AdvancedLiterateMachinery

AdvancedLiterateMachinery copied to clipboard

Lister converges extremely slow

1

Have not changed anything and uses Synthtext for training, loss still stays around 3.0 after 2hours training?

Apostatee

some questions about embedding in the code and in the paper

1

 if inputs_embeds is None: inputs_embeds = self.word_embeddings(input_ids) token_type_embeddings = self.token_type_embeddings(token_type_ids) embeddings = inputs_embeds + token_type_embeddings if self.position_embedding_type == "absolute": position_embeddings = self.position_embeddings(position_ids) embeddings += position_embeddings if "line_bbox" in kwargs:...

kenneys-bot

Final results' huge gap while using & not using pretrained weights?

3

I train the model without pretrained weights, the final results are as below [according to the paper, ser f1 should be 83.39%, re f1 should be 74.91%]:  However, when...

tonylin52

light42

Did you guys share the pretraining code

3

I'm reading the paper. And for our project, the pretraining part is more helpful so I wonder if you guys also shared these code somewhere? Also I have a question...

menglin0320

作者，您好！VLPT-STD是个很棒的工作，但是目前发布的只有预训练的代码，能否提供测试模型的代码和交叉注意力可视化的代码？

jiangduwang

Is it possible to fine-tune Geolayout LM only for entity labelling ?

2

The data on which we want to finetune Geolayout LM model doesnt have key-value pairs. For ex Name,Address,Dates etc Is it possible to ignore entity linking part and just train...

SR1608

Problems with H1, H2, and H3 calculations in the code in the paper

1

 Thank you very much for your help before, I've been working on your model recently, but I'm confused about the calculation of H1, H2, and H3 in the graph,...

kenneys-bot

The "test_data_path" for LISTER

1

Thank you for your excellent work and releasing the source code. LISTER is very inspiring for solving the corner cases of STR on long text. Could you give more details...

icecream-Tnak

ManikantaNT