|
|
|
|
|
|
|
|
|
|
|
model:
|
|
arch: blip2
|
|
model_type: coco
|
|
use_grad_checkpoint: False
|
|
|
|
datasets:
|
|
coco_retrieval:
|
|
vis_processor:
|
|
train:
|
|
name: "blip_image_train"
|
|
image_size: 364
|
|
eval:
|
|
name: "blip_image_eval"
|
|
image_size: 364
|
|
text_processor:
|
|
train:
|
|
name: "blip_caption"
|
|
eval:
|
|
name: "blip_caption"
|
|
|
|
|
|
|
|
run:
|
|
task: retrieval
|
|
|
|
|
|
num_workers: 4
|
|
batch_size_train: 16
|
|
batch_size_eval: 32
|
|
|
|
train_splits: ["train"]
|
|
valid_splits: ["val"]
|
|
test_splits: ["test"]
|
|
|
|
|
|
device: "cuda"
|
|
world_size: 1
|
|
dist_url: "env://"
|
|
distributed: True
|
|
use_dist_eval_sampler: False
|
|
|
|
|
|
k_test: 128
|
|
|
|
|
|
seed: 42
|
|
output_dir: "output/BLIP2/Retrieval_COCO"
|
|
|
|
evaluate: True
|
|
|