mirror of
https://github.com/Vision-CAIR/MiniGPT-4.git
synced 2025-04-05 18:40:46 +00:00
48 lines
1.0 KiB
YAML
48 lines
1.0 KiB
YAML
|
model:
|
||
|
|
arch: minigpt4
|
||
|
|
model_type: pretrain_llama2
|
||
|
|
max_txt_len: 160
|
||
|
|
end_sym: "</s>"
|
||
|
|
low_resource: True
|
||
|
|
prompt_template: '[INST] {} [/INST] '
|
||
|
|
ckpt: 'please set this value to the path of pretrained checkpoint'
|
||
|
|
|
||
|
|
datasets:
|
||
|
|
coco_vqa: # not used
|
||
|
|
type: eval
|
||
|
|
vis_processor:
|
||
|
|
eval:
|
||
|
|
name: "blip2_image_eval"
|
||
|
|
image_size: 336
|
||
|
|
text_processor:
|
||
|
|
eval:
|
||
|
|
name: "blip_caption"
|
||
|
|
|
||
|
|
run:
|
||
|
|
task: vqa
|
||
|
|
# optimization-specific
|
||
|
|
batch_size_train: 32
|
||
|
|
batch_size_eval: 128
|
||
|
|
# if do not use OCR token: batch=32, Evaluation Total time: 0:23:47 (9.0954 s / it)
|
||
|
|
use_ocr: True
|
||
|
|
# use OCR token: batch_size=16; Evaluation Total time: 0:36:34 (7.0116 s / it)
|
||
|
|
num_workers: 8
|
||
|
|
|
||
|
|
# inference-specific
|
||
|
|
max_len: 10
|
||
|
|
min_len: 1
|
||
|
|
num_beams: 1
|
||
|
|
inference_method: "generate"
|
||
|
|
prompt: "Question: {} Short answer:"
|
||
|
|
|
||
|
|
seed: 42
|
||
|
|
output_dir: "results"
|
||
|
|
|
||
|
|
evaluate: True
|
||
|
|
test_splits: ["val"]
|
||
|
|
|
||
|
|
# distribution-specific
|
||
|
|
device: "cuda"
|
||
|
|
world_size: 1
|
||
|
|
dist_url: "env://"
|
||
|
|
distributed: True
|