Closed howiejayz closed 1 year ago
Hello! I am trying to run the PerfZero RetinaNet benchmark.
However, I found in the model file the path for the COCO dataset for PerfZero is written as google bucket path:
COCO_TRAIN_DATA = 'gs://tf-perfzero-data/coco/train*' COCO_EVAL_DATA = 'gs://tf-perfzero-data/coco/val*' COCO_EVAL_JSON = 'gs://tf-perfzero-data/coco/instances_val2017.json' RESNET_CHECKPOINT_PATH = 'gs://cloud-tpu-checkpoints/retinanet/resnet50-checkpoint-2018-02-07'
which seems not to be publically available.
Is there a way to run this benchmark with a propoer dataset?
Solved by using the script in https://github.com/tensorflow/tpu
Hello! I am trying to run the PerfZero RetinaNet benchmark.
However, I found in the model file the path for the COCO dataset for PerfZero is written as google bucket path:
which seems not to be publically available.
Is there a way to run this benchmark with a propoer dataset?