KIE: Segmentation fault (core dumped)

[botastark]

I can't infer (tools/infer_kie_token_ser.py) from finetuned ser_vi_layoutxlm_xfund_zh_udml. I use ser_vi_layoutxlm_xfund_zh.yml for inference similar as config used for training ser_vi_layoutxlm_xfund_zh_udml.yml after loading the model, it says Segmentation fault (core dumped)

[2022/11/09 16:02:33] ppocr INFO: Architecture : [2022/11/09 16:02:33] ppocr INFO: Backbone : [2022/11/09 16:02:33] ppocr INFO: checkpoints : output/ser_vi_layoutxlm_xfund_zh_udml_play/best_accuracy [2022/11/09 16:02:33] ppocr INFO: mode : vi [2022/11/09 16:02:33] ppocr INFO: name : LayoutXLMForSer [2022/11/09 16:02:33] ppocr INFO: num_classes : 89 [2022/11/09 16:02:33] ppocr INFO: pretrained : True [2022/11/09 16:02:33] ppocr INFO: Transform : None [2022/11/09 16:02:33] ppocr INFO: algorithm : LayoutXLM [2022/11/09 16:02:33] ppocr INFO: model_type : kie [2022/11/09 16:02:33] ppocr INFO: Eval : [2022/11/09 16:02:33] ppocr INFO: dataset : [2022/11/09 16:02:33] ppocr INFO: data_dir : train_data_play/val [2022/11/09 16:02:33] ppocr INFO: label_file_list : ['train_data_play/val/val.txt'] [2022/11/09 16:02:33] ppocr INFO: name : SimpleDataSet [2022/11/09 16:02:33] ppocr INFO: transforms : [2022/11/09 16:02:33] ppocr INFO: DecodeImage : [2022/11/09 16:02:33] ppocr INFO: channel_first : False [2022/11/09 16:02:33] ppocr INFO: img_mode : RGB [2022/11/09 16:02:33] ppocr INFO: VQATokenLabelEncode : [2022/11/09 16:02:33] ppocr INFO: algorithm : LayoutXLM [2022/11/09 16:02:33] ppocr INFO: class_path : train_data_play/class_list.txt [2022/11/09 16:02:33] ppocr INFO: contains_re : False [2022/11/09 16:02:33] ppocr INFO: order_method : tb-yx [2022/11/09 16:02:33] ppocr INFO: use_textline_bbox_info : True [2022/11/09 16:02:33] ppocr INFO: VQATokenPad : [2022/11/09 16:02:33] ppocr INFO: max_seq_len : 512 [2022/11/09 16:02:33] ppocr INFO: return_attention_mask : True [2022/11/09 16:02:33] ppocr INFO: VQASerTokenChunk : [2022/11/09 16:02:33] ppocr INFO: max_seq_len : 512 [2022/11/09 16:02:33] ppocr INFO: Resize : [2022/11/09 16:02:33] ppocr INFO: size : [224, 224] [2022/11/09 16:02:33] ppocr INFO: NormalizeImage : [2022/11/09 16:02:33] ppocr INFO: mean : [123.675, 116.28, 103.53] [2022/11/09 16:02:33] ppocr INFO: order : hwc [2022/11/09 16:02:33] ppocr INFO: scale : 1 [2022/11/09 16:02:33] ppocr INFO: std : [58.395, 57.12, 57.375] [2022/11/09 16:02:33] ppocr INFO: ToCHWImage : None [2022/11/09 16:02:33] ppocr INFO: KeepKeys : [2022/11/09 16:02:33] ppocr INFO: keep_keys : ['input_ids', 'bbox', 'attention_mask', 'token_type_ids', 'image', 'labels'] [2022/11/09 16:02:33] ppocr INFO: loader : [2022/11/09 16:02:33] ppocr INFO: batch_size_per_card : 2 [2022/11/09 16:02:33] ppocr INFO: drop_last : False [2022/11/09 16:02:33] ppocr INFO: num_workers : 4 [2022/11/09 16:02:33] ppocr INFO: shuffle : False [2022/11/09 16:02:33] ppocr INFO: Global : [2022/11/09 16:02:33] ppocr INFO: cal_metric_during_train : False [2022/11/09 16:02:33] ppocr INFO: class_path : train_data_play/class_list.txt [2022/11/09 16:02:33] ppocr INFO: distributed : False [2022/11/09 16:02:33] ppocr INFO: epoch_num : 50 [2022/11/09 16:02:33] ppocr INFO: eval_batch_step : [0, 19] [2022/11/09 16:02:33] ppocr INFO: infer_img : train_data_play/val/20223000527387379915_Librettodicircolazione_page_0-fs8.png [2022/11/09 16:02:33] ppocr INFO: kie_det_model_dir : None [2022/11/09 16:02:33] ppocr INFO: kie_rec_model_dir : None [2022/11/09 16:02:33] ppocr INFO: log_smooth_window : 10 [2022/11/09 16:02:33] ppocr INFO: print_batch_step : 10 [2022/11/09 16:02:33] ppocr INFO: save_epoch_step : 2000 [2022/11/09 16:02:33] ppocr INFO: save_inference_dir : None [2022/11/09 16:02:33] ppocr INFO: save_model_dir : ./output/ser_vi_layoutxlm_xfund_zh_udml_play [2022/11/09 16:02:33] ppocr INFO: save_res_path : output/ser_res_play/ [2022/11/09 16:02:33] ppocr INFO: seed : 2022 [2022/11/09 16:02:33] ppocr INFO: use_gpu : True [2022/11/09 16:02:33] ppocr INFO: use_visualdl : True [2022/11/09 16:02:33] ppocr INFO: Loss : [2022/11/09 16:02:33] ppocr INFO: key : backbone_out [2022/11/09 16:02:33] ppocr INFO: name : VQASerTokenLayoutLMLoss [2022/11/09 16:02:33] ppocr INFO: num_classes : 89 [2022/11/09 16:02:33] ppocr INFO: Metric : [2022/11/09 16:02:33] ppocr INFO: main_indicator : hmean [2022/11/09 16:02:33] ppocr INFO: name : VQASerTokenMetric [2022/11/09 16:02:33] ppocr INFO: Optimizer : [2022/11/09 16:02:33] ppocr INFO: beta1 : 0.9 [2022/11/09 16:02:33] ppocr INFO: beta2 : 0.999 [2022/11/09 16:02:33] ppocr INFO: lr : [2022/11/09 16:02:33] ppocr INFO: epochs : 50 [2022/11/09 16:02:33] ppocr INFO: learning_rate : 5e-05 [2022/11/09 16:02:33] ppocr INFO: name : Linear [2022/11/09 16:02:33] ppocr INFO: warmup_epoch : 2 [2022/11/09 16:02:33] ppocr INFO: name : AdamW [2022/11/09 16:02:33] ppocr INFO: regularizer : [2022/11/09 16:02:33] ppocr INFO: factor : 0.0 [2022/11/09 16:02:33] ppocr INFO: name : L2 [2022/11/09 16:02:33] ppocr INFO: PostProcess : [2022/11/09 16:02:33] ppocr INFO: class_path : train_data_play/class_list.txt [2022/11/09 16:02:33] ppocr INFO: name : VQASerTokenLayoutLMPostProcess [2022/11/09 16:02:33] ppocr INFO: Train : [2022/11/09 16:02:33] ppocr INFO: dataset : [2022/11/09 16:02:33] ppocr INFO: data_dir : train_data_play/train [2022/11/09 16:02:33] ppocr INFO: label_file_list : ['train_data_play/train/train.txt'] [2022/11/09 16:02:33] ppocr INFO: name : SimpleDataSet [2022/11/09 16:02:33] ppocr INFO: ratio_list : [1.0] [2022/11/09 16:02:33] ppocr INFO: transforms : [2022/11/09 16:02:33] ppocr INFO: DecodeImage : [2022/11/09 16:02:33] ppocr INFO: channel_first : False [2022/11/09 16:02:33] ppocr INFO: img_mode : RGB [2022/11/09 16:02:33] ppocr INFO: VQATokenLabelEncode : [2022/11/09 16:02:33] ppocr INFO: algorithm : LayoutXLM [2022/11/09 16:02:33] ppocr INFO: class_path : train_data_play/class_list.txt [2022/11/09 16:02:33] ppocr INFO: contains_re : False [2022/11/09 16:02:33] ppocr INFO: order_method : tb-yx [2022/11/09 16:02:33] ppocr INFO: use_textline_bbox_info : True [2022/11/09 16:02:33] ppocr INFO: VQATokenPad : [2022/11/09 16:02:33] ppocr INFO: max_seq_len : 512 [2022/11/09 16:02:33] ppocr INFO: return_attention_mask : True [2022/11/09 16:02:33] ppocr INFO: VQASerTokenChunk : [2022/11/09 16:02:33] ppocr INFO: max_seq_len : 512 [2022/11/09 16:02:33] ppocr INFO: Resize : [2022/11/09 16:02:33] ppocr INFO: size : [224, 224] [2022/11/09 16:02:33] ppocr INFO: NormalizeImage : [2022/11/09 16:02:33] ppocr INFO: mean : [123.675, 116.28, 103.53] [2022/11/09 16:02:33] ppocr INFO: order : hwc [2022/11/09 16:02:33] ppocr INFO: scale : 1 [2022/11/09 16:02:33] ppocr INFO: std : [58.395, 57.12, 57.375] [2022/11/09 16:02:33] ppocr INFO: ToCHWImage : None [2022/11/09 16:02:33] ppocr INFO: KeepKeys : [2022/11/09 16:02:33] ppocr INFO: keep_keys : ['input_ids', 'bbox', 'attention_mask', 'token_type_ids', 'image', 'labels'] [2022/11/09 16:02:33] ppocr INFO: loader : [2022/11/09 16:02:33] ppocr INFO: batch_size_per_card : 2 [2022/11/09 16:02:33] ppocr INFO: drop_last : False [2022/11/09 16:02:33] ppocr INFO: num_workers : 4 [2022/11/09 16:02:33] ppocr INFO: shuffle : True [2022/11/09 16:02:33] ppocr INFO: profiler_options : None [2022/11/09 16:02:33] ppocr INFO: train with paddle 2.3.2 and device Place(gpu:0) W1109 16:02:33.859344 5909 gpu_resources.cc:61] Please NOTE: device: 0, GPU Compute Capability: 7.5, Driver API Version: 11.4, Runtime API Version: 10.2 W1109 16:02:33.863689 5909 gpu_resources.cc:91] device: 0, cuDNN Version: 8.5. [2022/11/09 16:02:36] ppocr INFO: resume from output/ser_vi_layoutxlm_xfund_zh_udml_play/best_accuracy [2022-11-09 16:02:37,441] [ INFO] - Already cached /root/.paddlenlp/models/layoutxlm-base-uncased/sentencepiece.bpe.model [2022-11-09 16:02:38,004] [ INFO] - tokenizer config file saved in /root/.paddlenlp/models/layoutxlm-base-uncased/tokenizer_config.json [2022-11-09 16:02:38,022] [ INFO] - Special tokens file saved in /root/.paddlenlp/models/layoutxlm-base-uncased/special_tokens_map.json Segmentation fault (core dumped)