ilsilfverskiold/bart-keyword-extractor

Model description

This model is a fine-tuned version of facebook/bart-large on a dataset in the hub called sunhaozhepy/ag_news_keywords_embeddings to extract main keywords from text. It achieves the following results on the evaluation set:

• Loss: 0.6179

Intended use

from transformers import pipeline

pipe = pipeline('summarization', model='bart_keywords_model')
print(pipe("Aria Opera GPT version - All the browsers come with their own version of AI. So I gave it a try and ask it with LLM it was using. First if all it didn't understand the question. Then I explained and asked which version. I got the usual answer about a language model that is not aware of it's own model I find that curious, but also not transparent. My laptop, software all state their versions and critical information. But something that can easily fool a lot of people doesn't. What I also wonder if the general public will be stuck to ChatGPT 3.5 for ever while better models are behind expensive paywalls."))

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 5e-05
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 16
• total_train_batch_size: 128
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• num_epochs: 3

Training results

Training Loss	Epoch	Step	Validation Loss
0.7701	0.57	500	0.7390
0.5804	1.14	1000	0.7056
0.5395	1.71	1500	0.6811
0.4036	2.28	2000	0.6504
0.3763	2.85	2500	0.6179

Framework versions

• Transformers 4.35.2
• Pytorch 2.1.0+cu118
• Datasets 2.15.0
• Tokenizers 0.15.0