Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

update config to increase mAP for model, result increase mAP by 87.5% #4

Open
wants to merge 3 commits into
base: denis-update-to-setup-guide
Choose a base branch
from
Open

update config to increase mAP for model, result increase mAP by 87.5% #4

Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
69e5fd6
update config to increase mAP for model, result increase mAP by 87.5%
interactivetech Aug 28, 2023
64cc149
update config to increase mAP for model, result increase mAP by 87.5%
interactivetech Aug 28, 2023
64b4e3e
update config to increase mAP for model, result increase mAP by 87.5%
interactivetech Aug 28, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
11 changes: 6 additions & 5 deletions examples/object-detection/experiment/const.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -15,20 +15,21 @@ profiling:
begin_on_batch: 0
end_after_batch: null
hyperparameters:
lr: 0.02
lr: .0004261457
momentum: 0.9
global_batch_size: 16
weight_decay: 1.0e-4
gamma: 0.1
warmup: linear
warmup_iters: 200
warmup_ratio: 0.001
warmup_ratio: 0.724677 # The fraction between 0-1 that defines the percentage of lr you want to start the lr at
pretrained_model: "https://storage.googleapis.com/ai-at-scale-pdk-assets/sample-data/pdk-object-detection/pretrained-model/frcnn_xview.pth"
#finetune_ckpt: "/lus/aiholus1/disk/andrew.mendez/model_479.pth"
step1: 504 # 14 epochs: 14*36 == 504
step2: 540 # 15 epochs: 15*36 == 540
model: fasterrcnn_resnet50_fpn
num_workers: 4
#device: cuda
environment:
image: mendeza/obj-det-pdk-train-env:0.0.2
environment_variables:
Expand All @@ -38,15 +39,15 @@ environment:

scheduling_unit: 400
min_validation_period:
batches: 36 # For training
batches: 5 # For training

searcher:
name: single
metric: mAP
smaller_is_better: true
max_length:
batches: 2 # 1*(579/16) = 1*36
records_per_epoch: 32 # 32 records / 16
batches: 10 # 1*(9/16) = 1*1
records_per_epoch: 9 # 9 records / 16
resources:
slots_per_trial: 1
resource_pool: gpu-pool
Expand Down