bert-base-spanish-wwm-cased-finetuned-tweets

This model is a fine-tuned version of bert-base-spanish-wwm-cased on the dataset petro-tweets. It achieves the following results on the evaluation set:

• Loss: 2.7191
• Perplexity: 15.17

Model description

This model is a fine-tuned version of bert-base-spanish-wwm-cased using the dataset from petro-tweets (available in Huggin Face). The purpose is to extend the bert-base-spanish-cased domain, which, once fine-tuned, will be modified for the fill-in-the-gaps task.

Intended uses & limitations

The use is limited to school use and the limitations have to do with the size of the dataset, since it does not allow for a large contribution, a larger dataset would have to be used to get a larger contribution.

Training and evaluation data

I did a training that gives the training and validation set loss. (It takes a lot of time. If you're using colab, I recommend to use less Epochs because the result does not change too much, and even though the loss is quite high, the performance of the model based on the perplexity is quite good) Also, I checked the perplexity, which is one good measure for Languages Models. As seen before, it gives a good perplexity, because it's quite low.

• Evaluation: I checked the performance of my model in the notebook provided, just by generating examples.

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 2e-05
• train_batch_size: 64
• eval_batch_size: 64
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 20

Training results

Training Loss	Epoch	Step	Validation Loss
1.9867	1.0	79	2.8373
2.1251	2.0	158	2.8581
2.0998	3.0	237	2.7878
2.1207	4.0	316	2.8046
2.1268	5.0	395	2.7893
2.1453	6.0	474	2.8231
2.1548	7.0	553	2.7538
2.1582	8.0	632	2.7992
2.1775	9.0	711	2.6910
2.1962	10.0	790	2.7582
2.226	11.0	869	2.7252
2.2153	12.0	948	2.7437
2.2421	13.0	1027	2.7776
2.2297	14.0	1106	2.7257
2.2265	15.0	1185	2.7722
2.2187	16.0	1264	2.7475
2.2272	17.0	1343	2.7317
2.2315	18.0	1422	2.6876
2.2619	19.0	1501	2.7518
2.2313	20.0	1580	2.7373

Framework versions

• Transformers 4.26.1
• Pytorch 1.13.1+cu116
• Datasets 2.10.1
• Tokenizers 0.13.2