pepi99

Results 24 comments of pepi99

New bitstamp api version (v2)

https://www.bitstamp.net/api/

GPU accelerated UMAP and HDBSCAN issues: memory and predict

Hello, I get same problem as Second issue: model.predict Did you manage to solve it?

GPU accelerated UMAP and HDBSCAN issues: memory and predict

Yes, as beckernick said, just remove the hdbscan=... in your bertopic model when you initialized, no problems after that! :)

Fine-tuning the model to add specific words and phrases with polarity

Same. I need to fine tune the model to recognize the sentiment for specific words in a specific domain/meaning. I suspect that this can be done by using transfer learning...

Training custom model

any solution?

How to fine tune this model

I thought about this, and this approach might lead to catastrophic forgetting. What if I just add some new examples into the SemEval task 4 dataset and retrain the model...

AttributeError with Tokenizer

Thank you for your solution. Now I get another error: Traceback (most recent call last): File "/Users/petar.ulev/Documents/prepare_sentiment_data/spacystuff.py", line 29, in task = nlp.preprocess(text=text, aspects=aspects) File "/Users/petar.ulev/Documents/prepare_sentiment_data/venv/lib/python3.8/site-packages/aspect_based_sentiment_analysis/pipelines.py", line 213, in preprocess...

How to pretrain mDeBERTa ?

What would be the difference between pre-training deberta from entirely new dataset, and fine-tuning the pre-trained deberta on MLM?

DeBERTa-V3 Pre-training code ?

UP. Is it now possible to pretrain Deberta-V3?

[Q] Is it possible to log a custom timestamp and use it as the "x-axis" for plots?

Bump