FedML-AI
fedcv-image_segmentation problem
Hi ,
I have already install FedML (via conda) and try to run the example of fedcv-image_segmentation .
I would like to run 2 client & 1 server with simulation on single machine.
But i met some problems when executing bash run_simulation.sh 1 , there is console message:
======== FedML (https://fedml.ai) ========
FedML version: 0.7.286
Execution path:/home/pomchi/miniconda3/envs/fedml/lib/python3.7/site-packages/fedml/__init__.py
======== Running Environment ========
OS: Linux-5.4.0-122-generic-x86_64-with-debian-buster-sid
Hardware: x86_64
Python version: 3.7.13 (default, Mar 29 2022, 02:18:16)
[GCC 7.5.0]
PyTorch version: 1.12.1+cu102
MPI4py is installed
======== CPU Configuration ========
The CPU usage is : 0%
Available CPU Memory: 60.0 G / 62.521270751953125G
======== GPU Configuration ========
NVIDIA GPU Info: <pynvml.nvml.LP_struct_c_nvmlDevice_t object at 0x7f91302019e0>
Available GPU memory: 23.6 G / 24.0G
[]
args.client_id_list = None
training_type != FEDML_TRAINING_PLATFORM_CROSS_DEVICE and training_type != FEDML_TRAINING_PLATFORM_CROSS_SILO
[FedML-Server(0) @device-id-0] [Mon, 08 Aug 2022 10:18:41] [INFO] [__init__.py:531:setup_log_mqtt_mgr] mlops log metrics agent config: mqtt.fedml.ai,1883
ERROR:/home/pomchi/fedml-server/fedml/logs/fedml-run-0-edge-0.log:Uncaught exception
Traceback (most recent call last):
File "main_fedml_image_segmentation.py", line 18, in <module>
dataset, class_num = load_data(args)
File "/home/pomchi/shuyu/FedML/python/app/fedcv/image_segmentation/data/data_loader.py", line 12, in load
return load_synthetic_data(args)
File "/home/pomchi/shuyu/FedML/python/app/fedcv/image_segmentation/data/data_loader.py", line 54, in load_synthetic_data
batch_size=args.batch_size,
TypeError: load_partition_data_cityscapes() missing 1 required positional argument: 'image_size'
My device only have one GPU ( NVIDIA GeForce RTX 3090 ), and i change the setting of fedml_config.yaml
train_args:
client_num_in_total: 2
client_num_per_round: 2
device_args:
worker_num: 1
using_gpu: true
gpu_mapping_file: config/gpu_mapping.yaml
gpu_mapping_key: mapping_default
gpu_mapping.yaml
mapping_default:
pomchi-X299-UD4: [1]
I'm not sure if this setting is correct
Another question is there is no pycocotools module when i import ( at image_segmentation/data/data_loader.py)
from .coco.segmentation.data_loader import load_partition_data_coco_segmentation
Traceback (most recent call last):
File "main_fedml_image_segmentation.py", line 4, in <module>
from data.data_loader import load as load_data
File "/home/pomchi/shuyu/FedML/python/app/fedcv/image_segmentation/data/data_loader.py", line 6, in <module>
from .coco.segmentation.data_loader import load_partition_data_coco_segmentation
File "/home/pomchi/shuyu/FedML/python/app/fedcv/image_segmentation/data/coco/segmentation/data_loader.py", line 11, in <module>
from .dataset import CocoSegmentation
File "/home/pomchi/shuyu/FedML/python/app/fedcv/image_segmentation/data/coco/segmentation/dataset.py", line 12, in <module>
import pycocotools.mask as coco_mask
ModuleNotFoundError: No module named 'pycocotools'
Any hint and suggestion is highly appreciated !!!
Hi @shuyu0914! For FedCV image segmentation, you need to run the bash script bootstrap.sh in config to install dependencies. You can see the latest code here, and we will release it soon!
