yushuang88 commited on
Commit
b144968
·
verified ·
1 Parent(s): 3cf50bd

Upload folder using huggingface_hub

Browse files
This view is limited to 50 files because it contains too many changes.   See raw diff
Files changed (50) hide show
  1. .gitattributes +51 -0
  2. README.md +44 -0
  3. config.json +11 -0
  4. config/config.yaml +60 -0
  5. model/DSENO.py +79 -0
  6. model/__init__.py +4 -0
  7. requirements.txt +4 -0
  8. results/coordinates.npy +3 -0
  9. results/figures/pipe_1000_grid.pdf +3 -0
  10. results/figures/pipe_1000_grid.png +3 -0
  11. results/figures/pipe_1000_physical.pdf +3 -0
  12. results/figures/pipe_1000_physical.png +3 -0
  13. results/figures/pipe_1050_grid.pdf +3 -0
  14. results/figures/pipe_1050_grid.png +3 -0
  15. results/figures/pipe_1050_physical.pdf +3 -0
  16. results/figures/pipe_1050_physical.png +3 -0
  17. results/figures/pipe_1100_grid.pdf +3 -0
  18. results/figures/pipe_1100_grid.png +3 -0
  19. results/figures/pipe_1100_physical.pdf +3 -0
  20. results/figures/pipe_1100_physical.png +3 -0
  21. results/figures/pipe_1199_grid.pdf +3 -0
  22. results/figures/pipe_1199_grid.png +3 -0
  23. results/figures/pipe_1199_physical.pdf +3 -0
  24. results/figures/pipe_1199_physical.png +3 -0
  25. results/figures/test_error_distribution.png +0 -0
  26. results/figures/training_curve.pdf +0 -0
  27. results/figures/training_curve.png +3 -0
  28. results/metrics.json +20 -0
  29. results/per_sample_metrics.csv +201 -0
  30. results/predictions.npy +3 -0
  31. results/reproduction_report.md +28 -0
  32. results/run_metadata.json +128 -0
  33. results/sample_indices.npy +3 -0
  34. results/seed_0/coordinates.npy +3 -0
  35. results/seed_0/figures/pipe_1000_grid.pdf +3 -0
  36. results/seed_0/figures/pipe_1000_grid.png +3 -0
  37. results/seed_0/figures/pipe_1000_physical.pdf +3 -0
  38. results/seed_0/figures/pipe_1000_physical.png +3 -0
  39. results/seed_0/figures/pipe_1050_grid.pdf +3 -0
  40. results/seed_0/figures/pipe_1050_grid.png +3 -0
  41. results/seed_0/figures/pipe_1050_physical.pdf +3 -0
  42. results/seed_0/figures/pipe_1050_physical.png +3 -0
  43. results/seed_0/figures/pipe_1100_grid.pdf +3 -0
  44. results/seed_0/figures/pipe_1100_grid.png +3 -0
  45. results/seed_0/figures/pipe_1100_physical.pdf +3 -0
  46. results/seed_0/figures/pipe_1100_physical.png +3 -0
  47. results/seed_0/figures/pipe_1199_grid.pdf +3 -0
  48. results/seed_0/figures/pipe_1199_grid.png +3 -0
  49. results/seed_0/figures/pipe_1199_physical.pdf +3 -0
  50. results/seed_0/figures/pipe_1199_physical.png +3 -0
.gitattributes CHANGED
@@ -33,3 +33,54 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
33
  *.zip filter=lfs diff=lfs merge=lfs -text
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
33
  *.zip filter=lfs diff=lfs merge=lfs -text
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
36
+ results/figures/pipe_1000_grid.pdf filter=lfs diff=lfs merge=lfs -text
37
+ results/figures/pipe_1000_grid.png filter=lfs diff=lfs merge=lfs -text
38
+ results/figures/pipe_1000_physical.pdf filter=lfs diff=lfs merge=lfs -text
39
+ results/figures/pipe_1000_physical.png filter=lfs diff=lfs merge=lfs -text
40
+ results/figures/pipe_1050_grid.pdf filter=lfs diff=lfs merge=lfs -text
41
+ results/figures/pipe_1050_grid.png filter=lfs diff=lfs merge=lfs -text
42
+ results/figures/pipe_1050_physical.pdf filter=lfs diff=lfs merge=lfs -text
43
+ results/figures/pipe_1050_physical.png filter=lfs diff=lfs merge=lfs -text
44
+ results/figures/pipe_1100_grid.pdf filter=lfs diff=lfs merge=lfs -text
45
+ results/figures/pipe_1100_grid.png filter=lfs diff=lfs merge=lfs -text
46
+ results/figures/pipe_1100_physical.pdf filter=lfs diff=lfs merge=lfs -text
47
+ results/figures/pipe_1100_physical.png filter=lfs diff=lfs merge=lfs -text
48
+ results/figures/pipe_1199_grid.pdf filter=lfs diff=lfs merge=lfs -text
49
+ results/figures/pipe_1199_grid.png filter=lfs diff=lfs merge=lfs -text
50
+ results/figures/pipe_1199_physical.pdf filter=lfs diff=lfs merge=lfs -text
51
+ results/figures/pipe_1199_physical.png filter=lfs diff=lfs merge=lfs -text
52
+ results/figures/training_curve.png filter=lfs diff=lfs merge=lfs -text
53
+ results/seed_0/figures/pipe_1000_grid.pdf filter=lfs diff=lfs merge=lfs -text
54
+ results/seed_0/figures/pipe_1000_grid.png filter=lfs diff=lfs merge=lfs -text
55
+ results/seed_0/figures/pipe_1000_physical.pdf filter=lfs diff=lfs merge=lfs -text
56
+ results/seed_0/figures/pipe_1000_physical.png filter=lfs diff=lfs merge=lfs -text
57
+ results/seed_0/figures/pipe_1050_grid.pdf filter=lfs diff=lfs merge=lfs -text
58
+ results/seed_0/figures/pipe_1050_grid.png filter=lfs diff=lfs merge=lfs -text
59
+ results/seed_0/figures/pipe_1050_physical.pdf filter=lfs diff=lfs merge=lfs -text
60
+ results/seed_0/figures/pipe_1050_physical.png filter=lfs diff=lfs merge=lfs -text
61
+ results/seed_0/figures/pipe_1100_grid.pdf filter=lfs diff=lfs merge=lfs -text
62
+ results/seed_0/figures/pipe_1100_grid.png filter=lfs diff=lfs merge=lfs -text
63
+ results/seed_0/figures/pipe_1100_physical.pdf filter=lfs diff=lfs merge=lfs -text
64
+ results/seed_0/figures/pipe_1100_physical.png filter=lfs diff=lfs merge=lfs -text
65
+ results/seed_0/figures/pipe_1199_grid.pdf filter=lfs diff=lfs merge=lfs -text
66
+ results/seed_0/figures/pipe_1199_grid.png filter=lfs diff=lfs merge=lfs -text
67
+ results/seed_0/figures/pipe_1199_physical.pdf filter=lfs diff=lfs merge=lfs -text
68
+ results/seed_0/figures/pipe_1199_physical.png filter=lfs diff=lfs merge=lfs -text
69
+ results/seed_0/figures/training_curve.png filter=lfs diff=lfs merge=lfs -text
70
+ results/seed_17/figures/pipe_1000_grid.pdf filter=lfs diff=lfs merge=lfs -text
71
+ results/seed_17/figures/pipe_1000_grid.png filter=lfs diff=lfs merge=lfs -text
72
+ results/seed_17/figures/pipe_1000_physical.pdf filter=lfs diff=lfs merge=lfs -text
73
+ results/seed_17/figures/pipe_1000_physical.png filter=lfs diff=lfs merge=lfs -text
74
+ results/seed_17/figures/pipe_1050_grid.pdf filter=lfs diff=lfs merge=lfs -text
75
+ results/seed_17/figures/pipe_1050_grid.png filter=lfs diff=lfs merge=lfs -text
76
+ results/seed_17/figures/pipe_1050_physical.pdf filter=lfs diff=lfs merge=lfs -text
77
+ results/seed_17/figures/pipe_1050_physical.png filter=lfs diff=lfs merge=lfs -text
78
+ results/seed_17/figures/pipe_1100_grid.pdf filter=lfs diff=lfs merge=lfs -text
79
+ results/seed_17/figures/pipe_1100_grid.png filter=lfs diff=lfs merge=lfs -text
80
+ results/seed_17/figures/pipe_1100_physical.pdf filter=lfs diff=lfs merge=lfs -text
81
+ results/seed_17/figures/pipe_1100_physical.png filter=lfs diff=lfs merge=lfs -text
82
+ results/seed_17/figures/pipe_1199_grid.pdf filter=lfs diff=lfs merge=lfs -text
83
+ results/seed_17/figures/pipe_1199_grid.png filter=lfs diff=lfs merge=lfs -text
84
+ results/seed_17/figures/pipe_1199_physical.pdf filter=lfs diff=lfs merge=lfs -text
85
+ results/seed_17/figures/pipe_1199_physical.png filter=lfs diff=lfs merge=lfs -text
86
+ results/seed_17/figures/training_curve.png filter=lfs diff=lfs merge=lfs -text
README.md ADDED
@@ -0,0 +1,44 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: other
3
+ library_name: pytorch
4
+ tags:
5
+ - OneScience
6
+ - neural-operator
7
+ - pipe-flow
8
+ ---
9
+
10
+ # D-SENO: Pipe-G flow
11
+
12
+ This package reproduces the Pipe-G experiment from [Efficient Dilated Squeeze and Excitation Neural Operator for Differential Equations](https://arxiv.org/pdf/2601.17407). A seven-block, width-96 D-SENO maps curved-pipe X/Y coordinates to horizontal velocity on a 129x129 grid. Three seeds each completed 500 epochs with 1,000 training and 200 test cases.
13
+
14
+ ## Results
15
+
16
+ The supplied seed-11 checkpoint gives test relative L2 **0.00292652** after reload. The best-checkpoint three-seed mean is **0.00289914**, and the final-epoch three-seed mean is **0.00307909**. The paper reports **0.0030**. Best checkpoints were selected using the same test set, so those best scores are not independent final-test results; see `results/three_seed_summary.json` for the complete record.
17
+
18
+ ## Download
19
+
20
+ ```bash
21
+ hf download OneScience-Group/D-SENO --local-dir ./D-SENO
22
+ cd D-SENO
23
+ ```
24
+
25
+ The original Geo-FNO Pipe arrays are available from the [OneScience/HO-FNO-Pipe ModelScope dataset](https://www.modelscope.cn/datasets/OneScience/HO-FNO-Pipe). This Hugging Face repository contains no raw dataset.
26
+
27
+ ```bash
28
+ modelscope download --dataset OneScience/HO-FNO-Pipe --local_dir ./data
29
+ ```
30
+
31
+ `Pipe_X.npy` and `Pipe_Y.npy` are coordinate arrays, while `Pipe_Q.npy` contains flow fields. Each has 2,310 cases on a 129x129 grid; the model uses Q channel 0. Run all commands from the model directory; the published config reads `data/` and writes to `results/`.
32
+
33
+ ## Environment and usage
34
+
35
+ Use Python 3.10 and PyTorch on an allocated GPU/DCU, plus `onescience[cfd-gpu]` or `onescience[cfd-dcu]` and `requirements.txt`. On an HPC system, use a compute-node allocation.
36
+
37
+ ```bash
38
+ python scripts/inference.py
39
+ python scripts/result.py
40
+ ```
41
+
42
+ To retrain, move the included `weight/` and `results/` outside the package and run `python scripts/train.py --seed 11`. Other seeds need distinct `--output-root` directories. Published inference accepts a relocated data directory while still checking model and preprocessing settings.
43
+
44
+ Checkpoint: `weight/best_model.pth` (SHA256 `f6464c8bd42694228df105abd651dfbe075fb92ba1464ad4835cc889c35bad95`). Source data: [Geo-FNO](https://github.com/neuraloperator/Geo-FNO). Check original data terms before redistribution.
config.json ADDED
@@ -0,0 +1,11 @@
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model_type": "d_seno_pipe_g",
3
+ "architectures": ["DSENO"],
4
+ "framework": "pytorch",
5
+ "experiment": "Pipe-G",
6
+ "input_channels": 2,
7
+ "output_channels": 1,
8
+ "resolution": [129, 129],
9
+ "checkpoint": "weight/best_model.pth",
10
+ "dataset": "OneScience/HO-FNO-Pipe"
11
+ }
config/config.yaml ADDED
@@ -0,0 +1,60 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # Formal Pipe-G experiment. Paper omissions are recorded in provenance below.
2
+ data:
3
+ directory: data
4
+ ntrain: 1000
5
+ ntest: 200
6
+ target_channel: 0
7
+ mesh: [129, 129]
8
+ split: original_order_train_0_1000_test_1000_1200
9
+ normalization: train_only_pointwise_mean_std
10
+ normalization_eps: 1.0e-5
11
+ normalization_unbiased: true
12
+ model:
13
+ width: 96
14
+ dilations: [23, 17, 13, 9, 7, 3, 1]
15
+ projection_width: 128
16
+ input_channels: 2
17
+ output_channels: 1
18
+ se_reduction: 1
19
+ training:
20
+ epochs: 500
21
+ batch_size: 20
22
+ lr: 0.001
23
+ weight_decay: 0.0001
24
+ optimizer: Adam
25
+ betas: [0.9, 0.999]
26
+ optimizer_eps: 1.0e-8
27
+ scheduler: StepLR
28
+ step_size: 100
29
+ gamma: 0.5
30
+ seed: 11
31
+ seeds: [11, 0, 17]
32
+ precision: float32
33
+ loss: decoded_physical_relative_l2
34
+ backward_reduction: sum
35
+ logging_reduction: mean
36
+ checkpoint_metric: test_relative_l2
37
+ runtime:
38
+ device: cuda
39
+ num_workers: 0
40
+ threads: 8
41
+ require_slurm: false
42
+ memory_format: channels_last
43
+ deterministic_convolution: false
44
+ output:
45
+ root: .
46
+ provenance:
47
+ paper: https://arxiv.org/pdf/2601.17407
48
+ paper_model: Pipe-G
49
+ paper_reference: sections_3_4.2_5.2_tables_3_4_7_8
50
+ paper_test_relative_l2: 0.0030
51
+ expected_parameters: 1305761
52
+ official_reference_mode: audited_read_only_facts_for_paper_omissions_only
53
+ official_reference: https://github.com/pj1911/Efficient-Dilated-Squeeze-and-Excitation-Neural-Operator-for-Differential-Equations/blob/main/scripts/d_seno_pipe.py
54
+ official_filled_omissions: [normalization, normalization_epsilon, unbiased_std, Adam, batch_sum_loss, intermediate_DC_GELU, zero_padding, projection_width_128, seeds]
55
+ resolved_conflict: paper_XY_two_inputs_no_extra_positional_grid
56
+ initialization_assumption: PyTorch_Conv2d_default
57
+ selection_assumption: test_set_selects_best_checkpoint_no_separate_validation
58
+ units: unspecified_preserve_original_values
59
+ paper_hardware: NVIDIA_A100_80GB_PCIe
60
+ paper_epoch_time_conflict: Table2_3.30s_vs_Table7_4.28s
model/DSENO.py ADDED
@@ -0,0 +1,79 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ """Pipe-G architecture from arXiv:2601.17407, independently implemented.
2
+
3
+ The paper defines two physical coordinate inputs, width 96, seven residual
4
+ DS blocks and the dilation order. Paper omissions (DC intermediate GELU,
5
+ zero padding and projection width 128) follow the separately audited
6
+ read-only official reference; no repository code is copied here.
7
+ """
8
+ from __future__ import annotations
9
+
10
+ from collections.abc import Sequence
11
+
12
+ import torch
13
+ from torch import nn
14
+
15
+
16
+ class SqueezeExcitation(nn.Module):
17
+ """Spatial mean and two pointwise convolutions produce channel gates."""
18
+
19
+ def __init__(self, channels: int, reduction: int = 1):
20
+ super().__init__()
21
+ if reduction < 1 or channels % reduction:
22
+ raise ValueError("SE reduction must divide the channel width")
23
+ hidden = channels // reduction
24
+ self.reduce = nn.Conv2d(channels, hidden, 1, bias=True)
25
+ self.activate = nn.GELU()
26
+ self.expand = nn.Conv2d(hidden, channels, 1, bias=True)
27
+
28
+ def forward(self, features: torch.Tensor) -> torch.Tensor:
29
+ pooled = features.mean(dim=(-2, -1), keepdim=True)
30
+ gates = torch.sigmoid(self.expand(self.activate(self.reduce(pooled))))
31
+ return features * gates
32
+
33
+
34
+ class DSBlock(nn.Module):
35
+ """Two dilated convolutions, SE gating, residual addition and GELU."""
36
+
37
+ def __init__(self, channels: int, dilation: int, se_reduction: int = 1):
38
+ super().__init__()
39
+ if dilation < 1:
40
+ raise ValueError("Dilation must be positive")
41
+ self.first = nn.Conv2d(channels, channels, 3, padding=dilation,
42
+ dilation=dilation, bias=True)
43
+ self.second = nn.Conv2d(channels, channels, 3, padding=dilation,
44
+ dilation=dilation, bias=True)
45
+ self.se = SqueezeExcitation(channels, se_reduction)
46
+ self.activate = nn.GELU()
47
+
48
+ def forward(self, features: torch.Tensor) -> torch.Tensor:
49
+ transformed = self.second(self.activate(self.first(features)))
50
+ return self.activate(features + self.se(transformed))
51
+
52
+
53
+ class DSENO(nn.Module):
54
+ """Normalized coordinates [B,H,W,2] to normalized velocity [B,H,W]."""
55
+
56
+ def __init__(self, width: int = 96,
57
+ dilations: Sequence[int] = (23, 17, 13, 9, 7, 3, 1),
58
+ projection_width: int = 128, input_channels: int = 2,
59
+ output_channels: int = 1, se_reduction: int = 1):
60
+ super().__init__()
61
+ if input_channels != 2 or output_channels != 1:
62
+ raise ValueError("Pipe requires exactly X/Y inputs and one velocity output")
63
+ if width < 1 or projection_width < 1 or not dilations:
64
+ raise ValueError("Widths and dilation sequence must be nonempty and positive")
65
+ self.input_channels = input_channels
66
+ self.lifting = nn.Conv2d(input_channels, width, 1, bias=True)
67
+ self.blocks = nn.Sequential(*(DSBlock(width, int(d), se_reduction)
68
+ for d in dilations))
69
+ self.projection = nn.Sequential(nn.Conv2d(width, projection_width, 1, bias=True),
70
+ nn.GELU(),
71
+ nn.Conv2d(projection_width, output_channels, 1,
72
+ bias=True))
73
+
74
+ def forward(self, coordinates: torch.Tensor) -> torch.Tensor:
75
+ if coordinates.ndim != 4 or coordinates.shape[-1] != self.input_channels:
76
+ raise ValueError("DSENO expects coordinates shaped [B,H,W,2]")
77
+ features = coordinates.permute(0, 3, 1, 2).contiguous(memory_format=getattr(self, "memory_format", torch.channels_last))
78
+ features = self.blocks(self.lifting(features))
79
+ return self.projection(features).squeeze(1)
model/__init__.py ADDED
@@ -0,0 +1,4 @@
 
 
 
 
 
1
+ """Independent D-SENO implementation."""
2
+ from .DSENO import DSENO
3
+
4
+ __all__ = ["DSENO"]
requirements.txt ADDED
@@ -0,0 +1,4 @@
 
 
 
 
 
1
+ torch>=2.0
2
+ numpy>=1.24
3
+ PyYAML>=6.0
4
+ matplotlib>=3.7
results/coordinates.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4d727a1b4081fa15a2c56edc913f1c86e7229cf3436078e8e51e68d2a6f10b4b
3
+ size 26625728
results/figures/pipe_1000_grid.pdf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:dc337aac8d792164cfd0176b76d1b2de4d43261d49909776d04319f922f947f0
3
+ size 136817
results/figures/pipe_1000_grid.png ADDED

Git LFS Details

  • SHA256: 239a90e487ecc4153f22b13c892b9c209c0c6758db1b5f6179fcca926ed684d8
  • Pointer size: 131 Bytes
  • Size of remote file: 366 kB
results/figures/pipe_1000_physical.pdf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f9040eb393f95158696a669f6784d253bbfeb6236171a4bc63d6f3b64e1dc902
3
+ size 437107
results/figures/pipe_1000_physical.png ADDED

Git LFS Details

  • SHA256: cf39752d9f767811e167bf559e1e34d755d0b4d5bcb5e9a581d044678a92b92b
  • Pointer size: 131 Bytes
  • Size of remote file: 748 kB
results/figures/pipe_1050_grid.pdf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3aa18e20e1a5fda06198043f6c4d5a9dddd9f44caf10adfc92f663eb0af48414
3
+ size 142525
results/figures/pipe_1050_grid.png ADDED

Git LFS Details

  • SHA256: e8bb25e5bea9ef8f34caf3c59f164cafbd822bffd13fbcb583a87570a34ca4cb
  • Pointer size: 131 Bytes
  • Size of remote file: 383 kB
results/figures/pipe_1050_physical.pdf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b0031d3a159a16ba5ece54a971163ac634cc68697ec027889a5a26be3f7e7c5d
3
+ size 405605
results/figures/pipe_1050_physical.png ADDED

Git LFS Details

  • SHA256: 13a98812dca6cffd8ba45730477b7daa5ea840ad59863f8bdd2331474e0b7d42
  • Pointer size: 131 Bytes
  • Size of remote file: 689 kB
results/figures/pipe_1100_grid.pdf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8c95a827e0e3b54c20e8f2c300421965a61e3c72fade0dd29ee9a14fc34ef789
3
+ size 154160
results/figures/pipe_1100_grid.png ADDED

Git LFS Details

  • SHA256: e3e5cce8af5147ef6bf14c4998b84eb67c034b53f0ae347583f132464c007e68
  • Pointer size: 131 Bytes
  • Size of remote file: 405 kB
results/figures/pipe_1100_physical.pdf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b05c701f854aa887f8efd223ad3da9de20a850d996ed0ee25eadd4312a967695
3
+ size 470662
results/figures/pipe_1100_physical.png ADDED

Git LFS Details

  • SHA256: 7fd3dbc768edf612eb1c339555ff2c2cf8a6b90fd86eb157dd195b534a6b9bfe
  • Pointer size: 131 Bytes
  • Size of remote file: 749 kB
results/figures/pipe_1199_grid.pdf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1f3de6459ce066cf0179025a17ce67cb43f165137147ac37cce047a8c1619e92
3
+ size 121380
results/figures/pipe_1199_grid.png ADDED

Git LFS Details

  • SHA256: 77e859563b9d259c9cd583720632b8f1623688f07d689dd22e2a6fe04b325961
  • Pointer size: 131 Bytes
  • Size of remote file: 326 kB
results/figures/pipe_1199_physical.pdf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b5d50775df1e5efe131327c1bb817e97774b5ded68f546766519a13d537afa3d
3
+ size 341789
results/figures/pipe_1199_physical.png ADDED

Git LFS Details

  • SHA256: f5d20a45e625c84341808a96ecf3cd4126cd13d051825e628b34cfc9c8e96cdd
  • Pointer size: 131 Bytes
  • Size of remote file: 615 kB
results/figures/test_error_distribution.png ADDED
results/figures/training_curve.pdf ADDED
Binary file (23.4 kB). View file
 
results/figures/training_curve.png ADDED

Git LFS Details

  • SHA256: ec0451f4efb3c04063a5ddf2d87c24ea50e5737487cb850c3a41ef3d1d791237
  • Pointer size: 131 Bytes
  • Size of remote file: 168 kB
results/metrics.json ADDED
@@ -0,0 +1,20 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "seed": 11,
3
+ "n_test": 200,
4
+ "test_indices": [
5
+ 1000,
6
+ 1200
7
+ ],
8
+ "mean_relative_l2": 0.0029265209022339956,
9
+ "std_relative_l2": 0.007569868265989835,
10
+ "rmse": 0.001499550589995751,
11
+ "mae": 0.00020542695495852963,
12
+ "checkpoint_epoch": 498,
13
+ "checkpoint_sha256": "f6464c8bd42694228df105abd651dfbe075fb92ba1464ad4835cc889c35bad95",
14
+ "training_epochs_completed": 500,
15
+ "selection": "minimum test relative L2; not independent validation",
16
+ "recorded_best_test_l2": 0.002926521198824048,
17
+ "inference_seconds": 18.213220912963152,
18
+ "paper_relative_l2": 0.003,
19
+ "paper_conflict_resolution": "two physical coordinate channels; no auxiliary positional grid"
20
+ }
results/per_sample_metrics.csv ADDED
@@ -0,0 +1,201 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ sample_index,relative_l2,rmse,mae
2
+ 1000,0.0022977016971890587,0.0004245754313880101,0.0002276376779195422
3
+ 1001,0.0009355462133983215,0.00017355530559102128,0.00012269084207574044
4
+ 1002,0.0009579512074462476,0.00017667103535009727,0.00013060494806220194
5
+ 1003,0.007588256749845576,0.0013986817083775805,0.00018994845852645563
6
+ 1004,0.0009834269914827838,0.0001823138593873096,0.00013108079630119086
7
+ 1005,0.0022942312352457095,0.00042552421223652066,0.00018080477699000707
8
+ 1006,0.020979144566417723,0.003871005218703833,0.0009736213340485714
9
+ 1007,0.015538678107729577,0.0028694061889648166,0.0006332248364347396
10
+ 1008,0.0008220501528656173,0.00015153254054848104,0.00011671719137806147
11
+ 1009,0.00630868345690982,0.0011808832074697668,0.0002852170511810016
12
+ 1010,0.0009333171304368908,0.00017214202143860512,0.00012707607123308326
13
+ 1011,0.003717492819913857,0.0006872466435271469,0.0002227929635751145
14
+ 1012,0.0013822804531449233,0.0002575288645429171,0.00015094858332386312
15
+ 1013,0.0012144350165589806,0.00022501069548731925,0.00016869351097876923
16
+ 1014,0.0007362988432270753,0.00013565810126500908,9.763407749237556e-05
17
+ 1015,0.0008588700118662614,0.00015784889035429155,8.977126847496339e-05
18
+ 1016,0.004742412575458205,0.0008774364843839872,0.0001620188517866924
19
+ 1017,0.0008530768028333167,0.00015792296054885526,0.00010754887072075123
20
+ 1018,0.000846265056295879,0.00015456735010246057,0.0001115566099804369
21
+ 1019,0.0009633063161182866,0.00017861070708712217,0.00012165482814783595
22
+ 1020,0.0010321970815419224,0.00018933041135299675,0.00011386788286498813
23
+ 1021,0.000801463309102273,0.0001480309861426715,0.00010832617148365584
24
+ 1022,0.0007096999889835328,0.00013039326640384183,9.745509360180851e-05
25
+ 1023,0.0013004771416415453,0.00023911548202775756,0.00017129955815287353
26
+ 1024,0.0010761076534151978,0.00019975603720064089,0.00014587319934584634
27
+ 1025,0.00849468617940855,0.0015762108277457057,0.00022389399105764843
28
+ 1026,0.0023482201412570293,0.0004297308525828248,0.0001940773547112913
29
+ 1027,0.0007479547856859341,0.00013818211044551792,0.00010476128115973037
30
+ 1028,0.0005697949238513385,0.00010410760493527848,8.039713167473276e-05
31
+ 1029,0.0006727747418239229,0.0001232870749556826,8.612091245550801e-05
32
+ 1030,0.0014780111173713956,0.0002716086542444767,0.00012407379985850407
33
+ 1031,0.0007655144205784743,0.0001404061984199884,0.00010050341548231311
34
+ 1032,0.0012919967608733816,0.00024019003458012632,0.0001774827298640307
35
+ 1033,0.05638630275076043,0.010480767189407866,0.002634466938680346
36
+ 1034,0.006805181457526845,0.0012685690351054413,0.0005110909174390179
37
+ 1035,0.003301845600391504,0.0006066696057609299,0.00015178226096948148
38
+ 1036,0.04546523301388522,0.00845919978292304,0.002082080084477659
39
+ 1037,0.001119869822462183,0.00020833646356562707,0.00014141181980599504
40
+ 1038,0.001849730221498252,0.000345061514487934,0.0001698192455512112
41
+ 1039,0.0011285203253390754,0.0002093506848347038,0.0001588046950919262
42
+ 1040,0.0013543753462992593,0.0002507209535883456,0.0001204671867313044
43
+ 1041,0.0006031487434448164,0.00011008934220886636,8.564335341675656e-05
44
+ 1042,0.017152561712124263,0.003151748386340724,0.0008383699570487989
45
+ 1043,0.0016018176200358863,0.00029510738438431395,0.00013533552420985274
46
+ 1044,0.0007572224057361603,0.00013900851242574057,0.00010562155428299064
47
+ 1045,0.00086367197693138,0.00015920729229644396,0.00011885613693708692
48
+ 1046,0.0010130275382823667,0.0001859238225812783,0.00012023190561324224
49
+ 1047,0.0010324460571464057,0.0001892847910254007,0.00013888009698623873
50
+ 1048,0.0013205765745390584,0.00024360825821875763,0.0001383758235447551
51
+ 1049,0.0017712798700589762,0.0003279237824762317,0.00022497556868890674
52
+ 1050,0.0009394588094960563,0.0001732251449296919,0.00012366504813720455
53
+ 1051,0.0009657987545307382,0.00017824917729044505,0.00012317288276320762
54
+ 1052,0.0006159110915112526,0.00011272447622890263,7.816328442488843e-05
55
+ 1053,0.001034653194565893,0.000189006728185694,0.000118754219418392
56
+ 1054,0.0006781103576569754,0.00012454936806775518,9.38438913116273e-05
57
+ 1055,0.0006776929606386686,0.00012401087250032592,8.781409999274772e-05
58
+ 1056,0.0011400751608975507,0.00021164217937611997,0.00016344733584644386
59
+ 1057,0.0009939909708218076,0.00018422021711680055,0.00013894827854024454
60
+ 1058,0.004488609463591939,0.0008286230805626377,0.00014921853196101157
61
+ 1059,0.0013168868003477364,0.000244300663069608,0.00013305210940201486
62
+ 1060,0.0016561690213986735,0.0003030785611783716,0.00016903283187998593
63
+ 1061,0.001147994357722968,0.0002133376202795569,0.0001496051999191273
64
+ 1062,0.0008517186862832542,0.00015709443400084195,0.00011326235802214076
65
+ 1063,0.0006760948141603517,0.00012439256950708076,9.389692235306607e-05
66
+ 1064,0.0008452223603963892,0.0001561550838956058,0.0001176069513203391
67
+ 1065,0.036484209725246596,0.006856290132121632,0.0016165546056406794
68
+ 1066,0.0011395849662459224,0.00021127461950608118,0.00015668197287476278
69
+ 1067,0.000558902786830984,0.00010283690839968707,7.80038658473539e-05
70
+ 1068,0.0010138837382998446,0.00018714669644808026,0.00012516630299730825
71
+ 1069,0.0007995749980783808,0.00014728080535644157,0.00011064583163534145
72
+ 1070,0.0008490727370857537,0.00015655225860224385,0.00011979705568366495
73
+ 1071,0.0010174222989029777,0.00018815305408345553,0.00012927022262309863
74
+ 1072,0.06044199894012382,0.011097546538434758,0.0027078719514793165
75
+ 1073,0.0021639177428917113,0.00040015323139415377,0.0001831771996208958
76
+ 1074,0.0005774802473599186,0.00010587244703631121,8.298565473376917e-05
77
+ 1075,0.0006181020342738633,0.00011342395926161676,8.409809169343256e-05
78
+ 1076,0.000532389694570264,9.717937388541613e-05,7.417479729359971e-05
79
+ 1077,0.002197728807997973,0.00040629609741957463,0.0002333838137122211
80
+ 1078,0.0007417403124897907,0.0001359657728373943,9.718706091001445e-05
81
+ 1079,0.0011552623676149725,0.00021467507017855822,0.00016342583789174386
82
+ 1080,0.0007695169843042761,0.00014154506809490344,9.402958032978281e-05
83
+ 1081,0.0012986821272850941,0.000241064102254564,0.00017644433948642486
84
+ 1082,0.0013342217762410198,0.00024543294330772226,0.00017979272302455601
85
+ 1083,0.0009131510282826279,0.0001686387986686369,0.00011613111940475288
86
+ 1084,0.0007821005494168509,0.00014400874501019158,0.00010638969291361162
87
+ 1085,0.0007680070766600492,0.00014107960200776182,0.00010347834126168737
88
+ 1086,0.0023585301256076347,0.0004378944738996479,0.0002019308667385528
89
+ 1087,0.0006712772431522356,0.0001231346806861526,9.273606584200095e-05
90
+ 1088,0.0012165263598888068,0.00022528978806594913,0.00016035029980949447
91
+ 1089,0.0008453393156466277,0.00015462316263034078,9.723704759437082e-05
92
+ 1090,0.0012311797433055176,0.00022858209322826226,0.00016225749690269292
93
+ 1091,0.0022236822292763334,0.0004101494129611896,0.00019562022405528819
94
+ 1092,0.0006565415006102689,0.00012033902084263929,9.236098857300036e-05
95
+ 1093,0.002407775709750736,0.0004515170568402421,0.0002771225213333326
96
+ 1094,0.0032559343386269917,0.0006059755953024647,0.00028591900705730877
97
+ 1095,0.0009068555523692832,0.00016702214865458,0.0001276859481384651
98
+ 1096,0.0042962572535882315,0.0008012155225704341,0.00024266058614205743
99
+ 1097,0.0009897686783963027,0.00018337280897205518,0.00012255817529437327
100
+ 1098,0.0006251331796918896,0.00011481888723935643,8.634965445480484e-05
101
+ 1099,0.0008525459392920498,0.0001572505628320289,0.00011997162070797442
102
+ 1100,0.0016340001025245189,0.0003040115645065819,0.00021864103428116137
103
+ 1101,0.002991226089770646,0.0005558742857152084,0.00017526372195301497
104
+ 1102,0.001092191147823935,0.0002018002707007483,0.00014979999174539218
105
+ 1103,0.0008355553606041598,0.00015360119613062935,0.00011530872780713066
106
+ 1104,0.0009190318833032241,0.00016913296539373295,0.00011573782979037699
107
+ 1105,0.02959502204165551,0.005434646635295581,0.0010253679255135186
108
+ 1106,0.007946408760413392,0.0014661123114346841,0.0004096465091477316
109
+ 1107,0.00084552894903526,0.00015608104546367604,0.00011602352577642531
110
+ 1108,0.0008015539692081439,0.00014697409667690465,0.00010814439970185653
111
+ 1109,0.00076303964761529,0.000140554911172631,0.00010692800418013271
112
+ 1110,0.0009523566703482207,0.00017626597639217975,0.00013391424730188392
113
+ 1111,0.0009601475818734708,0.00017729475474701235,0.00013293182019058893
114
+ 1112,0.0012484732361080682,0.00023219158781780512,0.00017602928772836323
115
+ 1113,0.0010973965496280254,0.0002036292321147137,0.00012427589689359723
116
+ 1114,0.0012845879353802062,0.00023762355051105074,0.00015231071865883108
117
+ 1115,0.0022097661609067577,0.00040763908003514155,0.00019361480279461474
118
+ 1116,0.0008349466331337646,0.00015265966575286492,9.711535809468999e-05
119
+ 1117,0.001725632501883809,0.00032076153964824024,0.0002155065474182872
120
+ 1118,0.0005799995096566365,0.00010659035998356608,8.232806320133226e-05
121
+ 1119,0.0008161137269070354,0.00015120651764964762,0.00011524458195375179
122
+ 1120,0.0011937227596327802,0.00022169525431854837,0.0001655357999501664
123
+ 1121,0.000875016132714357,0.00016054725179663127,0.00012173570746404323
124
+ 1122,0.0009531562998083918,0.00017596961202332042,0.00012664539567658892
125
+ 1123,0.004631123986488434,0.000867197808880576,0.00024275791485868286
126
+ 1124,0.0018277563735145995,0.00033950523468276333,0.00023545528638064243
127
+ 1125,0.0006487314863840367,0.00011864017071388083,8.913001616607496e-05
128
+ 1126,0.0007442499103442979,0.00013716705727062403,0.00010432873897365773
129
+ 1127,0.0007339917057630392,0.00013472045013150166,0.00010339105032455937
130
+ 1128,0.000946501772354011,0.00017478808776483175,0.0001325568643284434
131
+ 1129,0.00111621896617694,0.0002078006513956315,0.00012849223893468176
132
+ 1130,0.007487256515832383,0.0013798649234393473,0.00019469864097408998
133
+ 1131,0.0028589492342166677,0.000528684977848869,0.0002825639644897904
134
+ 1132,0.0008114404478121972,0.00014932778709032823,0.00010542534769085327
135
+ 1133,0.000971069411042974,0.00018063642111678372,0.00012204860068615412
136
+ 1134,0.0010193840544397202,0.00018814303311417776,0.00013187477210592385
137
+ 1135,0.0007030661115065838,0.00012870077267819366,9.309878086117515e-05
138
+ 1136,0.0007222968428765814,0.00013275895369207116,9.632787140294805e-05
139
+ 1137,0.0030347028587965853,0.0005618232252352535,0.0002549555995820534
140
+ 1138,0.0013211805240559203,0.0002445087531503473,0.00016753417795406685
141
+ 1139,0.0008566967045259716,0.00015714381478484403,8.942407914922234e-05
142
+ 1140,0.0008290580288422926,0.00015384184310281206,0.0001150290162578837
143
+ 1141,0.0010851264511966146,0.00019919466121172248,0.0001442072103022809
144
+ 1142,0.0008817647535937265,0.0001624980028408971,0.00012000579474122921
145
+ 1143,0.0009978307484903148,0.00018408708408734919,0.00013639106389291838
146
+ 1144,0.000989182451363875,0.00018339510540413094,0.0001340986705231566
147
+ 1145,0.0008339264360171897,0.0001528476363142045,0.00011776072878627639
148
+ 1146,0.0019464943419983992,0.000363696329986131,0.00018189524335649474
149
+ 1147,0.0010523259996906024,0.00019634557754849813,0.00013673786163585512
150
+ 1148,0.0006298391959858486,0.00011488609605924065,8.804865432639884e-05
151
+ 1149,0.0007400749910818521,0.00013615398937248045,0.00010085979460868274
152
+ 1150,0.0005691933839183395,0.00010409713240422187,7.72865461093694e-05
153
+ 1151,0.0009238890452542942,0.0001714380155009082,0.0001259274074983328
154
+ 1152,0.0017514444551631743,0.0003247496393141526,0.00018890120059879446
155
+ 1153,0.001008329301655796,0.00018519998616774904,0.00012289615188599835
156
+ 1154,0.0008387190824464518,0.00015447934991504327,0.00010892729044702579
157
+ 1155,0.0007161161941318931,0.00013134827696631475,9.167503053759951e-05
158
+ 1156,0.0011085028414543874,0.000204901396346653,0.00015487695233642294
159
+ 1157,0.000701122239925638,0.00012872052475129016,9.849797066508096e-05
160
+ 1158,0.0012725565634488728,0.00023715088364013843,0.00017512835913176668
161
+ 1159,0.0008560518025343002,0.0001574813663673128,0.00011323871474110905
162
+ 1160,0.0010551986913992935,0.00019524229677503593,0.0001418537190988494
163
+ 1161,0.0005799995789190519,0.00010595064298882985,7.525239303914193e-05
164
+ 1162,0.004028811557730473,0.0007443511444256793,0.0003040327663930429
165
+ 1163,0.0006157767775139202,0.00011301313196690456,8.430169792319143e-05
166
+ 1164,0.009976381985231033,0.0018433299789428965,0.0005804000318968206
167
+ 1165,0.0008021983072542625,0.0001486867874261117,0.00010968682156438235
168
+ 1166,0.003146225484794383,0.0005782230991834915,0.00025101014403656715
169
+ 1167,0.00125452806109533,0.00023319204555053197,0.0001749577121179709
170
+ 1168,0.0008863241396464565,0.00016200458313205847,0.00010509499263067124
171
+ 1169,0.0010202505958216267,0.00018878855756502972,0.00014057289840147948
172
+ 1170,0.0011441372878127334,0.00021237559838892006,0.0001518607555753167
173
+ 1171,0.001278653464241886,0.0002351225698374442,0.00012123394271727299
174
+ 1172,0.0006403365405048999,0.00011757913199902482,8.921624433487477e-05
175
+ 1173,0.001223068857194006,0.00022747490282254806,0.0001619286236703965
176
+ 1174,0.0010386081945899074,0.0001925482230834899,0.00014525292102138978
177
+ 1175,0.0007893340854312947,0.00014651239691335167,0.00011416500630059159
178
+ 1176,0.0008775398215861142,0.0001619004580871263,0.00012107880486997805
179
+ 1177,0.0009272360121559145,0.00017002176652530643,0.00010877193237529514
180
+ 1178,0.0005163100931139649,9.45874131043955e-05,7.258090553886136e-05
181
+ 1179,0.0006312902488991588,0.00011620843437668474,8.523230875004013e-05
182
+ 1180,0.0008785915458249157,0.0001629518283321546,0.00012515406039145208
183
+ 1181,0.0016418225379868626,0.00030238017606301013,0.00015488589210587523
184
+ 1182,0.007984512813961547,0.0014738368601886253,0.0005360701496602789
185
+ 1183,0.0008857516081738958,0.00016327829693785805,0.00012177002068175525
186
+ 1184,0.0014423659335239803,0.000266394107051537,0.0001786742306400993
187
+ 1185,0.0006280949881908072,0.00011515813226136313,8.66733615862304e-05
188
+ 1186,0.000549226688842606,0.00010054009005113506,7.446237085036372e-05
189
+ 1187,0.002152342428092499,0.0003985778044943974,0.00019703817374971114
190
+ 1188,0.005401676993235297,0.0010061923145903368,0.00022626211390530765
191
+ 1189,0.0007296747442614421,0.00013447813570395013,0.0001042460774536284
192
+ 1190,0.0013055229990481026,0.0002409155279513708,0.0001699858828880192
193
+ 1191,0.0011197362740212986,0.00020679827664853097,0.0001602481731927612
194
+ 1192,0.0007923231080816391,0.00014507497076340388,0.00011186703636331646
195
+ 1193,0.003600441995798049,0.0006674382941777328,0.0001820855061522862
196
+ 1194,0.0009122843467053009,0.0001674279719142483,0.00012071699589197319
197
+ 1195,0.0006560739247864402,0.00011994505941693953,9.057031968433275e-05
198
+ 1196,0.0011128618402373653,0.0002054458211992111,0.00014611055840365683
199
+ 1197,0.010400202205231539,0.001939437256982087,0.0006364977968048975
200
+ 1198,0.0006534973071351172,0.00011989702022768055,9.157126160500665e-05
201
+ 1199,0.0012121079146577287,0.00022336840027976866,0.00013498923415559733
results/predictions.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:29aef8bcedaadbce81ca19d12a894019eba69ead274d5495cea015ab5bab6ff2
3
+ size 13312928
results/reproduction_report.md ADDED
@@ -0,0 +1,28 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # D-SENO Pipe reproduction
2
+
3
+ Completed **500 epochs**, seed 11. Input: physical X/Y coordinates, target: Q channel 0. Fixed split: [0,1000) train, [1000,1200) test, full 129 x 129 resolution.
4
+
5
+ Best-checkpoint mean relative L2: **0.00292652** at epoch **498**. Final-epoch test relative L2: 0.00303409. RMSE: 0.00149955; MAE: 0.00020543. Paper reference: 0.0030 averaged over three seeds, not a guaranteed target for one run.
6
+
7
+ Training uses width96 and 7 DS blocks with dilation [23,17,13,9,7,3,1], relative L2, Adam lr0.001 and weight decay0.0001, StepLR step100 gamma0.5, batch20, FP32 and no early stopping. Each training batch sums sample losses; reported scores average sample-relative L2 errors.
8
+
9
+ The checkpoint is selected by lowest test metric as requested. These 200 samples are therefore a selection set; this score is not an estimate on a separate unseen final holdout.
10
+
11
+ Paper precedence: Section5.2 specifies two input coordinates, and the conclusion says no positional encoding. The extra grid channels in the official script are omitted. Normalization (training-only pointwise unbiased std plus1e-5), Adam implementation details and projection96->128->1 fill missing paper details using online read-only official references. No official repository was downloaded or copied. Independent implementation has 1,305,761 parameters; the paper rounds to1.306M.
12
+
13
+ Hardware: allocated Hygon DCU compute node instead of the paper's NVIDIA A10080GB. Exact device, node, environment, source hashes and SLURM job ID are in run_metadata.json. No training or inference ran on login nodes.
14
+
15
+ Only horizontal velocity is available to this surrogate; no vector streamlines, pressure, lift, drag, or divergence-free claim is derived. Coordinate and velocity units and Reynolds number are not specified in the supplied protocol and are not invented. Figures use the true curved mesh, matched truth/prediction color ranges and separate error colorbars; computational-grid views are also supplied.
16
+
17
+ Sources: [paper](https://arxiv.org/pdf/2601.17407), [official missing-detail reference](https://github.com/pj1911/Efficient-Dilated-Squeeze-and-Excitation-Neural-Operator-for-Differential-Equations/blob/main/scripts/d_seno_pipe.py).
18
+
19
+
20
+ ## Three-seed result
21
+
22
+ - Seed 11: 500 epochs; best relative L2 0.00292652 at epoch 498; final 0.00303409.
23
+ - Seed 0: 500 epochs; best relative L2 0.00293700 at epoch 491; final 0.00317688.
24
+ - Seed 17: 500 epochs; best relative L2 0.00283390 at epoch 479; final 0.00302631.
25
+
26
+ Mean test-selected best relative L2: **0.00289914 ± 0.00005674** (sample standard deviation across 3 seeds). Mean final-epoch relative L2: **0.00307909 ± 0.00008478**. Paper reference: 0.0030. Total budget: 1500 epochs.
27
+
28
+ All three SLURM training/inference/visualization jobs completed with exit0. Each best checkpoint was actually loaded and independently evaluated on200 test samples on its compute node. The final audit used only standard-library file/code checks:500-row histories and exact LR schedules, checkpoint scalar metadata without tensor loading, saved checkpoint hashes, NPY headers, sample IDs, frozen core source hashes, figure completion, and unchanged original data metadata. No model, tensor storage, or data array was loaded on the login node. The redundant queued aggregate jobs1130371 and1132308 were cancelled after all scientific jobs completed.
results/run_metadata.json ADDED
@@ -0,0 +1,128 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "seed": 11,
3
+ "config": {
4
+ "data": {
5
+ "directory": "/public/home/liuyushuang/zhaomingsi/onescience-cfd-data/HO-FNO-Pipe-Flow",
6
+ "ntrain": 1000,
7
+ "ntest": 200,
8
+ "target_channel": 0,
9
+ "mesh": [
10
+ 129,
11
+ 129
12
+ ],
13
+ "split": "original_order_train_0_1000_test_1000_1200",
14
+ "normalization": "train_only_pointwise_mean_std",
15
+ "normalization_eps": 1e-05,
16
+ "normalization_unbiased": true
17
+ },
18
+ "model": {
19
+ "width": 96,
20
+ "dilations": [
21
+ 23,
22
+ 17,
23
+ 13,
24
+ 9,
25
+ 7,
26
+ 3,
27
+ 1
28
+ ],
29
+ "projection_width": 128,
30
+ "input_channels": 2,
31
+ "output_channels": 1,
32
+ "se_reduction": 1
33
+ },
34
+ "training": {
35
+ "epochs": 500,
36
+ "batch_size": 20,
37
+ "lr": 0.001,
38
+ "weight_decay": 0.0001,
39
+ "optimizer": "Adam",
40
+ "betas": [
41
+ 0.9,
42
+ 0.999
43
+ ],
44
+ "optimizer_eps": 1e-08,
45
+ "scheduler": "StepLR",
46
+ "step_size": 100,
47
+ "gamma": 0.5,
48
+ "seed": 11,
49
+ "seeds": [
50
+ 11,
51
+ 0,
52
+ 17
53
+ ],
54
+ "precision": "float32",
55
+ "loss": "decoded_physical_relative_l2",
56
+ "backward_reduction": "sum",
57
+ "logging_reduction": "mean",
58
+ "checkpoint_metric": "test_relative_l2"
59
+ },
60
+ "runtime": {
61
+ "device": "cuda",
62
+ "num_workers": 0,
63
+ "threads": 8,
64
+ "require_slurm": true,
65
+ "memory_format": "channels_last",
66
+ "deterministic_convolution": false
67
+ },
68
+ "output": {
69
+ "root": "/public/home/liuyushuang/zhaomingsi/onescience-cfd-repro-packages/D-SENO"
70
+ },
71
+ "provenance": {
72
+ "paper": "https://arxiv.org/pdf/2601.17407",
73
+ "paper_model": "Pipe-G",
74
+ "paper_reference": "sections_3_4.2_5.2_tables_3_4_7_8",
75
+ "paper_test_relative_l2": 0.003,
76
+ "expected_parameters": 1305761,
77
+ "official_reference_mode": "audited_read_only_facts_for_paper_omissions_only",
78
+ "official_reference": "https://github.com/pj1911/Efficient-Dilated-Squeeze-and-Excitation-Neural-Operator-for-Differential-Equations/blob/main/scripts/d_seno_pipe.py",
79
+ "official_filled_omissions": [
80
+ "normalization",
81
+ "normalization_epsilon",
82
+ "unbiased_std",
83
+ "Adam",
84
+ "batch_sum_loss",
85
+ "intermediate_DC_GELU",
86
+ "zero_padding",
87
+ "projection_width_128",
88
+ "seeds"
89
+ ],
90
+ "resolved_conflict": "paper_XY_two_inputs_no_extra_positional_grid",
91
+ "initialization_assumption": "PyTorch_Conv2d_default",
92
+ "selection_assumption": "test_set_selects_best_checkpoint_no_separate_validation",
93
+ "units": "unspecified_preserve_original_values",
94
+ "paper_hardware": "NVIDIA_A100_80GB_PCIe",
95
+ "paper_epoch_time_conflict": "Table2_3.30s_vs_Table7_4.28s"
96
+ }
97
+ },
98
+ "parameters": 1305761,
99
+ "node": "f11r3n12",
100
+ "slurm_job_id": "1130384",
101
+ "device": "BW",
102
+ "torch": "2.4.1",
103
+ "hip": "6.1.25065",
104
+ "cuda": null,
105
+ "numpy": "1.26.4",
106
+ "python": "3.10.16 (main, Dec 11 2024, 16:24:50) [GCC 11.2.0]",
107
+ "source_hashes": {
108
+ "model/DSENO.py": "93e7656ae3dcf8cbcc5324d0fc029ade7724f1ca677e3982cb25cd6074cc2f33",
109
+ "model/__init__.py": "98dc01f23cf769bfaa3b91ec4227557236d10845d3c7624d58e788b77f07f171",
110
+ "scripts/aggregate.py": "3eafe439d9506a00144335f1896a62fd0d3953d2f87b96cb5bf26fb466ff7168",
111
+ "scripts/benchmark.py": "13f88aa4c188e3b6febc0824b75203feeefbe118dfb77aede1084b368a682f37",
112
+ "scripts/common.py": "8a41cdfbac0f1a876f90590c469a09f269e611ff336a5937fec948fce7d22996",
113
+ "scripts/inference.py": "f146165d0612d651379c0fb69c4f5dfa8440e11710a7fbf4d0aacc6dd06957f8",
114
+ "scripts/preflight.py": "6d2da79fd02ea1d795ec62a40f3c56a53aa58eabec222ae27b80d074897e54cc",
115
+ "scripts/result.py": "314b315275721cd656140cce2ea2723af838f4655f6c43864281f88972faeea8",
116
+ "scripts/smoke.py": "29f796134cb616a99e83afa921cf0b97c0ad5fa46c2b27dd356c22a42eb3b62c",
117
+ "scripts/train.py": "a4698d0653dc5b0fd94ff19cd0a23b3eb6c095b3be3594a2708ecd503a393fb6",
118
+ "config/config.yaml": "569c7cfb6fbef269fd6a67121c63df8b8e9b3945864c56b3023af7c555d3fabf"
119
+ },
120
+ "checkpoint_selection": "minimum mean test relative L2 (test-selected)",
121
+ "paper_test_l2": 0.003,
122
+ "paper_conflict": "No auxiliary positional grid; paper two-channel input takes priority",
123
+ "precision": "float32; AMP/TF32 disabled",
124
+ "normalization_fit_indices": [
125
+ 0,
126
+ 1000
127
+ ]
128
+ }
results/sample_indices.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7ba3d9e686aed626dd4cae2d589a5dafc2fff8719600d9c26f03f53563c1674a
3
+ size 1728
results/seed_0/coordinates.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4d727a1b4081fa15a2c56edc913f1c86e7229cf3436078e8e51e68d2a6f10b4b
3
+ size 26625728
results/seed_0/figures/pipe_1000_grid.pdf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4685b730060fd7db8593eb9e2a21dbac23295e57673ba1cd1d22570246d3dee6
3
+ size 124024
results/seed_0/figures/pipe_1000_grid.png ADDED

Git LFS Details

  • SHA256: 0b2ff37a35b7331006093576d62fbb4f86dce399a46313dcd72df54fefac5f19
  • Pointer size: 131 Bytes
  • Size of remote file: 328 kB
results/seed_0/figures/pipe_1000_physical.pdf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a4b3d836b5383451b5f26dcb605b4750142e244152367c2e0ea1cbf47bdcf2e4
3
+ size 389842
results/seed_0/figures/pipe_1000_physical.png ADDED

Git LFS Details

  • SHA256: 1bd354eb9e5bada6be3193e67325d9b39e480fe0be9da965c5425a5c344fe52c
  • Pointer size: 131 Bytes
  • Size of remote file: 691 kB
results/seed_0/figures/pipe_1050_grid.pdf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cb97488100a16ace11a564a25ca9fff4a6cab08c3463fb3a4986c9ad394f79d4
3
+ size 142088
results/seed_0/figures/pipe_1050_grid.png ADDED

Git LFS Details

  • SHA256: 45012dbb80233ad79cafa89188b3a7351b41b50a2f75f8217d6d17ee1c006518
  • Pointer size: 131 Bytes
  • Size of remote file: 383 kB
results/seed_0/figures/pipe_1050_physical.pdf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:abe85d626ec4015797c46bc68bb61ba7f211d661b71b614cc0f1aaba0acf6150
3
+ size 404769
results/seed_0/figures/pipe_1050_physical.png ADDED

Git LFS Details

  • SHA256: d742e28b6d7ad68a53ee32b318cb8d7d971b7e865625bacfc20ce22eb55fc5b6
  • Pointer size: 131 Bytes
  • Size of remote file: 689 kB
results/seed_0/figures/pipe_1100_grid.pdf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bf67fa61fe38228da072c2cd1250c72d65210013737c955ca32c564618a97ded
3
+ size 149747
results/seed_0/figures/pipe_1100_grid.png ADDED

Git LFS Details

  • SHA256: 115a2577a80dcbe9598c5416c1b60387b2f5fd71c6e4361f80efa75a890a6459
  • Pointer size: 131 Bytes
  • Size of remote file: 412 kB
results/seed_0/figures/pipe_1100_physical.pdf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6dd066b4ab12901166c37f9c4304f27cbd94877264e1f069093273da6cd4fbbb
3
+ size 445757
results/seed_0/figures/pipe_1100_physical.png ADDED

Git LFS Details

  • SHA256: 0fd69414055f1fbfabfe4726f442f74bfce5945bdf7f891559ced968f9a2bab4
  • Pointer size: 131 Bytes
  • Size of remote file: 740 kB
results/seed_0/figures/pipe_1199_grid.pdf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:83707b2da6cfdc4edba5d8c7cfb5d08e9b9944d389d0bcc9a0eaeb23f245cb93
3
+ size 123273
results/seed_0/figures/pipe_1199_grid.png ADDED

Git LFS Details

  • SHA256: bb273c6148e5a59016dc1722c231334bfd0122014a1ab5d3424d53b49af6bf25
  • Pointer size: 131 Bytes
  • Size of remote file: 328 kB
results/seed_0/figures/pipe_1199_physical.pdf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6409016b74f2e9f1c74168de70fbbe54fcef0767c239f8b607137b1f1ad2c495
3
+ size 347047
results/seed_0/figures/pipe_1199_physical.png ADDED

Git LFS Details

  • SHA256: 6bdb6371e324d7829308fea047b572791c414f5b07f0ba39bb52c03a22c52e20
  • Pointer size: 131 Bytes
  • Size of remote file: 617 kB