transformers版本问题

[AI-WangCG]

Hi,
感谢分享工作!

transformers 4.5.1无法下载RobertaTokenizerFast和RobertaModel
修改如下
【transformers 4.5.1】 ->【transformers 4.46.3】
if "roberta" in text_encoder_type:
self.tokenizer = RobertaTokenizerFast.from_pretrained(text_encoder_type)
self.text_encoder = RobertaModel.from_pretrained(text_encoder_type)

运行脚本如下：
export PYTHONPATH=./
export HF_ENDPOINT="https://hf-mirror.com"
python inference_on_custom_imgs_pseudo_coco.py
--param_path weights/RLIP_PDA_v2_HICO_R50_VGCOO365_COO365det_RQL_LSE_RPL_20e_L1_20e_checkpoint0019.pth
--save_path custom_imgs/result/custom_imgs.pickle
--batch_size 1
--num_obj_classes 80
--num_verb_classes 117
--dec_layers 3
--enc_layers 6
--num_queries 128
--use_no_obj_token
--backbone resnet50
--RLIP_ParSeDA_v2
--with_box_refine
--num_feature_levels 4
--num_patterns 0
--pe_temperatureH 20
--pe_temperatureW 20
--dim_feedforward 2048
--dropout 0.0
--subject_class
--use_no_obj_token
--fusion_type GLIP_attn
--gating_mechanism VXAc
--verb_query_tgt_type vanilla_MBF
--fusion_interval 2
--fusion_last_vis
--lang_aux_loss
异常如下
traceback (most recent call last):
File "inference_on_custom_imgs_pseudo_coco.py", line 918, in
main(args)
File "inference_on_custom_imgs_pseudo_coco.py", line 379, in main
load_info = model.load_state_dict(checkpoint['model'])
File "/usr/local/lib/python3.8/dist-packages/torch/nn/modules/module.py", line 1497, in load_state_dict
raise RuntimeError('Error(s) in loading state_dict for {}:\n\t{}'.format(
RuntimeError: Error(s) in loading state_dict for RLIP_ParSeDA:
Unexpected key(s) in state_dict: "transformer.text_encoder.embeddings.position_ids".