SPG: Sequential Policy Gradient: Lightweight Reinforcement Learning for Model Performance

🚀 If you're using Jupyter or Colab, you can follow the demo and run it on a single GPU:

Model Zoo: Adaptive Hyperparameter Optimization (HPO) via SPG Algorithm

• GLUE (Text classification: BERT on CoLA, SST-2, MRPC, QQP, QNLI, and RTE task)
• SQuAD (Question answering: BERT)
• SUPERB (Speech classification: Wav2Vec2 for Audio Classification (AC))

Requirements

1. Install torch>=2.0.0+cu118.
2. To install other pip packages:
   setup

   1 cd examples 2 pip install -r requirements.txt
3. Prepare the ImageNet dataset manually and place it in /path/to/imagenet. For image classification examples, pass the argument --data-path=/path/to/imagenet to the training script. The extracted dataset directory should follow this structure:
   setup

   1/path/to/imagenet/: 2 train/: 3 n01440764: 4 n01440764_18.JPEG ... 5 n01443537: 6 n01443537_2.JPEG ... 7 val/: 8 n01440764: 9 ILSVRC2012_val_00000293.JPEG ... 10 n01443537: 11 ILSVRC2012_val_00000236.JPEG ...
4. Prepare the MS-COCO 2017 dataset manually and place it in /path/to/coco. For semantic segmentation examples, pass the argument --data-path=/path/to/coco to the training script. The extracted dataset directory should follow this structure:
   setup

   1/path/to/coco/: 2 annotations: 3 many_json_files.json ... 4 train2017: 5 000000000009.jpg ... 6 val2017: 7 000000000139.jpg ...
5. Prepare the GSM8K dataset manually and place it in /path/to/gsm8k. For language modeling examples, pass the argument --data-path=/path/to/gsm8k to the training script. The extracted dataset directory should follow this structure:
   setup

   1/path/to/gsm8k/: 2 train.parquet 3 test.parquet
6. For 🗣️ Keyword Spotting subset, Common Language, SQuAD, Common Voice, GLUE and WMT datasets, manual downloading is not required — they will be automatically loaded via the Hugging Face Datasets library when running our audio-classification, question-answering, speech-recognition, text-classification, or translation examples.