Anup K. Kalia comments

Results 8 comments of


                                            Anup K. Kalia

How to load rembert distilled models?

I tried the following ```python def convert_tf_checkpoint_to_pytorch(tf_checkpoint_path, bert_config_file, pytorch_dump_path): # Initialise PyTorch model config = RemBertConfig.from_json_file(bert_config_file) print(f"Building PyTorch model from configuration: {config}") model = RemBertModel(config) # Load weights from tf...

RuntimeError: The expanded size of the tensor (514) must match the existing size (512) at non-singleton dimension 1. Target sizes: [8, 514]. Tensor sizes: [1, 512]

I am having similar issues with Flair.

RuntimeError: The expanded size of the tensor (514) must match the existing size (512) at non-singleton dimension 1. Target sizes: [8, 514]. Tensor sizes: [1, 512]

Hey @stefan-it I was using a custom pretrain model based on google/rembert. For some odd reasons in self.tokenizer.model_max_length was coming larger than LARGE_INTEGER. This resulted in making self.truncate = False....

RuntimeError: The expanded size of the tensor (514) must match the existing size (512) at non-singleton dimension 1. Target sizes: [8, 514]. Tensor sizes: [1, 512]

Yes I am using the latest version and also I guess the error came from creating a custom pre-training model using the Rembert architecture.

Can I convert the flair fine-tuned model into a hugging face model?

Thanks !!

Get only the text of a span

This is not working

Techtokens get split based on colon

Even (64-bit) getting split into (64-bit and )

Documentation needed to interface with TCA from Tackle Inventory

Hi Derek (@djwhatle) Thanks for reaching out. You can follow through the following documentation. 1. https://github.com/konveyor/tackle-container-advisor (this will help you to get started e.g., deploying the API) 2. https://github.com/konveyor/tackle-container-advisor/tree/main/aca_backend_api (more...