rdj-034/lab2_efficient

1---2library_name: transformers3language:4- en5- fr6tags:7- translation8- seq2seq9metrics:10- bleu11model-index:12- name: lab2_efficient13  results:14  - task:15      type: translation16    dataset:17      name: kde418      type: kde419    metrics:20    - type: bleu21      value: 44.11322      name: BLEU23    - type: loss24      value: 1.354625      name: Loss26---27 28# lab2_efficient29 30## Hyperparameters31- learning_rate: 2e-532- per_device_train_batch_size: 12833- effective_batch_size: 12834- gradient_accumulation_steps: 135- weight_decay: 0.136- optimizer: adamw_torch37- fp16: True38- gradient_checkpointing: True39- lr_scheduler: cosine40- warmup_ratio: 0.141- max_steps: 10042 43## Results44| Metric | Value |45|--------|-------|46| BLEU | 44.113 |47| Eval Loss | 1.3546 |48| Train Steps | 100 |49| Epoch | 0.0615 |50