From a5b59fd31db6f44988f2b061ebffb927f35bfd6b Mon Sep 17 00:00:00 2001 From: Grzegorz Kwasniewski <213329731+greg-kwasniewski1@users.noreply.github.com> Date: Fri, 3 Oct 2025 00:41:59 +0200 Subject: [PATCH] [TRTLLM-6342][bug] Patched incorrect starcoder tp config (#8118) Signed-off-by: greg-kwasniewski1 <213329731+greg-kwasniewski1@users.noreply.github.com> --- tensorrt_llm/_torch/auto_deploy/models/patches/starcoder.py | 4 ++++ 1 file changed, 4 insertions(+) create mode 100644 tensorrt_llm/_torch/auto_deploy/models/patches/starcoder.py diff --git a/tensorrt_llm/_torch/auto_deploy/models/patches/starcoder.py b/tensorrt_llm/_torch/auto_deploy/models/patches/starcoder.py new file mode 100644 index 0000000000..4d28bec3d1 --- /dev/null +++ b/tensorrt_llm/_torch/auto_deploy/models/patches/starcoder.py @@ -0,0 +1,4 @@ +from transformers.models.starcoder2.configuration_starcoder2 import Starcoder2Config + +# Remove this patch after TRT-LLM upgrades to the HF transformers version >= 4.57 +Starcoder2Config.base_model_tp_plan["layers.*.mlp.c_proj"] = "rowwise"