metadata

language:
  - en
license: mit
base_model: gpt2
tags:
  - pytorch
  - GPT2ForSequenceClassification
  - generated_from_trainer
metrics:
  - accuracy
  - matthews_correlation
model-index:
  - name: GPT2-genre-detection
    results: []
library_name: transformers
pipeline_tag: text-classification
datasets:
  - datadrivenscience/movie-genre-prediction

GPT2-genre-detection

This model is a fine-tuned version of gpt2 on the datadrivenscience/movie-genre-prediction dataset. It achieves the following results on the evaluation set:

Loss: 1.5267
Accuracy: 0.4593
Matthews Correlation: 0.1010

description

Data-Driven Science organized a competition where in the goal was to fine tune a model that can predict the genre of a movie from a given synopsis. There were a total of 10 genres as follows:

{
    "0": "horror",
    "1": "adventure",
    "2": "action",
    "3": "crime",
    "4": "mystery",
    "5": "family",
    "6": "scifi",
    "7": "thriller",
    "8": "fantasy",
    "9": "romance"
  }

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 2e-05
train_batch_size: 8
eval_batch_size: 32
seed: 85855289
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: cosine
lr_scheduler_warmup_ratio: 0.05
num_epochs: 2

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Matthews Correlation
1.5765	1.0	10125	1.5562	0.4589	0.0899
1.5058	2.0	20250	1.5267	0.4593	0.1010

Framework versions

Transformers 4.36.2
Pytorch 2.1.0+cu121
Datasets 2.16.0
Tokenizers 0.15.0