From 03e6f3412c72082eb62a6f2e53bc8e2bf79d0bc8 Mon Sep 17 00:00:00 2001 From: ADou Date: Wed, 9 Sep 2026 07:54:03 +0000 Subject: [PATCH] [None][docs] drop ghost --backend tensorrt from trtllm-eval docs trtllm-eval Click Choice for --backend is only ["pytorch"]; docs still told users to pass --backend tensorrt and that --model accepts pre-built TensorRT engines. Align docs/source/commands/trtllm-eval.rst and examples/trtllm-eval/README.md with live argparse after engine-backend removal. Signed-off-by: ADou --- docs/source/commands/trtllm-eval.rst | 4 +--- examples/trtllm-eval/README.md | 2 +- 2 files changed, 2 insertions(+), 4 deletions(-) diff --git a/docs/source/commands/trtllm-eval.rst b/docs/source/commands/trtllm-eval.rst index 7525551e9048..2b175c35492c 100644 --- a/docs/source/commands/trtllm-eval.rst +++ b/docs/source/commands/trtllm-eval.rst @@ -85,9 +85,7 @@ Here are some examples: trtllm-eval --model --max_seq_len 36864 aime25 trtllm-eval --model --max_seq_len 36864 aime26 -The ``--model`` argument accepts either a Hugging Face model ID or a local checkpoint path. By default, ``trtllm-eval`` runs the model with the PyTorch backend; you can pass ``--backend tensorrt`` to switch to the TensorRT backend. - -Alternatively, the ``--model`` argument also accepts a local path to pre-built TensorRT engines. In this case, you should pass the Hugging Face tokenizer path to the ``--tokenizer`` argument. +The ``--model`` argument accepts either a Hugging Face model ID or a local checkpoint path. ``trtllm-eval`` runs with the PyTorch backend (``--backend`` accepts only ``pytorch``); the TensorRT engine backend and pre-built engine paths were removed. For more details, see ``trtllm-eval --help`` and ``trtllm-eval --help``. diff --git a/examples/trtllm-eval/README.md b/examples/trtllm-eval/README.md index 3b9a389b2a2f..5437a852f96f 100644 --- a/examples/trtllm-eval/README.md +++ b/examples/trtllm-eval/README.md @@ -19,7 +19,7 @@ trtllm-eval --model meta-llama/Llama-3.1-8B-Instruct gsm8k trtllm-eval --model meta-llama/Llama-3.3-70B-Instruct gpqa_diamond ``` -The `--model` argument accepts either a Hugging Face model ID or a local checkpoint path. By default, `trtllm-eval` runs the model with the PyTorch backend; pass `--backend tensorrt` to switch to the TensorRT backend. Alternatively, the `--model` argument also accepts a local path to pre-built TensorRT engines; in that case, please pass the Hugging Face tokenizer path to the `--tokenizer` argument. +The `--model` argument accepts either a Hugging Face model ID or a local checkpoint path. `trtllm-eval` runs with the PyTorch backend (`--backend` accepts only `pytorch`); the TensorRT engine backend and pre-built engine paths were removed. See more details by `trtllm-eval --help`.