Skip to content

Commit fc1945d

Browse files
HydrogenSulfateHydrogenSulfate
authored
[Example] Add ldc 2d Re1000 and Re3200 example (#924)
* update piranet for ldc re3200 case(WIP) * add ldc_2d_Re1000 plain example * add ldc_2d_Re3200_piratenet and ldc_2d_Re3200_sota * refine code * refine docstrings * add missing blank line for docstrings * refine docstrings * replace lowercase to uppercase for first word in sentence * add ldc_2d_re3200_piratenet and ldc_2d_re3200_sota * remove redundant files * fix linenum * fix doctest * fix for allen_cahn
1 parent 04d54dd commit fc1945d

Some content is hidden

Large Commits have some content hidden by default. Use the searchbox below for content that may be hidden.

59 files changed

+1729
-347
lines changed

‎docs/zh/examples/allen_cahn.md

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -40,7 +40,7 @@
4040

4141
| 预训练模型 | 指标 |
4242
|:--| :--|
43-
| [allen_cahn_piratenet_pretrained.pdparams](https://paddle-org.bj.bcebos.com/paddlescience/models/AllenCahn/allen_cahn_piratenet_pretrained.pdparams) | L2Rel.u: 8.32403e-06 |
43+
| [allen_cahn_piratenet_pretrained.pdparams](https://paddle-org.bj.bcebos.com/paddlescience/models/AllenCahn/allen_cahn_piratenet_pretrained.pdparams) | L2Rel.u: 1.2e-05 |
4444

4545
## 1. 背景简介
4646

‎docs/zh/examples/ldc2d_steady.md

Lines changed: 214 additions & 162 deletions
Large diffs are not rendered by default.

‎examples/ldc/conf/ldc_2d_Re1000_plain.yaml

Lines changed: 88 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,88 @@
1+
defaults:
2+
- ppsci_default
3+
- TRAIN: train_default
4+
- TRAIN/ema: ema_default
5+
- TRAIN/swa: swa_default
6+
- EVAL: eval_default
7+
- INFER: infer_default
8+
- hydra/job/config/override_dirname/exclude_keys: exclude_keys_default
9+
- _self_
10+
11+
hydra:
12+
run:
13+
# dynamic output directory according to running time and override name
14+
dir: outputs_ldc_2d_Re1000_plain/${now:%Y-%m-%d}/${now:%H-%M-%S}/${hydra.job.override_dirname}
15+
job:
16+
name: ${mode} # name of logfile
17+
chdir: false # keep current working directory unchanged
18+
callbacks:
19+
init_callback:
20+
_target_: ppsci.utils.callbacks.InitCallback
21+
sweep:
22+
# output directory for multirun
23+
dir: ${hydra.run.dir}
24+
subdir: ./
25+
26+
# general settings
27+
mode: train # running mode: train/eval
28+
seed: 42
29+
output_dir: ${hydra:run.dir}
30+
log_freq: 100
31+
use_tbd: false
32+
33+
# working conditions
34+
Re: [100, 400, 1000]
35+
epochs: [20, 40, 140]
36+
EVAL_DATA_PATH: ./data/ldc_Re1000.mat
37+
38+
# model settings
39+
MODEL:
40+
input_keys: ["x", "y"]
41+
output_keys: ["u", "v", "p"]
42+
num_layers: 4
43+
hidden_size: 256
44+
activation: tanh
45+
46+
# training settings
47+
TRAIN:
48+
epochs: 20
49+
iters_per_epoch: 1000
50+
save_freq: 10
51+
eval_during_train: true
52+
eval_freq: 1
53+
lr_scheduler:
54+
epochs: ${sum:${epochs}}
55+
iters_per_epoch: ${TRAIN.iters_per_epoch}
56+
learning_rate: 1.0e-3
57+
gamma: 0.9
58+
decay_steps: 2000
59+
by_epoch: false
60+
batch_size:
61+
pde: 1024
62+
bc: 256
63+
pretrained_model_path: null
64+
checkpoint_path: null
65+
66+
# evaluation settings
67+
EVAL:
68+
pretrained_model_path: null
69+
eval_with_no_grad: true
70+
batch_size: 4096
71+
72+
# inference settings
73+
INFER:
74+
pretrained_model_path: null
75+
export_path: ./inference/ldc_2d_re3200
76+
pdmodel_path: ${INFER.export_path}.pdmodel
77+
pdiparams_path: ${INFER.export_path}.pdiparams
78+
onnx_path: ${INFER.export_path}.onnx
79+
device: gpu
80+
engine: native
81+
precision: fp32
82+
ir_optim: true
83+
min_subgraph_size: 5
84+
gpu_mem: 2000
85+
gpu_id: 0
86+
max_batch_size: 1024
87+
num_cpu_threads: 10
88+
batch_size: 1024

‎examples/ldc/conf/ldc_2d_Re3200_piratenet.yaml

Lines changed: 99 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,99 @@
1+
defaults:
2+
- ppsci_default
3+
- TRAIN: train_default
4+
- TRAIN/ema: ema_default
5+
- TRAIN/swa: swa_default
6+
- EVAL: eval_default
7+
- INFER: infer_default
8+
- hydra/job/config/override_dirname/exclude_keys: exclude_keys_default
9+
- _self_
10+
11+
hydra:
12+
run:
13+
# dynamic output directory according to running time and override name
14+
dir: outputs_ldc_2d_Re3200_piratenet/${now:%Y-%m-%d}/${now:%H-%M-%S}/${hydra.job.override_dirname}
15+
job:
16+
name: ${mode} # name of logfile
17+
chdir: false # keep current working directory unchanged
18+
callbacks:
19+
init_callback:
20+
_target_: ppsci.utils.callbacks.InitCallback
21+
sweep:
22+
# output directory for multirun
23+
dir: ${hydra.run.dir}
24+
subdir: ./
25+
26+
# general settings
27+
mode: train # running mode: train/eval
28+
seed: 42
29+
output_dir: ${hydra:run.dir}
30+
log_freq: 100
31+
use_tbd: false
32+
33+
# working conditions
34+
Re: [100, 400, 1000, 1600, 3200]
35+
epochs: [10, 20, 50, 50, 500]
36+
EVAL_DATA_PATH: ./data/ldc_Re3200.mat
37+
38+
# model settings
39+
MODEL:
40+
input_keys: ["x", "y"]
41+
output_keys: ["u", "v", "p"]
42+
num_blocks: 4
43+
hidden_size: 256
44+
activation: tanh
45+
fourier:
46+
scale: 15.0
47+
dim: 256
48+
random_weight:
49+
mean: 1.0
50+
std: 0.1
51+
52+
# training settings
53+
TRAIN:
54+
epochs: 10
55+
iters_per_epoch: 1000
56+
save_freq: 100
57+
eval_during_train: true
58+
eval_freq: 1
59+
lr_scheduler:
60+
epochs: ${sum:${epochs}}
61+
iters_per_epoch: ${TRAIN.iters_per_epoch}
62+
learning_rate: 1.0e-3
63+
gamma: 0.9
64+
decay_steps: 10000
65+
warmup_epoch: 5
66+
by_epoch: false
67+
batch_size:
68+
pde: 4096
69+
bc: 256
70+
pretrained_model_path: null
71+
checkpoint_path: null
72+
grad_norm:
73+
update_freq: 1000
74+
momentum: 0.9
75+
init_weights: [10, 1, 1, 100, 100]
76+
77+
# evaluation settings
78+
EVAL:
79+
pretrained_model_path: null
80+
eval_with_no_grad: true
81+
batch_size: 4096
82+
83+
# inference settings
84+
INFER:
85+
pretrained_model_path: null
86+
export_path: ./inference/ldc_2d_re3200
87+
pdmodel_path: ${INFER.export_path}.pdmodel
88+
pdiparams_path: ${INFER.export_path}.pdiparams
89+
onnx_path: ${INFER.export_path}.onnx
90+
device: gpu
91+
engine: native
92+
precision: fp32
93+
ir_optim: true
94+
min_subgraph_size: 5
95+
gpu_mem: 2000
96+
gpu_id: 0
97+
max_batch_size: 1024
98+
num_cpu_threads: 10
99+
batch_size: 1024

‎examples/ldc/conf/ldc_2d_Re3200_sota.yaml

Lines changed: 97 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,97 @@
1+
defaults:
2+
- ppsci_default
3+
- TRAIN: train_default
4+
- TRAIN/ema: ema_default
5+
- TRAIN/swa: swa_default
6+
- EVAL: eval_default
7+
- INFER: infer_default
8+
- hydra/job/config/override_dirname/exclude_keys: exclude_keys_default
9+
- _self_
10+
11+
hydra:
12+
run:
13+
# dynamic output directory according to running time and override name
14+
dir: outputs_ldc_2d_Re3200_sota/${now:%Y-%m-%d}/${now:%H-%M-%S}/${hydra.job.override_dirname}
15+
job:
16+
name: ${mode} # name of logfile
17+
chdir: false # keep current working directory unchanged
18+
callbacks:
19+
init_callback:
20+
_target_: ppsci.utils.callbacks.InitCallback
21+
sweep:
22+
# output directory for multirun
23+
dir: ${hydra.run.dir}
24+
subdir: ./
25+
26+
# general settings
27+
mode: train # running mode: train/eval
28+
seed: 42
29+
output_dir: ${hydra:run.dir}
30+
log_freq: 100
31+
use_tbd: false
32+
33+
# working conditions
34+
Re: [100, 400, 1000, 3200]
35+
epochs: [50, 50, 100, 500]
36+
EVAL_DATA_PATH: ./data/ldc_Re1000.mat
37+
38+
# model settings
39+
MODEL:
40+
input_keys: ["x", "y"]
41+
output_keys: ["u", "v", "p"]
42+
num_layers: 5
43+
hidden_size: 256
44+
activation: tanh
45+
fourier:
46+
scale: 10.0
47+
dim: 128
48+
random_weight:
49+
mean: 1.0
50+
std: 0.1
51+
52+
# training settings
53+
TRAIN:
54+
epochs: 50
55+
iters_per_epoch: 1000
56+
save_freq: 100
57+
eval_during_train: true
58+
eval_freq: 1
59+
lr_scheduler:
60+
epochs: ${sum:${epochs}}
61+
iters_per_epoch: ${TRAIN.iters_per_epoch}
62+
learning_rate: 1.0e-3
63+
gamma: 0.9
64+
decay_steps: 10000
65+
by_epoch: false
66+
batch_size:
67+
pde: 8192
68+
bc: 256
69+
pretrained_model_path: null
70+
checkpoint_path: null
71+
grad_norm:
72+
update_freq: 1000
73+
momentum: 0.9
74+
75+
# evaluation settings
76+
EVAL:
77+
pretrained_model_path: null
78+
eval_with_no_grad: true
79+
batch_size: 4096
80+
81+
# inference settings
82+
INFER:
83+
pretrained_model_path: https://paddle-org.bj.bcebos.com/paddlescience/models/ldc/ldc_re1000_sota_pretrained.pdparams
84+
export_path: ./inference/ldc_2d_re1000
85+
pdmodel_path: ${INFER.export_path}.pdmodel
86+
pdiparams_path: ${INFER.export_path}.pdiparams
87+
onnx_path: ${INFER.export_path}.onnx
88+
device: gpu
89+
engine: native
90+
precision: fp32
91+
ir_optim: true
92+
min_subgraph_size: 5
93+
gpu_mem: 2000
94+
gpu_id: 0
95+
max_batch_size: 1024
96+
num_cpu_threads: 10
97+
batch_size: 1024

0 commit comments

Comments
 (0)