Loading README.md +10 −4 Changes for README.md: 10 added lines, 4 removed lines. Original line number Diff line number Diff line Loading @@ -12,22 +12,28 @@ Once you have the data, it has to be organized in the following structure: ├── ... ├── data │ ├── transcripts │ │ ├── 300_TRANSCRIPT.csv │ │ ├── 301_TRANSCRIPT.csv │ │ ├── 302_TRANSCRIPT.csv │ │ └── ... │ ├── train_split_Depression_AVEC2017.csv │ ├── dev_split_Depression_AVEC2017.csv │ └── test_split_Depression_AVEC2017.csv └── ... After, you have to run all the cells of the `TranscriptsToCONLLU.ipynb` and `prepare_data.py`. After that, you have to run `prepare_data.py`. ### Training the model To train the model, run the following command: Training is done with [HuggingFace 🤗 Accelerate](https://huggingface.co/docs/accelerate/index) library. Before training, run `accelerate config` to setup your environment. Set the training parameters in the `config.py` file. To start training, run: ``` python ./train_bert_dist.py --conf_file config_bert.ini --seed 0 --chunking_method lines --data_oversampling none --loss_averaging none --attention_type hierarchical --pooling mean --bidirectional --batch_size 8 --num_iters 200 --num_classes 8 --multilabel --binary_only --no_regularization_loss --save_every_epoch accelerate launch train.py ``` ### Testing the model Loading Loading
README.md +10 −4 Changes for README.md: 10 added lines, 4 removed lines. Original line number Diff line number Diff line Loading @@ -12,22 +12,28 @@ Once you have the data, it has to be organized in the following structure: ├── ... ├── data │ ├── transcripts │ │ ├── 300_TRANSCRIPT.csv │ │ ├── 301_TRANSCRIPT.csv │ │ ├── 302_TRANSCRIPT.csv │ │ └── ... │ ├── train_split_Depression_AVEC2017.csv │ ├── dev_split_Depression_AVEC2017.csv │ └── test_split_Depression_AVEC2017.csv └── ... After, you have to run all the cells of the `TranscriptsToCONLLU.ipynb` and `prepare_data.py`. After that, you have to run `prepare_data.py`. ### Training the model To train the model, run the following command: Training is done with [HuggingFace 🤗 Accelerate](https://huggingface.co/docs/accelerate/index) library. Before training, run `accelerate config` to setup your environment. Set the training parameters in the `config.py` file. To start training, run: ``` python ./train_bert_dist.py --conf_file config_bert.ini --seed 0 --chunking_method lines --data_oversampling none --loss_averaging none --attention_type hierarchical --pooling mean --bidirectional --batch_size 8 --num_iters 200 --num_classes 8 --multilabel --binary_only --no_regularization_loss --save_every_epoch accelerate launch train.py ``` ### Testing the model Loading
