Tutorial on Training and Testing on other datasets

The main model architecture difference with V2X-Real and other datasets such as OPV2V and DAIR-V2X is the usage of multi-class classification in V2X-Real dataset compared with the single-class classification (only vehicle detection) used in other datasets. Please be aware that we arrange separate python files for running on different datasets. Please refer to Tutorial of Baseline Training and Inference on V2X-Real dataset, Tutorial of Codebook Learning on V2X-Real dataset, and Tutorial of PTQ on V2X-Real dataset before reading this documentation.

Train the model

We uses yaml file to configure all the parameters for training. We have already configured the dataset and core method in the yaml file. To train your own model from scratch or a continued checkpoint, run the following commonds:

python opencood/tools/train.py -y ${CONFIG_FILE} [--model_dir ${CHECKPOINT_FOLDER}]

Codebook Learning

Stage 1: Train the full-precision model

We uses yaml file to configure all the parameters for training. To train your own model from scratch or a continued checkpoint, run the following commonds (with opv2v as examples):

python ./opencood/tools/train.py -y ./opencood/hypes_yaml/opv2v/Codebook/Pyramid/lidar_pyramid_stage1.yaml

Stage 2: Codebook-only Training

python ./opencood/tools/train_stage2.py --hypes_yaml ./opencood/hypes_yaml/opv2v/Codebook/Pyramid/lidar_pyramid_stage2.yaml --stage1_model your_path_to_stage1_model.pth

• stage1_model points to the pretrained checkpoint from Stage 1.

Stage 3: Codebook Co-training

python ./opencood/tools/train_stage3.py --hypes_yaml ./opencood/hypes_yaml/opv2v/Codebook/Pyramid/lidar_pyramid_stage3.yaml --stage2_model your_path_to_stage2_model.pth

• stage2_model points to the pretrained checkpoint from Stage 2.

Test the model

Note that we use inference.py instead of inference_mc.py for evaluating on other datasets:

python opencood/tools/inference.py --model_dir ${CHECKPOINT_FOLDER} [--fusion_method intermediate]

Post-Training Quantization (PTQ)

Note that we use inference_quant.py instead of inference_mc_quant.py for evaluating on other datasets:

python opencood/tools/inference_quant.py ${CHECKPOINT_FOLDER} [--fusion_method intermediate] --num_cali_batches 16 --n_bits_w 8 --n_bits_a 8 --iters_w 5000

• num_cali_batches refers to the size of the calibration dataset.
• n_bits_w refers to the bitwidth for weight quantization.
• n_bits_a refers to the bitwidth for activation quantization.
• iters_w refers to the number of calibration steps.

Notes:

• You could refer to /scripts folder for more examples.
• When modifying yaml files in ./opencood/hypes_yaml, pay attention to the assignment_path, core_method, dataset fields to ensure the consistency of the functions and corresponding datasets.