From d73ea263855e614a160b4cb33291a504340ea857 Mon Sep 17 00:00:00 2001 From: "Zhou, Huijuan" Date: Tue, 16 Sep 2025 11:16:05 +0800 Subject: [PATCH] fix for GenerationConfig issue for pipeline init changes by transformer update --- optimum/habana/transformers/modeling_utils.py | 4 ++++ optimum/habana/transformers/pipelines/__init__.py | 1 + optimum/habana/transformers/pipelines/image_to_text.py | 9 +++++++++ 3 files changed, 14 insertions(+) create mode 100644 optimum/habana/transformers/pipelines/__init__.py create mode 100644 optimum/habana/transformers/pipelines/image_to_text.py diff --git a/optimum/habana/transformers/modeling_utils.py b/optimum/habana/transformers/modeling_utils.py index 0c00d27017..0d1706ec0b 100644 --- a/optimum/habana/transformers/modeling_utils.py +++ b/optimum/habana/transformers/modeling_utils.py @@ -315,6 +315,7 @@ gaudi_XLMRoberta_Sdpa_SelfAttention_forward, ) from .models.deepseek_v2.modeling_deepseek_v2 import DeepseekV2ForCausalLM as GaudiDeepseekV2ForCausalLM +from .pipelines import GaudiImageToTextPipeline def adapt_transformers_to_gaudi(): @@ -394,6 +395,9 @@ def adapt_transformers_to_gaudi(): transformers.generation.MaxTimeCriteria.__call__ = gaudi_MaxTimeCriteria_call transformers.generation.EosTokenCriteria.__call__ = gaudi_EosTokenCriteria_call transformers.generation.StoppingCriteriaList.__call__ = gaudi_StoppingCriteriaList_call + transformers.pipelines.image_to_text.ImageToTextPipeline._default_generation_config = ( + GaudiImageToTextPipeline._default_generation_config + ) # Optimization for BLOOM generation on Gaudi transformers.models.bloom.modeling_bloom.BloomAttention.forward = gaudi_bloom_attention_forward diff --git a/optimum/habana/transformers/pipelines/__init__.py b/optimum/habana/transformers/pipelines/__init__.py new file mode 100644 index 0000000000..9377c50d9a --- /dev/null +++ b/optimum/habana/transformers/pipelines/__init__.py @@ -0,0 +1 @@ +from .image_to_text import GaudiImageToTextPipeline diff --git a/optimum/habana/transformers/pipelines/image_to_text.py b/optimum/habana/transformers/pipelines/image_to_text.py new file mode 100644 index 0000000000..fa1ecf4592 --- /dev/null +++ b/optimum/habana/transformers/pipelines/image_to_text.py @@ -0,0 +1,9 @@ +from transformers.pipelines.image_to_text import ImageToTextPipeline + +from ..generation import GaudiGenerationConfig + + +class GaudiImageToTextPipeline(ImageToTextPipeline): + _default_generation_config = GaudiGenerationConfig( + max_new_tokens=256, + )