diff --git a/energy_star/text_generation.yaml b/energy_star/text_generation.yaml index cef9bf39..348a6b51 100644 --- a/energy_star/text_generation.yaml +++ b/energy_star/text_generation.yaml @@ -17,8 +17,8 @@ backend: device_ids: 0 no_weights: False task: text-generation - model: openai/gpt-oss-20b - torch_dtype: auto + model: HuggingFaceTB/SmolLM3-3B + torch_dtype: bfloat16 device_map: auto scenario: @@ -26,9 +26,7 @@ scenario: text_column_name: text num_samples: 1000 truncation: True - reasoning: True - reasoning_params: - reasoning_effort: high + reasoning: False input_shapes: batch_size: 1 diff --git a/optimum_benchmark/preprocessors/dataset_preprocessor.py b/optimum_benchmark/preprocessors/dataset_preprocessor.py index ab0a92a4..84c4f647 100644 --- a/optimum_benchmark/preprocessors/dataset_preprocessor.py +++ b/optimum_benchmark/preprocessors/dataset_preprocessor.py @@ -287,7 +287,7 @@ def tokenize_function(examples): return pretrained_processor( examples[scenario_config.text_column_name], truncation=scenario_config.truncation, - max_length=max_length - new_tokens, + max_length=min(max_length, 2048) - new_tokens, return_token_type_ids=False, padding=padding, )