Instructions to use dzungpham/graphcodebert-code-classification with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use dzungpham/graphcodebert-code-classification with Transformers:
# Load model directly from transformers import AutoModel model = AutoModel.from_pretrained("dzungpham/graphcodebert-code-classification", dtype="auto") - Notebooks
- Google Colab
- Kaggle
graphcodebert-code-classification / fourier-spectral-norm-classifier /checkpoint-1500 /train_config.yaml
| TEST: false | |
| model_name: /kaggle/input/models/dzung271828/microsoft-graphcodebert-base/transformers/default/1 | |
| output_dir: training/fourier-spectral-norm-classifier/ | |
| num_epochs: 5 | |
| max_steps: -1 | |
| batch_size: 512 | |
| learning_rate: 1.0e-06 | |
| max_length: 512 | |
| num_labels: 2 | |
| use_wandb: false | |
| freeze_base: true | |
| loss_type: ce | |
| focal_alpha: 1.0 | |
| focal_gamma: 2.0 | |
| r_drop_alpha: 6.0 | |
| infonce_temperature: 0.07 | |
| infonce_weight: 0.5 | |
| seed: 42 | |
| resume_from_checkpoint: null | |
| save_steps: 500 | |
| eval_steps: 500 | |
| logging_steps: 5 | |
| label_smoothing: 0.5 | |
| adversarial_epsilon: 0.5 | |
| use_swa: true | |
| swa_start_epoch: 0 | |
| swa_lr: 1.0e-06 | |
| data_augmentation: true | |
| aug_rename_prob: 0.7 | |
| aug_format_prob: 0.7 | |
| weight_decay: 0.1 | |
| mixup_alpha: 1.0 | |
| low_pass_keep_ratio: 0.5 | |
| freq_consistency_weight: 0.2 | |
| use_mixcode: true | |
| use_fgm: true | |
| fgm_freq: 5 | |
| use_r_drop: true | |
| use_freq_consistency_loss: true | |
| use_attn_spectral: false | |
| attn_spectral_weight: 0.1 | |
| attn_spectral_cutoff_ratio: 0.25 | |
| hidden_dropout_prob: 0.3 | |
| attention_probs_dropout_prob: 0.3 | |
| classifier_dropout: 0.4 | |
| device: cuda | |
| torch_compile: true | |
| cache_dir: ./tokenized_cache | |
| use_swa_actual: true | |
| use_fgm_actual: true | |
| use_r_drop_actual: true | |
| use_mixcode_actual: true | |
| use_attn_spectral_actual: false | |
| use_freq_consistency_loss_actual: true | |
| use_spectral_norm: true | |