gpt_oss_monkey_patch, ) + apply_gpt_oss_monkey_patch() from parallax.sglang.monkey_patch_utils.minimax_m2_model import ( apply_minimax_m2_monkey_patch, ) + apply_minimax_m2_monkey_patch() from parallax.sglang.monkey_patch_utils.glm4_moe_model import ( apply_glm4_moe_monkey_patch, ) + apply_glm4_moe_monkey_patch()