forked from fudan-generative-vision/dynamicPDB
-
Notifications
You must be signed in to change notification settings - Fork 0
/
run_eval.sh
23 lines (23 loc) · 812 Bytes
/
run_eval.sh
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
# atlas frame 1
model_path=../DFOLDv2_res/result_v2_selfdata/ckpt/force_jw_full_gen_10/10D_08M_2024Y_15h_12m_12s/step_9603.pth
start_idx=910000
python eval_DFOLD_dynamics.py eval.weights_path=$model_path \
experiment.use_ddp=False \
data.frame_time=2 \
model.cfg_drop_rate=0.0 \
experiment.batch_size=1 \
data.test_csv_path=$test_data \
experiment.noise_scale=0.1 \
data.frame_sample_step=1 \
data.fix_sample_start=$start_idx \
model.ipa.temporal=False \
diffuser.r3.coordinate_scaling=1.0 \
eval.name=debug \
data.test_csv_path=test.csv
# eval.name=exp_start_$start_idx \
# data.frame_sample_step=50 \
# model.ipa.temporal=False \
# model.ipa.temporal=True model.ipa.temporal_position_encoding=True \
# experiment.base_root=../DFOLDv2_res/result_v2_extension \
# data.time_norm=True
# model.ipa.num_blocks=1 \