From db02e45209c0d3a3c6329d9c89072aca840ca056 Mon Sep 17 00:00:00 2001 From: baptiste Date: Mon, 14 Oct 2024 12:16:42 +0000 Subject: [PATCH] fix bugs --- examples/trt_llama.yaml | 2 +- optimum_benchmark/backends/tensorrt_llm/config.py | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-) diff --git a/examples/trt_llama.yaml b/examples/trt_llama.yaml index bc3cb902..1aa5ecc2 100644 --- a/examples/trt_llama.yaml +++ b/examples/trt_llama.yaml @@ -20,7 +20,7 @@ backend: scenario: input_shapes: batch_size: 4 - sequence_length: 128 + sequence_length: 256 generate_kwargs: max_new_tokens: 100 min_new_tokens: 100 diff --git a/optimum_benchmark/backends/tensorrt_llm/config.py b/optimum_benchmark/backends/tensorrt_llm/config.py index d7f4b1cb..d3512b0e 100644 --- a/optimum_benchmark/backends/tensorrt_llm/config.py +++ b/optimum_benchmark/backends/tensorrt_llm/config.py @@ -24,7 +24,7 @@ class TRTLLMConfig(BackendConfig): world_size: int = 1 gpus_per_node: int = 1 - max_prompt_length: int = 128 + max_prompt_length: int = 256 max_new_tokens: int = -1 max_batch_size: int = 1 max_beam_width: int = 1