Keep Training
Instructions for Use
Training deep learning models is typically a time-consuming process. If the training process is interrupted, the results of previous training will be lost, wasting computational resources. If the training procedure is interrupted for some reason, this tool can be used to resume the training procedure based on the training log path. In addition, the Keep Training feature enables log continuation. After training is complete, users can open TensorBoard to see the complete training curve.
The model obtained through Keep Training is saved by default to the "Model Storage Path" set in the Model Training tool before the interruption. The model name remains consistent with the "Model Name" in the Model Training tool.
Parameter Description
- Training Log Path: Select the model training log folder where the training procedure was interrupted. The parameters from before the interruption are recorded in the log folder and can be automatically reproduced during Keep Training.
- Training Model Usage: Must remain consistent with the original training model's purpose.
- Processor Type: You can use the computer's central processing unit (CPU) or graphics processing unit (GPU) to process data. GPUs have faster computation speeds.
- GPU ID: Specify the GPU ID for data processing. Default is 0. CPU inference is fixed at -1. If using GPU inference, you need to query the GPU ID via the "nvidia-smi" command in the system command line. The GPU ID should be based on the query results. If there are multiple GPUs, you can specify the GPU ID for processing data. To use multiple GPUs, separate them with English commas ",", e.g., "0,2,3" indicates using GPUs with IDs 0, 2, and 3.
Related Topics