From 8d1bfd38745164e8bc586b4526700ea1e6b5fc3e Mon Sep 17 00:00:00 2001 From: Gang Wen Date: Tue, 7 Apr 2026 19:57:01 +0200 Subject: [PATCH] Use AutoTokenizer instead of BartTokenizerFast for broader model support Some Florence2 variants (e.g. Florence-2-Flux-Large) use a RoBERTa tokenizer. AutoTokenizer auto-detects the correct tokenizer class. Co-Authored-By: Claude Opus 4.6 --- py/florence2_ultra.py | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/py/florence2_ultra.py b/py/florence2_ultra.py index 44ea558..bc302fd 100644 --- a/py/florence2_ultra.py +++ b/py/florence2_ultra.py @@ -49,7 +49,7 @@ def _load_model_v5(model_path, attention, dtype): """Load Florence2 model for transformers >= 5.0.0""" log(f"[DEBUG] _load_model_v5 called with model_path={model_path}, attention={attention}, dtype={dtype}") from ..florence2_models.modeling_florence2 import Florence2ForConditionalGeneration, Florence2Config - from transformers import CLIPImageProcessor, BartTokenizerFast + from transformers import CLIPImageProcessor, AutoTokenizer from ..florence2_models.processing_florence2 import Florence2Processor from accelerate import init_empty_weights from accelerate.utils import set_module_tensor_to_device @@ -108,7 +108,7 @@ def _load_model_v5(model_path, attention, dtype): image_processor.image_seq_length = 577 log(f"[DEBUG] Loading tokenizer from {model_path}") - tokenizer = BartTokenizerFast.from_pretrained(model_path) + tokenizer = AutoTokenizer.from_pretrained(model_path) log(f"[DEBUG] Creating Florence2Processor") processor = Florence2Processor(image_processor=image_processor, tokenizer=tokenizer) log(f"[DEBUG] _load_model_v5 completed successfully")