distilbert-base-uncased-Regression-Simpsons_Plus_Others

This model is a fine-tuned version of distilbert-base-uncased on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.3754
• Mse: 0.3754
• Rmse: 0.6127
• Mae: 0.4651

Model description

This project works to predict the rating of episodes for the following TV shows:

• The Simpsons
• Brooklyn Nine Nine
• Seinfeld
• The Big Bang Theory
• 30 Rock
• Community
• Parks and Recreation
• The Office
• How I Met Your Mother
• Modern Family

For more information on how it was created, check out the following link: https://github.com/DunnBC22/NLP_Projects/blob/main/NLP%20Regression/NLP%20Regression%20-%20Simpsons%20Plus%20Other%20Series.ipynb

Intended uses & limitations

This model is intended to demonstrate my ability to solve a complex problem using technology.

Training and evaluation data

Data Sources:

• https://www.kaggle.com/datasets/mattbarty/the-simpsons-s1s32-imdb-scores-episode-info
• https://www.kaggle.com/datasets/maddyramsey/brookyln-nine-nine-imdb-ratings
• https://www.kaggle.com/datasets/hod101s/seinfeld-imdb-ratings
• https://www.kaggle.com/datasets/bcruise/big-bang-theory-episodes?select=big_bang_theory_imdb.csv
• https://www.kaggle.com/datasets/bcruise/30-rock-episode-data?select=30_rock_imdb.csv
• https://www.kaggle.com/datasets/imbenab/community-episodes-imdb-ratings
• https://www.kaggle.com/datasets/bcruise/parks-and-recreation-episode-data?select=parks_and_rec_imdb.csv
• https://www.kaggle.com/datasets/kapastor/the-office-imdb-ratings-per-episode
• https://www.kaggle.com/datasets/bcruise/how-i-met-your-mother-episodes-data?select=himym_imdb.csv
• https://www.kaggle.com/datasets/rprkh15/modern-family-dataset

Also, I pulled the episode description and rating from IMDb for the following TV shows:

• Two and a Half Men
• Young Sheldon
• Married... With Children
• Family Guy
• South Park
• That '70s Show
• It's Always Sunny in Philadelphia

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: 25

Training results

Training Loss	Epoch	Step	Validation Loss	Mse	Rmse	Mae
29.5977	1.0	51	7.9215	7.9215	2.8145	2.7032
4.4551	2.0	102	0.6728	0.6728	0.8202	0.6039
2.0068	3.0	153	0.6034	0.6034	0.7768	0.5882
1.8734	4.0	204	0.4423	0.4423	0.6651	0.4975
1.7607	5.0	255	0.3971	0.3971	0.6302	0.4725
1.6901	6.0	306	0.4005	0.4005	0.6328	0.4751
1.6525	7.0	357	0.4001	0.4001	0.6325	0.4766
1.6103	8.0	408	0.4278	0.4278	0.6541	0.4954
1.5659	9.0	459	0.3903	0.3903	0.6247	0.4618
1.4968	10.0	510	0.3987	0.3987	0.6314	0.4670
1.4983	11.0	561	0.4764	0.4764	0.6902	0.5324
1.4659	12.0	612	0.3913	0.3913	0.6256	0.4616
1.4532	13.0	663	0.4511	0.4511	0.6716	0.5153
1.4515	14.0	714	0.4009	0.4009	0.6332	0.4768
1.4506	15.0	765	0.4588	0.4588	0.6773	0.5160
1.4249	16.0	816	0.3940	0.3940	0.6277	0.4630
1.4254	17.0	867	0.4456	0.4456	0.6675	0.5084
1.4023	18.0	918	0.4517	0.4517	0.6721	0.5096
1.3754	19.0	969	0.4210	0.4210	0.6489	0.4869
1.3865	20.0	1020	0.4163	0.4163	0.6452	0.4830
1.3802	21.0	1071	0.4290	0.4290	0.6550	0.4904
1.4087	22.0	1122	0.4097	0.4097	0.6401	0.4745
1.3855	23.0	1173	0.4438	0.4438	0.6662	0.5027
1.3911	24.0	1224	0.4302	0.4302	0.6559	0.4906
1.3877	25.0	1275	0.4287	0.4287	0.6547	0.4887

Framework versions

• Transformers 4.22.1
• Pytorch 1.12.1
• Datasets 2.4.0
• Tokenizers 0.12.1

License Notice

This model is a fine-tuned derivative of a pretrained model. Users must comply with the original model license.

Dataset Notice

This model was fine-tuned on third-party datasets which may have separate licenses or usage restrictions.