search

NVIDIA / flownet2-pytorch

Pytorch implementation of FlowNet 2.0: Evolution of Optical Flow Estimation with Deep Networks
Other
3.09k stars 739 forks source link

checkpoint path #199

Open YoungJ-Baek opened 4 years ago

YoungJ-Baek commented 4 years ago

after I command `python main.py --inference --model FlowNet2 --save_flow --inference_dataset MpiSintelClean \

--inference_dataset_root /path/to/mpi-sintel/clean/dataset \ --resume /path/to/checkpoints `

I get `Parsing Arguments [0.020s] batch_size: 8 [0.020s] crop_size: [256, 256] [0.020s] fp16: False [0.020s] fp16_scale: 1024.0 [0.020s] gradient_clip: None [0.020s] inference: True [0.020s] inference_batch_size: 1 [0.020s] inference_dataset: MpiSintelClean [0.020s] inference_dataset_replicates: 1 [0.020s] inference_dataset_root: /path/to/mpi-sintel/clean/dataset [0.020s] inference_n_batches: -1 [0.020s] inference_size: [-1, -1] [0.020s] inference_visualize: False [0.020s] log_frequency: 1 [0.020s] loss: L1Loss [0.020s] model: FlowNet2 [0.020s] model_batchNorm: False [0.020s] model_div_flow: 20.0 [0.020s] name: run [0.020s] no_cuda: False [0.020s] number_gpus: 1 [0.020s] number_workers: 8 [0.020s] optimizer: Adam [0.020s] optimizer_amsgrad: False [0.020s] optimizer_betas: (0.9, 0.999) [0.020s] optimizer_eps: 1e-08 [0.021s] optimizer_lr: 0.001 [0.021s] optimizer_weight_decay: 0 [0.021s] render_validation: False [0.021s] resume: /path/to/checkpoints [0.021s] rgb_max: 255.0 [0.021s] save: ./work [0.021s] save_flow: True [0.021s] schedule_lr_fraction: 10 [0.021s] schedule_lr_frequency: 0 [0.021s] seed: 1 [0.021s] skip_training: False [0.021s] skip_validation: False [0.021s] start_epoch: 1 [0.021s] total_epochs: 10000 [0.021s] train_n_batches: -1 [0.021s] training_dataset: MpiSintelFinal [0.021s] training_dataset_replicates: 1 [0.021s] training_dataset_root: ./MPI-Sintel/flow/training [0.021s] validation_dataset: MpiSintelClean [0.021s] validation_dataset_replicates: 1 [0.021s] validation_dataset_root: ./MPI-Sintel/flow/training [0.021s] validation_frequency: 5 [0.021s] validation_n_batches: -1 [0.023s] Operation finished

Source Code Current Git Hash: b'21cd424e05b52c71c2079e405575a32a0726c92f'

Initializing Datasets [0.000s] Operation finished

Building FlowNet2 model [2.041s] Effective Batch Size: 8 [2.042s] Number of parameters: 162518834 [2.042s] Initializing CUDA [4.549s] Parallelizing [4.550s] nt founNo checkpoid at '/path/to/checkpoints' [4.550s] Operation failed `

How can I solve nt founNo checkpoid at '/path/to/checkpoints' problem? Do I have to make checkpoints folder in flownet2-pytorch folder? Also, I want to know is there anything I should do except things in readme

GUWOGANSHOU commented 3 years ago

You should download the checkpoints and put it in the right location.

GUWOGANSHOU commented 3 years ago

May be this blog will be helpful.https://cloud.tencent.com/developer/article/1461933