-
Currently, we use Cosine Similarity for similarity metric. With complex architectures like BERT, it may not be effective as the objective functions used for pre-training or fine-tuning does not direct…
-
Hi, thanks for providing this code! Could you please give more information (e.g. a brief explanation) of the following options?
- max_align=5
- top_k=3
- win=5
- skip=-0.1
- margin=True
- len_pe…
-
Subject: Unable to add todolist formatting that is distinct from individual todos
### Problem
- The todolist is actually output as one or more todo/paragraph pairs with no wrapper node that can be…
-
Hi,
Thanks so much for the great project - SBERT! I am fine-tuning a classification task using the pre-trained `bert-base-uncased` in SBERT. I am trying to understand how the accuracy of binary class…
-
Is this even possible? Can we with minimal a priori knowledge can we separate sentences in all languages in all scripts enough so that when combined with a Gale-Church sentence Aligner, we can get de…
-
The repo seems not compatible with cuda.
```
(webaskb) ➜ WebAsKB git:(master) ✗ python webaskb_run.py train_ptrnet
load the GloVe dataset sample that matches to our vocabula…
Vimos updated
6 years ago
-
Hi, I tried to train my own custom data using some large pretrained model (multilingual) such as multilingual BERT or XML roBERTa, and even if I tried reducing my batch size (for current my batch size…
-
### Metadata
- Authors: Guillaume Lample, Ludovic Denoyer, Marc'Aurelio Ranzato
- Organization: Facebook AI Research
- Conference: ICLR 2018
- Link: https://openreview.net/forum?id=rkYTTf-AZ
-
Hello,
I am working with data that contains a mix of languages (e.g. English, France, Spanish)
I would like to use pre-trained sentence-bert model to find similarities between pairs of text. In some…
-
Hi,
Thanks very much for your work, these datasets are great for evaluation. Given many positive examples in datasets like MRPC and TwitterURL are quasi-paraphrases (similar but not the same meanin…