yushuang88 commited on
Commit
db32cff
·
verified ·
1 Parent(s): 15294c2

Upload folder using huggingface_hub

Browse files
.gitattributes CHANGED
@@ -33,3 +33,5 @@ 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/learning_curves.png filter=lfs diff=lfs merge=lfs -text
37
+ results/pressure_fields.png filter=lfs diff=lfs merge=lfs -text
README.md ADDED
@@ -0,0 +1,44 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ library_name: pytorch
4
+ tags:
5
+ - OneScience
6
+ - neural-operator
7
+ - darcy-flow
8
+ ---
9
+
10
+ # MoNo-light: Darcy 85x85
11
+
12
+ This package reproduces the MoNo-light Darcy experiment in Table 3 of the [MoNo paper](https://arxiv.org/pdf/2608.09764v1). It predicts pressure from permeability on an 85x85 grid using multiscale latent variables and Sinkhorn transport. The single-seed run completed 500 epochs with 1,000 training and 200 test samples.
13
+
14
+ ## Results
15
+
16
+ The reloaded final-epoch checkpoint achieves mean test relative L2 **0.00693725**, versus **0.0070** in the paper, a numerical difference of about 0.9%. Hardware, software, and the Sinkhorn kernel differ from the authors' setup, so this is not evidence of strict same-condition superiority. Full metrics and plots are in `results/`.
17
+
18
+ ## Download
19
+
20
+ ```bash
21
+ hf download OneScience-Group/MoNo-light --local-dir ./MoNo-light
22
+ cd MoNo-light
23
+ ```
24
+
25
+ The two original 421x421 Darcy MAT files are also hosted in the [OneScience/HNO-Darcy ModelScope dataset](https://www.modelscope.cn/datasets/OneScience/HNO-Darcy). This Hugging Face repository contains the model only, not the raw data. Download the dataset from the model directory:
26
+
27
+ ```bash
28
+ modelscope download --dataset OneScience/HNO-Darcy --local_dir ./data
29
+ ```
30
+
31
+ The package reads `data/piececonst_r421_N1024_smooth1.mat` and `data/piececonst_r421_N1024_smooth2.mat`. Both contain `coeff` and `sol` arrays; the loader uses the first 1,000 and 200 samples respectively and strides the 421x421 grid by five. It creates a local cache on first use.
32
+
33
+ ## Environment and usage
34
+
35
+ Use Python 3.11, PyTorch on an allocated GPU/DCU, `onescience[cfd-gpu]` or `onescience[cfd-dcu]`, SciPy, PyYAML, and Matplotlib. The raw MAT files total about 3.39 GB; allow sufficient RAM for the first load.
36
+
37
+ ```bash
38
+ python scripts/inference.py
39
+ python scripts/result.py
40
+ ```
41
+
42
+ For a new 500-epoch training run, move the included `weight/` and `results/` outside the model directory before running `python scripts/train.py`. The published inference accepts a relocated data directory while retaining the model/training-configuration check. The weights are unchanged from the original run.
43
+
44
+ Checkpoint: `weight/best_model.pth` (SHA256 `d20393f9ae89d38f53b8da9e09799db57bef37154ec8c7fddb9a9dd4a6666122`). Original data source: [FNO Darcy files](https://drive.google.com/drive/folders/1UnbQh2WWc6knEHbLn-ZaXrKUZhp7pjt-). Check the source's terms before redistributing the data.
config.json ADDED
@@ -0,0 +1,11 @@
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model_type": "mono_light_darcy",
3
+ "architectures": ["MoNoLight"],
4
+ "framework": "pytorch",
5
+ "experiment": "Darcy_85_Table3",
6
+ "input_channels": 1,
7
+ "output_channels": 1,
8
+ "resolution": [85, 85],
9
+ "checkpoint": "weight/best_model.pth",
10
+ "dataset": "OneScience/HNO-Darcy"
11
+ }
config/config.yaml ADDED
@@ -0,0 +1,36 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ seed: 0
2
+ data:
3
+ directory: data
4
+ train_file: piececonst_r421_N1024_smooth1.mat
5
+ test_file: piececonst_r421_N1024_smooth2.mat
6
+ train_samples: 1000
7
+ test_samples: 200
8
+ stride: 5
9
+ grid_size: 85
10
+ cache: results/darcy85_data.npz
11
+ model:
12
+ width: 96
13
+ latent_sizes: [512, 256, 128, 64]
14
+ depths: [3, 1, 1, 1]
15
+ heads: [4, 4, 8, 8]
16
+ sinkhorn_iterations: 8
17
+ temperature: 1.0
18
+ anchor_grad_fraction: 0.25
19
+ output_dropout: 0.5
20
+ training:
21
+ epochs: 500
22
+ batch_size: 4
23
+ learning_rate: 0.001
24
+ warmup_epochs: 100
25
+ weight_decay: 0.00005
26
+ betas: [0.9, 0.99]
27
+ grad_clip: 1000.0
28
+ div_factor: 100
29
+ final_div_factor: 100
30
+ checkpoint: weight/best_model.pth
31
+ resume_checkpoint: weight/last_model.pth
32
+ selection: final_epoch_500_no_test_selection
33
+ precision: float32
34
+ print_every_steps: 25
35
+ results: results
36
+ paper_relative_l2: 0.0070
model/MoNo.py ADDED
@@ -0,0 +1,124 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ """Independent implementation of MoNo-light, paper equations (1)--(6), (29)--(30).
2
+
3
+ No author repository code is included. See .trainer_work/darcy85/source_capture.md.
4
+ Identical Darcy coordinates share one geometry route across a batch. All transport
5
+ iterations remain differentiable, including both encoder and decoder projections.
6
+ """
7
+ import copy
8
+ import math
9
+
10
+ import torch
11
+ from torch import nn
12
+ from torch.nn import functional as F
13
+
14
+
15
+ class ResidualProjector(nn.Module):
16
+ def __init__(self, inputs, width, outputs, dropout=0.0):
17
+ super().__init__()
18
+ self.lift = nn.Linear(inputs, width)
19
+ self.hidden = nn.ModuleList(nn.Linear(width, width) for _ in range(3))
20
+ self.drop = nn.Dropout(dropout)
21
+ self.readout = nn.Linear(width, outputs)
22
+
23
+ def forward(self, x):
24
+ x = F.gelu(self.lift(x))
25
+ for layer in self.hidden:
26
+ x = x + self.drop(F.gelu(layer(x)))
27
+ return self.readout(x)
28
+
29
+
30
+ def sinkhorn_transport(scores, iterations=8, temperature=1.0):
31
+ """Uniform-marginal log Sinkhorn; finite iterations, never detached."""
32
+ z = scores / temperature
33
+ n, m = z.shape[-2:]
34
+ v = z.new_zeros((*z.shape[:-2], 1, m))
35
+ for _ in range(iterations):
36
+ u = -math.log(n) - torch.logsumexp(z + v, dim=-1, keepdim=True)
37
+ v = -math.log(m) - torch.logsumexp(z + u, dim=-2, keepdim=True)
38
+ return torch.exp(z + u + v)
39
+
40
+
41
+ class TransformerLayer(nn.Module):
42
+ def __init__(self, width, heads):
43
+ super().__init__()
44
+ self.heads = heads
45
+ self.norm_attention = nn.LayerNorm(width)
46
+ self.qkv = nn.Linear(width, width * 3)
47
+ self.attention_out = nn.Linear(width, width)
48
+ self.norm_ffn = nn.LayerNorm(width)
49
+ self.ffn = nn.Sequential(nn.Linear(width, 4 * width), nn.GELU(),
50
+ nn.Linear(4 * width, width))
51
+
52
+ def forward(self, x):
53
+ b, n, d = x.shape
54
+ q, k, v = self.qkv(self.norm_attention(x)).reshape(
55
+ b, n, 3, self.heads, d // self.heads).permute(2, 0, 3, 1, 4).unbind(0)
56
+ a = F.scaled_dot_product_attention(q, k, v, dropout_p=0.0)
57
+ x = x + self.attention_out(a.transpose(1, 2).reshape(b, n, d))
58
+ return x + self.ffn(self.norm_ffn(x))
59
+
60
+
61
+ def initialize_readout(module):
62
+ if isinstance(module, nn.Linear):
63
+ nn.init.trunc_normal_(module.weight, std=0.02)
64
+ if module.bias is not None:
65
+ nn.init.zeros_(module.bias)
66
+ elif isinstance(module, nn.LayerNorm):
67
+ nn.init.ones_(module.weight)
68
+ nn.init.zeros_(module.bias)
69
+
70
+
71
+ class MoNoLight(nn.Module):
72
+ def __init__(self, width=96, latent_sizes=(512, 256, 128, 64),
73
+ depths=(3, 1, 1, 1), heads=(4, 4, 8, 8),
74
+ sinkhorn_iterations=8, temperature=1.0,
75
+ anchor_grad_fraction=0.25, output_dropout=0.5):
76
+ super().__init__()
77
+ self.latent_sizes = tuple(latent_sizes)
78
+ self.iterations = sinkhorn_iterations
79
+ self.temperature = temperature
80
+ self.anchor_grad_fraction = anchor_grad_fraction
81
+ self.anchor = ResidualProjector(2, width, width)
82
+ self.frozen_anchor = copy.deepcopy(self.anchor).requires_grad_(False)
83
+ self.physical = ResidualProjector(3, width, width)
84
+ self.assignment = nn.ModuleList(
85
+ ResidualProjector(width, width, n) for n in latent_sizes)
86
+ self.encoder = nn.ModuleList(nn.Sequential(*[
87
+ TransformerLayer(width, h) for _ in range(d)]) for d, h in zip(depths, heads))
88
+ self.decoder = nn.ModuleList(nn.Sequential(*[
89
+ TransformerLayer(width, h) for _ in range(d)]) for d, h in zip(depths, heads))
90
+ self.output = ResidualProjector(width, width, 1, output_dropout)
91
+ self.encoder.apply(initialize_readout)
92
+ self.decoder.apply(initialize_readout)
93
+ self.output.apply(initialize_readout)
94
+
95
+ def embed_geometry(self, coordinates):
96
+ if not self.training:
97
+ return self.frozen_anchor(coordinates)
98
+ # Synchronize before the train forward, as documented in the existing spec.
99
+ with torch.no_grad():
100
+ for dst, src in zip(self.frozen_anchor.parameters(), self.anchor.parameters()):
101
+ dst.copy_(src)
102
+ n = coordinates.shape[-2]
103
+ chosen = torch.randperm(n, device=coordinates.device)[:math.ceil(n * self.anchor_grad_fraction)]
104
+ # Evaluate the frozen copy everywhere; replace chosen points by trainable outputs.
105
+ with torch.no_grad():
106
+ g = self.frozen_anchor(coordinates)
107
+ return g.index_copy(-2, chosen, self.anchor(coordinates.index_select(-2, chosen)))
108
+
109
+ def forward(self, coordinates, features):
110
+ """coordinates [1,N,2] (shared grid), normalized features [B,N,3]."""
111
+ g = self.embed_geometry(coordinates)
112
+ f = self.physical(features)
113
+ states, transports = [], []
114
+ for size, assignment, encoder in zip(self.latent_sizes, self.assignment, self.encoder):
115
+ p = sinkhorn_transport(assignment(g), self.iterations, self.temperature)
116
+ transports.append(p)
117
+ g = torch.matmul(p.transpose(-1, -2), g) * size
118
+ f = encoder(torch.matmul(p.transpose(-1, -2), f) * size)
119
+ states.append(f)
120
+ f = self.decoder[-1](states[-1])
121
+ for level in range(len(states) - 2, -1, -1):
122
+ n = states[level].shape[-2]
123
+ f = self.decoder[level](states[level] + torch.matmul(transports[level + 1], f) * n)
124
+ return self.output(torch.matmul(transports[0], f) * features.shape[-2])
results/artifact_audit.json ADDED
@@ -0,0 +1,12 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "status": "success",
3
+ "job_id": "1100920",
4
+ "epochs": 500,
5
+ "updates": 125000,
6
+ "test_samples": 200,
7
+ "mean_relative_l2_raw": 0.006937251414878938,
8
+ "signed_gap": -6.274858512106234e-05,
9
+ "selection": "final_epoch_500_no_test_selection",
10
+ "checkpoint": "/public/home/liuyushuang/zhaomingsi/onescience-cfd-repro-packages/MoNo-light/weight/best_model.pth",
11
+ "report": "/public/home/liuyushuang/zhaomingsi/onescience-cfd-repro-packages/MoNo-light/results/reproduction_report.md"
12
+ }
results/data_manifest.json ADDED
@@ -0,0 +1,121 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "sources": [
3
+ {
4
+ "file": "/public/home/liuyushuang/zhaomingsi/onescience-cfd-data/MoNo-light/piececonst_r421_N1024_smooth1.mat",
5
+ "size": 1692307453,
6
+ "mtime_ns": 1794146095000000000
7
+ },
8
+ {
9
+ "file": "/public/home/liuyushuang/zhaomingsi/onescience-cfd-data/MoNo-light/piececonst_r421_N1024_smooth2.mat",
10
+ "size": 1692748750,
11
+ "mtime_ns": 1794146184000000000
12
+ }
13
+ ],
14
+ "schema": {
15
+ "train": [
16
+ [
17
+ "Kcoeff",
18
+ [
19
+ 1024,
20
+ 421,
21
+ 421
22
+ ],
23
+ "double"
24
+ ],
25
+ [
26
+ "Kcoeff_x",
27
+ [
28
+ 1024,
29
+ 421,
30
+ 421
31
+ ],
32
+ "double"
33
+ ],
34
+ [
35
+ "Kcoeff_y",
36
+ [
37
+ 1024,
38
+ 421,
39
+ 421
40
+ ],
41
+ "double"
42
+ ],
43
+ [
44
+ "coeff",
45
+ [
46
+ 1024,
47
+ 421,
48
+ 421
49
+ ],
50
+ "double"
51
+ ],
52
+ [
53
+ "sol",
54
+ [
55
+ 1024,
56
+ 421,
57
+ 421
58
+ ],
59
+ "double"
60
+ ]
61
+ ],
62
+ "test": [
63
+ [
64
+ "Kcoeff",
65
+ [
66
+ 1024,
67
+ 421,
68
+ 421
69
+ ],
70
+ "double"
71
+ ],
72
+ [
73
+ "Kcoeff_x",
74
+ [
75
+ 1024,
76
+ 421,
77
+ 421
78
+ ],
79
+ "double"
80
+ ],
81
+ [
82
+ "Kcoeff_y",
83
+ [
84
+ 1024,
85
+ 421,
86
+ 421
87
+ ],
88
+ "double"
89
+ ],
90
+ [
91
+ "coeff",
92
+ [
93
+ 1024,
94
+ 421,
95
+ 421
96
+ ],
97
+ "double"
98
+ ],
99
+ [
100
+ "sol",
101
+ [
102
+ 1024,
103
+ 421,
104
+ 421
105
+ ],
106
+ "double"
107
+ ]
108
+ ]
109
+ },
110
+ "train_samples": 1000,
111
+ "test_samples": 200,
112
+ "split": "file1[0:1000], file2[0:200]",
113
+ "downsample": "[:,::5,::5]",
114
+ "grid": [
115
+ 85,
116
+ 85
117
+ ],
118
+ "coeff": "permeability",
119
+ "sol": "pressure",
120
+ "units": "stored numerical scale; unspecified"
121
+ }
results/history.csv ADDED
@@ -0,0 +1,501 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ epoch,updates,train_relative_l2,test_relative_l2,learning_rate,epoch_seconds,elapsed_seconds
2
+ 1,250,0.5640980224609375,0.5633376240730286,1.0244272158737352e-05,22.05616069212556,22.060828271321952
3
+ 2,500,0.5627725830078125,0.563075840473175,1.0976847548535692e-05,15.196564155165106,37.38052320340648
4
+ 3,750,0.562616943359375,0.5617297887802124,1.2197003148094014e-05,15.246734781190753,52.74816608335823
5
+ 4,1000,0.5282906494140625,0.4548191726207733,1.3903534714817352e-05,15.337613353971392,68.20793928811327
6
+ 5,1250,0.36952410888671877,0.2758006751537323,1.6094757973353054e-05,15.276880800258368,83.60551797412336
7
+ 6,1500,0.2460223388671875,0.22884009778499603,1.8768510277900616e-05,15.2220717552118,98.94957066699862
8
+ 7,1750,0.22257058715820313,0.21685856580734253,2.1922152746652306e-05,15.32394469389692,114.39563910197467
9
+ 8,2000,0.2173551025390625,0.21502025425434113,2.5552572866260155e-05,15.449282397981733,129.96481813304126
10
+ 9,2250,0.2127647247314453,0.21001358330249786,2.9656187563755084e-05,15.388207325246185,145.47735466621816
11
+ 10,2500,0.20764627075195313,0.20085503160953522,3.4228946742891845e-05,15.389676490798593,160.99212522804737
12
+ 11,2750,0.20070780944824218,0.19636540114879608,3.9266337281423534e-05,15.380585588049144,176.49442824022844
13
+ 12,3000,0.1762288513183594,0.1600719690322876,4.4763387485365264e-05,15.476150880102068,192.09338091220707
14
+ 13,3250,0.1496522979736328,0.13183708488941193,5.071467199584828e-05,15.430152419023216,207.64669869933277
15
+ 14,3500,0.12127027893066407,0.11030237376689911,5.711431714372256e-05,15.365771345794201,223.13404673803598
16
+ 15,3750,0.10115877532958985,0.09192697703838348,6.39560067466229e-05,15.401633621193469,238.65722757112235
17
+ 16,4000,0.09068042755126954,0.08262644708156586,7.12329883427765e-05,15.546202002093196,254.33277691714466
18
+ 17,4250,0.08418280792236328,0.08248597383499146,7.893807985540145e-05,15.381681535858661,269.83707828307524
19
+ 18,4500,0.08100576019287109,0.07988909631967545,8.70636766811154e-05,15.402473694179207,285.36845789523795
20
+ 19,4750,0.07927720642089844,0.08002491295337677,9.560175919536246e-05,15.432647463865578,300.92682423302904
21
+ 20,5000,0.0734571533203125,0.07149950414896011,0.00010454390066744678,15.48054573405534,316.5434751561843
22
+ 21,5250,0.07334333801269531,0.07302560657262802,0.00011388127557736348,15.359153665136546,332.02784189535305
23
+ 22,5500,0.06969951629638672,0.06441071629524231,0.00012360466832621905,15.438961117062718,347.5911403731443
24
+ 23,5750,0.06794630432128906,0.0626569390296936,0.00013370448233164176,15.424303850159049,363.1363980770111
25
+ 24,6000,0.06591207885742187,0.0619669146835804,0.0001441707494992076,15.486168215982616,378.7453774083406
26
+ 25,6250,0.06643695068359375,0.0640878826379776,0.0001549931400605328,15.405439985916018,394.2750057121739
27
+ 26,6500,0.06368424987792969,0.07481248676776886,0.00016616097276832417,15.400536146946251,409.80008955299854
28
+ 27,6750,0.06115339279174805,0.059370070695877075,0.00017766322543832306,15.351495707873255,425.278838430997
29
+ 28,7000,0.05921088409423828,0.0621337927877903,0.00018948854582774175,15.400949634145945,440.80178946629167
30
+ 29,7250,0.05940121841430664,0.05934016406536102,0.00020162526283945147,15.401934700086713,456.32512305723503
31
+ 30,7500,0.059197086334228516,0.05411706119775772,0.0002140613980408687,15.338326334021986,471.78568975720555
32
+ 31,7750,0.056619102478027344,0.05696960911154747,0.00022678467748616748,15.36466471478343,487.2719206660986
33
+ 32,8000,0.05422004318237305,0.05425222963094711,0.0002397825438301501,15.4189224219881,502.8154009124264
34
+ 33,8250,0.05224772262573242,0.04666374996304512,0.0002530421687218237,15.34491026122123,518.2819164372049
35
+ 34,8500,0.0503892936706543,0.04673701897263527,0.0002665504654654453,15.259591876994818,533.666005638428
36
+ 35,8750,0.05302828598022461,0.050601836293935776,0.00028029410193654443,15.340377296786755,549.1310077412054
37
+ 36,9000,0.04934526062011719,0.0488077737390995,0.00029425951374017286,15.413852998055518,564.6654764669947
38
+ 37,9250,0.049592453002929686,0.049851685762405396,0.0003084329175983958,15.379934263881296,580.1822014441714
39
+ 38,9500,0.048840450286865233,0.05279238149523735,0.0003228003249538103,15.371501198969781,595.6780225429684
40
+ 39,9750,0.04726791763305664,0.044977378100156784,0.00033734755577566854,15.402522661723197,611.2037809370086
41
+ 40,10000,0.04577093505859375,0.05512531101703644,0.0003520602525549746,15.280102051794529,626.6045299083926
42
+ 41,10250,0.045427753448486326,0.04747654125094414,0.0003669238944747457,15.193811983801425,641.9381919763982
43
+ 42,10500,0.043981048583984376,0.039034437388181686,0.0003819238117414532,15.232642125338316,657.2932522413321
44
+ 43,10750,0.043081672668457034,0.04412568733096123,0.00039704520006349437,15.33949240297079,672.7555879773572
45
+ 44,11000,0.04193961334228516,0.04103700816631317,0.00041227313526241065,15.271463744807988,688.1479191323742
46
+ 45,11250,0.04372662734985352,0.04078379645943642,0.0004275925880024274,15.29543927917257,703.5658997399732
47
+ 46,11500,0.04232806777954102,0.04233556613326073,0.00044298843862378135,15.253947435878217,718.94163778238
48
+ 47,11750,0.04042987442016602,0.044361427426338196,0.0004584454920651921,15.316700919996947,734.4668041411787
49
+ 48,12000,0.043145767211914064,0.03931807354092598,0.00047394849286075305,15.418614755850285,750.0063490821049
50
+ 49,12250,0.04043648910522461,0.03617968410253525,0.0004894821401964392,15.325307075399905,765.4576181070879
51
+ 50,12500,0.04068456268310547,0.042579833418130875,0.0005050311030113705,15.285789232235402,780.8695743330754
52
+ 51,12750,0.041836559295654295,0.044003572314977646,0.0005205800351289292,15.31165401590988,796.3031784463674
53
+ 52,13000,0.04253176116943359,0.04772301763296127,0.0005361135904027943,15.325383997056633,811.7507984051481
54
+ 53,13250,0.03879388809204101,0.0396014042198658,0.0005516164378629475,15.297281785868108,827.1712664794177
55
+ 54,13500,0.04148298645019531,0.03861913084983826,0.0005670732768466986,15.266439008060843,842.5618090801872
56
+ 55,13750,0.039040836334228515,0.04107655957341194,0.0005824688520998022,15.339295255951583,858.0251947282813
57
+ 56,14000,0.040483036041259766,0.04577549546957016,0.0005977879688327554,15.276119383983314,873.4289586353116
58
+ 57,14250,0.03761217880249024,0.04737388715147972,0.0006130155077174186,15.17102260235697,888.7249923651107
59
+ 58,14500,0.038804290771484375,0.03606722131371498,0.000628136439809164,15.189307058230042,904.0357237202115
60
+ 59,14750,0.03869889068603516,0.04100201651453972,0.0006431358413798123,15.315366561058909,919.4752776562236
61
+ 60,15000,0.03775595092773437,0.04575824365019798,0.0006579989086467321,15.21123363217339,934.8108012462035
62
+ 61,15250,0.037105056762695314,0.056044768542051315,0.0006727109723835523,15.149370691739023,950.092250067275
63
+ 62,15500,0.036269363403320314,0.036774445325136185,0.0006872575123980772,15.253172736149281,965.470517475158
64
+ 63,15750,0.03607638931274414,0.0407162606716156,0.0007016241718631087,15.371278842911124,980.9634833033197
65
+ 64,16000,0.03534313201904297,0.04330037161707878,0.0007157967714860355,15.404377884231508,996.4911038042046
66
+ 65,16250,0.03510810852050781,0.030752861872315407,0.0007297613235032012,15.43535693641752,1012.0532062463462
67
+ 66,16500,0.03420325469970703,0.031657055020332336,0.0007435040454852446,15.35016519203782,1027.5274272281677
68
+ 67,16750,0.03507833480834961,0.03517494350671768,0.0007570113739397797,15.481478838250041,1043.1351445023902
69
+ 68,17000,0.03511650085449219,0.029792701825499535,0.0007702699776979967,15.382697103079408,1058.6407919093035
70
+ 69,17250,0.03359023284912109,0.03468478098511696,0.0007832667710719691,15.367541668005288,1074.1318383701146
71
+ 70,17500,0.033206844329833986,0.03288806229829788,0.0007959889267696805,15.34105641208589,1089.595873025246
72
+ 71,17750,0.03296038055419922,0.0321698822081089,0.0008084238885550252,15.31781157432124,1105.0442680912092
73
+ 72,18000,0.03273040008544922,0.028872504830360413,0.0008205593836402895,15.365776046179235,1120.533270795364
74
+ 73,18250,0.03280875015258789,0.03128401190042496,0.0008323834347988805,15.371878099162132,1136.0308755552396
75
+ 74,18500,0.033220829010009764,0.03005462884902954,0.0008438843721863471,15.393524464219809,1151.547118339222
76
+ 75,18750,0.033482513427734374,0.03157724440097809,0.0008550508448580308,15.292654593940824,1166.9623817112297
77
+ 76,19000,0.0317713623046875,0.02860613912343979,0.0008658718319719713,15.2968532089144,1182.3833054522984
78
+ 77,19250,0.031253894805908206,0.034648723900318146,0.0008763366536660197,15.341989480890334,1197.8505415152758
79
+ 78,19500,0.03156649398803711,0.02745087817311287,0.0008864349815984138,15.36122271604836,1213.3326398599893
80
+ 79,19750,0.029795495986938477,0.032256998121738434,0.0008961568491414223,15.343163981102407,1228.8104129112326
81
+ 80,20000,0.030693483352661134,0.02920437604188919,0.0009054926612179872,15.330030777025968,1244.2606786959805
82
+ 81,20250,0.02921798324584961,0.026422323659062386,0.0009144332037716662,15.37628030218184,1259.7598919621669
83
+ 82,20500,0.03095605278015137,0.03660924732685089,0.0009229696528605205,15.370581715833396,1275.2609026981518
84
+ 83,20750,0.029778362274169922,0.032760728150606155,0.0009310935833659769,15.43544763000682,1290.8289489652961
85
+ 84,21000,0.03036695098876953,0.03138549625873566,0.0009387969773080677,15.392148920334876,1306.3441214202903
86
+ 85,21250,0.02906047248840332,0.027523793280124664,0.0009460722317588418,15.31413179077208,1321.7848892873153
87
+ 86,21500,0.029129297256469726,0.02902698516845703,0.0009529121663461376,15.370641463901848,1337.280976077076
88
+ 87,21750,0.0292639217376709,0.035469748079776764,0.0009593100303403094,15.307326765730977,1352.7108373800293
89
+ 88,22000,0.028019851684570312,0.031048152595758438,0.0009652595093169163,15.305147528182715,1368.1394503312185
90
+ 89,22250,0.029516941070556642,0.026217196136713028,0.0009707547313887947,15.413621977902949,1383.675802309066
91
+ 90,22500,0.027790195465087892,0.031737279146909714,0.0009757902730013659,15.452553453855217,1399.2623586733826
92
+ 91,22750,0.028081356048583986,0.026995060965418816,0.0009803611642854575,15.329735392238945,1414.718174222391
93
+ 92,23000,0.026730985641479493,0.027565333992242813,0.0009844628939623574,15.392243755981326,1430.2326527293772
94
+ 93,23250,0.025501590728759765,0.029425883665680885,0.0009880914137962566,15.377958202268928,1445.7398415412754
95
+ 94,23500,0.025796905517578125,0.024252695962786674,0.0009912431425896885,15.321847414597869,1461.1849262951873
96
+ 95,23750,0.026245546340942384,0.02474229782819748,0.000993914969718023,15.389415725134313,1476.6994751612656
97
+ 96,24000,0.025168867111206056,0.022422926500439644,0.0009961042581995204,15.342533451039344,1492.2555021443404
98
+ 97,24250,0.02692961311340332,0.02706611156463623,0.0009978088472979247,15.401350571773946,1507.7821266991086
99
+ 98,24500,0.02554169464111328,0.027285141870379448,0.0009990270546550183,15.377941998187453,1523.2863088161685
100
+ 99,24750,0.025728776931762697,0.02385311760008335,0.00099975767795104,15.386975768953562,1538.798459946178
101
+ 100,25000,0.024998138427734375,0.0211489200592041,0.0009999999997532846,15.31348335929215,1554.234198378399
102
+ 101,25250,0.024259906768798828,0.022689487785100937,0.000999984456761355,15.31342512788251,1569.6720179752447
103
+ 102,25500,0.02430233383178711,0.02138766273856163,0.0009999380754572309,15.417614683043212,1585.2099127583206
104
+ 103,25750,0.024668550491333006,0.023540282621979713,0.0009998608587019294,15.337571161799133,1600.6722320551053
105
+ 104,26000,0.02402964973449707,0.03208734095096588,0.000999752811258544,15.355833450332284,1616.1507841059938
106
+ 105,26250,0.02507335090637207,0.023593630641698837,0.0009996139397919497,15.351557733956724,1631.6260260073468
107
+ 106,26500,0.023035898208618164,0.026218533515930176,0.0009994442528683926,15.226642586290836,1646.9747782833874
108
+ 107,26750,0.022602928161621093,0.021744554862380028,0.000999243760954962,15.296616556122899,1662.3972467733547
109
+ 108,27000,0.023951927185058595,0.018859149888157845,0.0009990124764189428,15.300236796960235,1677.8220949131064
110
+ 109,27250,0.02381962013244629,0.02244798094034195,0.000998750413527055,15.2846939410083,1693.2297941599973
111
+ 110,27500,0.02210795211791992,0.02394491620361805,0.0009984575884445723,15.376196226105094,1708.7282623769715
112
+ 111,27750,0.022047779083251954,0.023899616673588753,0.0009981340192343248,15.253152412362397,1724.1042996603064
113
+ 112,28000,0.022575065612792967,0.02721894346177578,0.0009977797258555858,15.24655719101429,1739.4735127082095
114
+ 113,28250,0.02217424964904785,0.02268831804394722,0.0009973947301628397,15.28984751785174,1754.8849952360615
115
+ 114,28500,0.02159435272216797,0.02068471908569336,0.0009969790559044343,15.20725304307416,1770.212937453296
116
+ 115,28750,0.022252925872802733,0.020153038203716278,0.0009965327287211158,15.245433167088777,1785.5820914143696
117
+ 116,29000,0.02220391082763672,0.02230728790163994,0.0009960557761444474,15.22893429826945,1800.9351666644216
118
+ 117,29250,0.021181577682495116,0.02104806713759899,0.0009955482275951105,15.308647123165429,1816.3681807531975
119
+ 118,29500,0.021506397247314454,0.025478024035692215,0.0009950101143810907,15.25823638215661,1831.749237793032
120
+ 119,29750,0.021197025299072265,0.021358245983719826,0.0009944414696957454,15.300569294951856,1847.173497111071
121
+ 120,30000,0.02126473617553711,0.020482292398810387,0.0009938423286157576,15.237412501126528,1862.53337045433
122
+ 121,30250,0.02032640075683594,0.020366892218589783,0.0009932127280989711,15.259513173718005,1877.913269605022
123
+ 122,30500,0.019908342361450195,0.023982824757695198,0.0009925527069821118,15.195682135876268,1893.2314553833567
124
+ 123,30750,0.021352767944335938,0.021069303154945374,0.0009918623059783907,15.231249706819654,1908.5842112870887
125
+ 124,31000,0.020045007705688477,0.016888482496142387,0.000991141567674994,15.22820430668071,1923.935725021176
126
+ 125,31250,0.020962337493896484,0.018971426412463188,0.0009903905365304558,15.272018235176802,1939.3300762660801
127
+ 126,31500,0.019833038330078125,0.019927198067307472,0.0009896092588719137,15.270483948290348,1954.7253908463754
128
+ 127,31750,0.019819419860839845,0.01948411576449871,0.0009887977828922547,15.226408480666578,1970.0759828491136
129
+ 128,32000,0.019862510681152343,0.016500884667038918,0.0009879561586471392,15.176401157863438,1985.375199857168
130
+ 129,32250,0.01918644905090332,0.01813567988574505,0.000987084438051915,15.295288131572306,2000.7966925064102
131
+ 130,32500,0.0185704345703125,0.01739833503961563,0.000986182674878415,15.226731287315488,2016.1485259602778
132
+ 131,32750,0.020185272216796876,0.019044214859604836,0.000985250924751639,15.198386389762163,2031.4726655911654
133
+ 132,33000,0.018798723220825195,0.01962823048233986,0.0009842892451463238,15.309917441103607,2046.9038984691724
134
+ 133,33250,0.01889006996154785,0.017354736104607582,0.0009832976953833975,15.281233867164701,2062.3105547423474
135
+ 134,33500,0.01832650947570801,0.023082010447978973,0.0009822763366263202,15.338594720698893,2077.7893520384096
136
+ 135,33750,0.018963077545166015,0.015460704453289509,0.0009812252318773108,15.27336859004572,2093.1845557633787
137
+ 136,34000,0.01806884765625,0.021335110068321228,0.0009801444459734616,15.269187728874385,2108.5791012360714
138
+ 137,34250,0.017949586868286134,0.025710538029670715,0.000979034045582738,15.305923258885741,2124.011368006002
139
+ 138,34500,0.018485679626464843,0.025214292109012604,0.0009778940991998664,15.1576749230735,2139.292259396985
140
+ 139,34750,0.018244838714599608,0.01637757383286953,0.0009767246771421097,15.129236069973558,2154.5463391924277
141
+ 140,35000,0.017565183639526367,0.024474721401929855,0.0009755258515449284,15.228452512994409,2169.9035056452267
142
+ 141,35250,0.01811922836303711,0.016961760818958282,0.0009742976963575327,15.148994943127036,2185.1766669722274
143
+ 142,35500,0.018507806777954102,0.015517051331698895,0.0009730402873383196,15.227274203207344,2200.526412810199
144
+ 143,35750,0.01783013916015625,0.015871817246079445,0.0009717537020502002,15.232639902736992,2215.882810200099
145
+ 144,36000,0.017449542999267577,0.01804148219525814,0.0009704380198558156,15.242578560020775,2231.2459027403966
146
+ 145,36250,0.017318164825439453,0.018343903124332428,0.0009690933219126408,15.223069502040744,2246.5922766700387
147
+ 146,36500,0.01623832130432129,0.0188971646130085,0.0009677196911679792,15.121383598074317,2261.835593251977
148
+ 147,36750,0.01727017021179199,0.01840318739414215,0.0009663172123538455,15.130378675181419,2277.087757529225
149
+ 148,37000,0.01682404899597168,0.014540526084601879,0.0009648859719817391,15.153064768761396,2292.359020390082
150
+ 149,37250,0.01656670951843262,0.015768393874168396,0.0009634260583373079,15.196589762810618,2307.765496519394
151
+ 150,37500,0.0170563907623291,0.017894065007567406,0.0009619375614749023,15.241198915056884,2323.127799644135
152
+ 151,37750,0.016095441818237306,0.014568889513611794,0.0009604205732120202,15.31419395795092,2338.564770560246
153
+ 152,38000,0.017014595031738283,0.01633778214454651,0.0009588751871236435,15.307043139822781,2353.994678232353
154
+ 153,38250,0.016457786560058595,0.016152389347553253,0.0009573014985364653,15.275430395733565,2369.394355654251
155
+ 154,38500,0.016153282165527343,0.01560987439006567,0.0009556996045230106,15.29520108923316,2384.811823685188
156
+ 155,38750,0.01542067813873291,0.01660224050283432,0.0009540696038956477,15.271534443832934,2400.2075660252012
157
+ 156,39000,0.01629313850402832,0.016501756384968758,0.0009524115972004931,15.259340829681605,2415.591650712304
158
+ 157,39250,0.0160687255859375,0.017469491809606552,0.0009507256867112098,15.313146028202027,2431.029422186315
159
+ 158,39500,0.016311838150024415,0.017864223569631577,0.0009490119764226982,15.21291619585827,2446.363668011967
160
+ 159,39750,0.01574165725708008,0.015438240952789783,0.0009472705720446811,15.279652433935553,2461.76745643327
161
+ 160,40000,0.01565931510925293,0.014562265947461128,0.0009455015809951832,15.312816669698805,2477.204263358377
162
+ 161,40250,0.01634073257446289,0.018651096150279045,0.0009437051123939054,15.26649016700685,2492.593971647322
163
+ 162,40500,0.016218894958496093,0.014371490105986595,0.0009418812770554929,15.337469606194645,2508.0534672853537
164
+ 163,40750,0.01525035285949707,0.014260453172028065,0.0009400301874827003,15.329126324970275,2523.5071758762933
165
+ 164,41000,0.014816065788269044,0.014608028344810009,0.0009381519578594523,15.252174146007746,2538.8831661944278
166
+ 165,41250,0.015726679801940917,0.014140292070806026,0.000936246704043799,15.185148224234581,2554.1900756941177
167
+ 166,41500,0.015198017120361328,0.013827071525156498,0.0009343145435607702,15.200790227856487,2569.514027712401
168
+ 167,41750,0.01439855670928955,0.015459885820746422,0.000932355595595126,15.19557928899303,2584.8336760471575
169
+ 168,42000,0.0146421537399292,0.01328720897436142,0.0009303699809840044,15.240343700163066,2600.1996567002498
170
+ 169,42250,0.015046365737915039,0.013916640542447567,0.0009283578222094676,15.190979177132249,2615.5147514929995
171
+ 170,42500,0.014747878074645997,0.01571376994252205,0.0009263192433909469,15.231941368896514,2630.871354468167
172
+ 171,42750,0.014729454040527343,0.014488416723906994,0.0009242543702775865,15.267052461858839,2646.2602205569856
173
+ 172,43000,0.014419453620910645,0.012942531146109104,0.0009221633302404863,15.313674652017653,2661.6982831102796
174
+ 173,43250,0.014437777519226074,0.01695185713469982,0.0009200462522648456,15.38540296535939,2677.2095201490447
175
+ 174,43500,0.01370613193511963,0.013050079345703125,0.000917903266942006,15.304132082033902,2692.636337572243
176
+ 175,43750,0.014353767395019531,0.01453403476625681,0.0009157345064613971,15.28373791789636,2708.045928635169
177
+ 176,44000,0.015319687843322754,0.015654241666197777,0.0009135401046023807,15.437781134620309,2723.6080035581253
178
+ 177,44250,0.014509757041931153,0.013610059395432472,0.0009113201967260002,15.375121218152344,2739.10676933825
179
+ 178,44500,0.01414177131652832,0.016241272911429405,0.0009090749197666299,15.323394665960222,2754.5551522080787
180
+ 179,44750,0.01390151309967041,0.0131221953779459,0.000906804412223529,15.337718636728823,2770.0206147972494
181
+ 180,45000,0.015003868103027344,0.015681076794862747,0.0009045088141522979,15.332814750261605,2785.475346050225
182
+ 181,45250,0.013782325744628907,0.018389195203781128,0.0009021882671562387,15.317305179312825,2800.916840595193
183
+ 182,45500,0.01345441722869873,0.01753031648695469,0.0008998429143776209,15.208518150728196,2816.250805873424
184
+ 183,45750,0.013658947944641113,0.014348933473229408,0.0008974729004888517,15.247777451295406,2831.6205486231484
185
+ 184,46000,0.013163713455200196,0.013621244579553604,0.0008950783716835511,15.309364275075495,2847.0529541149735
186
+ 185,46250,0.014085173606872559,0.013025525957345963,0.0008926594756675355,15.419524621684104,2862.5934587293305
187
+ 186,46500,0.013021725654602051,0.01283149141818285,0.0008902163616497047,15.333673036191612,2878.0518516292796
188
+ 187,46750,0.013152226448059082,0.012210401706397533,0.00088774918033284,15.375478108879179,2893.5505612362176
189
+ 188,47000,0.012861348152160645,0.013034491799771786,0.0008852580839043061,15.362280137371272,2909.03780318005
190
+ 189,47250,0.013095233917236328,0.012526562437415123,0.0008827432260266648,15.24394895369187,2924.403976195026
191
+ 190,47500,0.013558761596679687,0.012882002629339695,0.0008802047618281961,15.297735143918544,2939.8276688773185
192
+ 191,47750,0.012929742813110351,0.014546799473464489,0.0008776428478933292,15.429319156799465,2955.3814552873373
193
+ 192,48000,0.013659934043884277,0.012128624133765697,0.000875057642252983,15.323774951975793,2970.827919722069
194
+ 193,48250,0.013184808731079101,0.014870553277432919,0.0008724493043748189,15.354452323168516,2986.3081362703815
195
+ 194,48500,0.012923858642578124,0.013121486641466618,0.0008698179951534037,15.313494220841676,3001.746706505306
196
+ 195,48750,0.01324846363067627,0.013164393603801727,0.0008671638769002846,15.362893811892718,3017.2346646701917
197
+ 196,49000,0.0125816650390625,0.011965429410338402,0.0008644871133339776,15.270664919633418,3032.6302361912094
198
+ 197,49250,0.01275950527191162,0.015404879115521908,0.000861787869569868,15.174907303415239,3047.9299003630877
199
+ 198,49500,0.01265218734741211,0.014406242407858372,0.0008590663121100259,15.282469796016812,3063.337697891053
200
+ 199,49750,0.01247257137298584,0.012895669788122177,0.0008563226088329349,15.34149232134223,3078.891110213008
201
+ 200,50000,0.012333866119384765,0.013602638617157936,0.0008535569289831373,15.244960744399577,3094.259712621104
202
+ 201,50250,0.012932848930358887,0.013560743071138859,0.0008507694431607938,15.322351478971541,3109.7063562143594
203
+ 202,50500,0.01204676628112793,0.011577198281884193,0.0008479603233111601,15.338768524117768,3125.1705671520904
204
+ 203,50750,0.011992451667785644,0.0128812650218606,0.0008451297427139804,15.37585277017206,3140.6663773939945
205
+ 204,51000,0.011982297897338868,0.013377590104937553,0.0008422778759727995,15.274415529798716,3156.063799967058
206
+ 205,51250,0.012460744857788086,0.015456019900739193,0.0008394048990041915,15.262975873891264,3171.4505955781788
207
+ 206,51500,0.012303523063659668,0.012236809358000755,0.0008365109890269085,15.260005007963628,3186.834569875151
208
+ 207,51750,0.012359234809875488,0.013492295518517494,0.0008335963245509494,15.318697806913406,3202.2767040519975
209
+ 208,52000,0.01206342887878418,0.013984747231006622,0.0008306610853665489,15.342009730171412,3217.7422932842746
210
+ 209,52250,0.012604329109191894,0.014308244921267033,0.0008277054525330855,15.379509671125561,3233.247409542091
211
+ 210,52500,0.011709978103637695,0.013472506776452065,0.000824729608367915,15.317263348028064,3248.689549744129
212
+ 211,52750,0.01151204776763916,0.014717795886099339,0.0008217337364351227,15.371119969058782,3264.1838476164266
213
+ 212,53000,0.01199649715423584,0.012130802497267723,0.0008187180215342015,15.376245946157724,3279.6838330151513
214
+ 213,53250,0.012148633003234863,0.012551362626254559,0.0008156826496886515,15.48570654913783,3295.296061153058
215
+ 214,53500,0.012505068778991699,0.01187924761325121,0.0008126278081345063,15.541852020192891,3310.9625411001034
216
+ 215,53750,0.011748524665832519,0.011167139746248722,0.0008095536853087823,15.462777303066105,3326.552950471174
217
+ 216,54000,0.01129755973815918,0.012634340673685074,0.0008064604708378561,15.490200845059007,3342.171700670384
218
+ 217,54250,0.011321988105773926,0.012500166893005371,0.0008033483555257664,15.481780932750553,3357.7779496582225
219
+ 218,54500,0.010962491989135742,0.011491253040730953,0.0008002175313424452,15.51824337663129,3373.420596969314
220
+ 219,54750,0.011959577560424805,0.011148912832140923,0.0007970681914118751,15.42715336382389,3388.9728386681527
221
+ 220,55000,0.011522401809692382,0.013529356569051743,0.0007939005300001785,15.412652869243175,3404.5104921814054
222
+ 221,55250,0.011493789672851562,0.011323336511850357,0.0007907147425036316,15.467607453931123,3420.1038267123513
223
+ 222,55500,0.010903006553649902,0.011946320533752441,0.000787511025436614,15.514765048865229,3435.742205671966
224
+ 223,55750,0.011269472122192382,0.011723630130290985,0.0007842895764194852,15.452445758040994,3451.3183482233435
225
+ 224,56000,0.010697693824768066,0.012571688741445541,0.0007810505941663948,15.392676630988717,3466.8361986433156
226
+ 225,56250,0.010927980422973632,0.012845725752413273,0.0007777942784730255,15.403129997197539,3482.3659384422936
227
+ 226,56500,0.010787775039672851,0.013036439195275307,0.0007745208302042673,15.459749687928706,3497.9514814293943
228
+ 227,56750,0.010916512489318847,0.010162041522562504,0.0007712304512818292,15.421556204091758,3513.4979962380603
229
+ 228,57000,0.01076762866973877,0.011156890541315079,0.000767923344671782,15.459682156797498,3529.085434406996
230
+ 229,57250,0.010857117652893067,0.01135971024632454,0.0007645997143720393,15.455717999953777,3544.6668224842288
231
+ 230,57500,0.01074506664276123,0.011491135694086552,0.0007612597653997739,15.508270544931293,3560.302213130053
232
+ 231,57750,0.010505529403686523,0.012468491680920124,0.0007579037037787706,15.432723210658878,3575.8612982803024
233
+ 232,58000,0.010227849960327148,0.010652260854840279,0.000754531736526719,15.442280960734934,3591.4305752911605
234
+ 233,58250,0.010378185272216798,0.015071863308548927,0.0007511440716424422,15.447220464702696,3607.0033647920936
235
+ 234,58500,0.011123008728027343,0.01193546038120985,0.0007477409180930681,15.468471732921898,3622.597145056352
236
+ 235,58750,0.01080122947692871,0.010721530765295029,0.0007443224858011374,15.474804973695427,3638.1986035052687
237
+ 236,59000,0.010108009338378906,0.013473277911543846,0.0007408889856316566,15.43142807064578,3653.755591872148
238
+ 237,59250,0.010526034355163574,0.01132947951555252,0.0007374406293790888,15.413584854919463,3669.2933070193976
239
+ 238,59500,0.009657110214233398,0.011614074930548668,0.0007339776297542914,15.534396037925035,3684.950793202035
240
+ 239,59750,0.01083056640625,0.013886978849768639,0.000730500200371393,15.365314326714724,3700.44052515598
241
+ 240,60000,0.010227033615112305,0.010669414885342121,0.0007270085557346186,15.389095155987889,3715.9560028431006
242
+ 241,60250,0.009885171890258788,0.011151749640703201,0.000723502911225056,15.330196803901345,3731.413451962173
243
+ 242,60500,0.010639625549316406,0.010999901220202446,0.000719983483087372,15.402732045855373,3746.9415380610153
244
+ 243,60750,0.009808377265930176,0.012618106789886951,0.000716450488416472,15.478569102007896,3762.546422822401
245
+ 244,61000,0.009884139060974122,0.010515531525015831,0.0007129041451441091,15.4023597240448,3778.0734005169943
246
+ 245,61250,0.009957359313964844,0.0104208430275321,0.000709344672025441,15.388122173957527,3793.585598933976
247
+ 246,61500,0.009777658462524414,0.009921019896864891,0.0007057722886255367,15.296892685815692,3809.0065187313594
248
+ 247,61750,0.009602244377136231,0.01146406028419733,0.0007021872153058314,15.317237399984151,3824.449861417059
249
+ 248,62000,0.009573838233947755,0.009792991913855076,0.0006985896732105347,15.345470807049423,3839.9176354971714
250
+ 249,62250,0.00972711181640625,0.011541123501956463,0.0006949798842529891,15.369084205012769,3855.41588580003
251
+ 250,62500,0.009251748085021973,0.010819674469530582,0.000691358071101981,15.30637610796839,3870.8441445729695
252
+ 251,62750,0.009925447463989257,0.009907126426696777,0.0006877244571680057,15.313810353167355,3886.283691046294
253
+ 252,63000,0.009571197509765626,0.010428155772387981,0.0006840792665894869,15.300154733937234,3901.795157614164
254
+ 253,63250,0.009432005882263183,0.010489230044186115,0.0006804227242189494,15.377300604712218,3917.296789376065
255
+ 254,63500,0.00907388401031494,0.009456549771130085,0.0006767550556091508,15.300551411230117,3932.7192705031484
256
+ 255,63750,0.00929893398284912,0.010493620298802853,0.000673076486999167,15.252002256922424,3948.1078813839704
257
+ 256,64000,0.009107268333435059,0.009785753674805164,0.0006693872453004376,15.249166336841881,3963.4809947181493
258
+ 257,64250,0.009025053024291992,0.00960873905569315,0.0006656875580827687,15.330673518590629,3978.936433318071
259
+ 258,64500,0.00888148307800293,0.011670705862343311,0.0006619776535602944,15.226177600678056,3994.2848726972006
260
+ 259,64750,0.009208748817443848,0.011243895627558231,0.0006582577605774012,15.218726764898747,4009.631045187358
261
+ 260,65000,0.00889033317565918,0.011285082437098026,0.0006545281085946105,15.22692823735997,4024.981797699351
262
+ 261,65250,0.009355822563171386,0.011386020109057426,0.0006507889276744248,15.273033517878503,4040.3813169002533
263
+ 262,65500,0.008524689674377441,0.010720652528107166,0.0006470404484671365,15.224637552164495,4055.735939532984
264
+ 263,65750,0.008564793586730957,0.010315613821148872,0.0006432829021965998,15.208647812716663,4071.067465438042
265
+ 264,66000,0.009095388412475586,0.009199217893183231,0.0006395165206459684,15.28220861684531,4086.4725624360144
266
+ 265,66250,0.008870038032531738,0.010986574925482273,0.0006357415361433977,15.255739254876971,4101.852650227025
267
+ 266,66500,0.00861741542816162,0.009909437038004398,0.0006319581815477133,15.256487828213722,4117.2321335393935
268
+ 267,66750,0.008890153884887696,0.009357238188385963,0.0006281666902340482,15.28647333290428,4132.642422389239
269
+ 268,67000,0.008696987152099609,0.01031520962715149,0.0006243672960794459,15.176955880131572,4147.9449570770375
270
+ 269,67250,0.008561318397521972,0.009135612286627293,0.0006205602334484348,15.223166173323989,4163.293312420137
271
+ 270,67500,0.008690460205078125,0.010690970346331596,0.0006167457371785704,15.201174871996045,4178.617394363973
272
+ 271,67750,0.008219593048095703,0.010743317194283009,0.000612924042565951,15.223127292003483,4193.96730112331
273
+ 272,68000,0.008633996963500977,0.00914310198277235,0.000609095385350702,15.344586045946926,4209.437186122406
274
+ 273,68250,0.008030777931213378,0.009057486429810524,0.0006052600017024348,15.292865309864283,4224.850707296282
275
+ 274,68500,0.008431404113769532,0.009402274154126644,0.0006014181282056794,15.192441636696458,4240.170652182307
276
+ 275,68750,0.00819501495361328,0.00936892256140709,0.0005975700018452896,15.182272035162896,4255.477418836206
277
+ 276,69000,0.0082567777633667,0.008979836478829384,0.0005937158599918258,15.20290220528841,4270.803790523205
278
+ 277,69250,0.007978610515594483,0.01045750267803669,0.0005898559403869121,15.276480403728783,4286.199918684084
279
+ 278,69500,0.008127251625061034,0.009448244236409664,0.0005859904811285718,15.38496467191726,4301.711809316184
280
+ 279,69750,0.00836260986328125,0.008864480070769787,0.0005821197206565393,15.406913831830025,4317.2419988974
281
+ 280,70000,0.00827000617980957,0.009777288883924484,0.0005782438977375541,15.393791210837662,4332.760176765267
282
+ 281,70250,0.008001694679260254,0.010214585810899734,0.0005743632514506301,15.410814831033349,4348.296583123971
283
+ 282,70500,0.008175464630126953,0.009548209607601166,0.0005704780211723102,15.234542266931385,4363.6557036913
284
+ 283,70750,0.008014219284057617,0.008965665474534035,0.0005665884465618987,15.311661596875638,4379.094147165306
285
+ 284,71000,0.007845052242279053,0.009278013370931149,0.0005626947675466792,15.39034919301048,4394.616083807312
286
+ 285,71250,0.007763772010803222,0.008648162707686424,0.0005587972243071141,15.309686679393053,4410.049885808025
287
+ 286,71500,0.007553804874420166,0.009719133377075195,0.0005548960572620298,15.403966567944735,4425.578192625195
288
+ 287,71750,0.007655575275421143,0.009281124919652939,0.0005509915070537853,15.445379047188908,4441.147477674298
289
+ 288,72000,0.007643978118896485,0.009164913557469845,0.0005470838145334299,15.43516123527661,4456.706701034214
290
+ 289,72250,0.00791244125366211,0.008949099108576775,0.0005431732207458452,15.452482333872467,4472.28340100497
291
+ 290,72500,0.0076171278953552244,0.009472756646573544,0.0005392599669148762,15.436600603163242,4487.844100234099
292
+ 291,72750,0.0076296329498291016,0.008763611316680908,0.0005353442944284525,15.412293762899935,4503.37943922542
293
+ 292,73000,0.007605233669281006,0.008907829411327839,0.0005314264448236975,15.514422257896513,4519.020059874281
294
+ 293,73250,0.007941579818725586,0.009304161183536053,0.0005275066597720295,15.403027215972543,4534.549850131385
295
+ 294,73500,0.0073641505241394045,0.009145785123109818,0.0005235851810642545,15.316140829119831,4549.989694059361
296
+ 295,73750,0.00729433536529541,0.009065118618309498,0.0005196622505956505,15.416471152566373,4565.53006906528
297
+ 296,74000,0.007370599269866943,0.009524212218821049,0.0005157381103510481,15.423613493796438,4581.0784383961
298
+ 297,74250,0.0072201089859008786,0.009242490865290165,0.0005118130023899017,15.361528988927603,4596.563940349035
299
+ 298,74500,0.00751288366317749,0.00960018951445818,0.0005078871688313591,15.29481904907152,4611.982134922408
300
+ 299,74750,0.007366073131561279,0.008793370798230171,0.0005039608518393266,15.350293786730617,4627.45754358219
301
+ 300,75000,0.007199713230133057,0.008389965631067753,0.000500034293607531,15.35459652915597,4642.937522358261
302
+ 301,75250,0.006991016387939453,0.008662061765789986,0.0004961077363445795,15.359479040838778,4658.4205291662365
303
+ 302,75500,0.007384171485900879,0.008812970481812954,0.0004921814222590202,15.39835077803582,4674.029453301337
304
+ 303,75750,0.007080995559692383,0.008651438169181347,0.0004882555935444001,15.361481255851686,4689.5185019811615
305
+ 304,76000,0.006745595455169678,0.008505295030772686,0.0004843304923643268,15.380346334073693,4705.021793841384
306
+ 305,76250,0.007025896072387695,0.009188290685415268,0.00048040636083752995,15.320213558152318,4720.473263306078
307
+ 306,76500,0.006919732570648193,0.009274912066757679,0.0004764834410229266,15.380388942081481,4735.977030242328
308
+ 307,76750,0.006911533832550049,0.008469218388199806,0.00047256197490468924,15.40273610316217,4751.5060099270195
309
+ 308,77000,0.0076693973541259765,0.00925340037792921,0.00046864220437731993,15.335301423911005,4766.9645213112235
310
+ 309,77250,0.0067156200408935544,0.008209007792174816,0.0004647243712307284,15.22858139174059,4782.31994543504
311
+ 310,77500,0.006845987319946289,0.008137403056025505,0.00046080871713531825,15.296688500791788,4797.7427554363385
312
+ 311,77750,0.006428201198577881,0.008116974495351315,0.0004568954836270779,15.352699527982622,4813.219640971161
313
+ 312,78000,0.006705593109130859,0.009596765041351318,0.0004529849120926834,15.304768678266555,4828.646861887071
314
+ 313,78250,0.006968065738677979,0.008465379476547241,0.00044907724375460736,15.28607281204313,4844.057585563976
315
+ 314,78500,0.006504374980926514,0.008379952982068062,0.00044517271965623985,15.33189497794956,4859.514655456413
316
+ 315,78750,0.0067659058570861816,0.010377254337072372,0.00044127158064701877,15.38024098938331,4875.020611867309
317
+ 316,79000,0.00640078067779541,0.008631445467472076,0.0004373740673675746,15.39508654968813,4890.535299775191
318
+ 317,79250,0.0067115740776062015,0.008841326460242271,0.00043348042023488506,15.336850216146559,4905.997421786189
319
+ 318,79500,0.006841792106628418,0.008266616612672806,0.0004295908794274463,15.306805898901075,4921.428570813965
320
+ 319,79750,0.006500396728515625,0.009085770696401596,0.0004257056848704565,15.389137608930469,4936.9420785224065
321
+ 320,80000,0.00627678394317627,0.008220870979130268,0.0004218250762210169,15.368169764988124,4952.432896788232
322
+ 321,80250,0.006312687397003174,0.008684355765581131,0.0004179492928533483,15.289663058705628,4967.845291079022
323
+ 322,80500,0.006229059219360351,0.007921664975583553,0.00041407857384402526,15.316254185978323,4983.284537752159
324
+ 323,80750,0.00615611743927002,0.008051615208387375,0.0004102131579572293,15.279143532272428,4998.687735865358
325
+ 324,81000,0.006120932102203369,0.008585390634834766,0.0004063532836300196,15.342967708129436,5014.14998458419
326
+ 325,81250,0.006660331249237061,0.009281967766582966,0.0004024991889576267,15.391251121647656,5029.663083721418
327
+ 326,81500,0.006155280113220215,0.008222008123993874,0.0003986511116787641,15.333136433269829,5045.122690435965
328
+ 327,81750,0.006034309387207031,0.008248448371887207,0.00039480928916096474,15.469891790766269,5060.719200700987
329
+ 328,82000,0.0061392135620117185,0.007818457670509815,0.0003909739583859378,15.363623300101608,5076.22423124034
330
+ 329,82250,0.006097472190856933,0.00786806270480156,0.00038714535593495174,15.357701260130852,5091.709708448965
331
+ 330,82500,0.0060310311317443846,0.007837338373064995,0.00038332371797423977,15.391943397000432,5107.230175657198
332
+ 331,82750,0.005867375373840332,0.008168465457856655,0.00037950928024043333,15.385206678882241,5122.743823876139
333
+ 332,83000,0.006037099838256836,0.008278660476207733,0.00037570227802601893,15.433668594341725,5138.300124822184
334
+ 333,83250,0.005818454265594483,0.008452964015305042,0.000371902946164826,15.38240901241079,5153.810973844025
335
+ 334,83500,0.005892965793609619,0.008528104051947594,0.0003681115190175401,15.355108694173396,5169.292115707416
336
+ 335,83750,0.005861095905303955,0.007897097617387772,0.0003643282304572471,15.404504340142012,5184.822761718184
337
+ 336,84000,0.0058440942764282225,0.007868960499763489,0.00036055331385500595,15.428232682868838,5200.377440058161
338
+ 337,84250,0.0057880220413208,0.008388528600335121,0.00035678700206545437,15.380744425114244,5215.890261965338
339
+ 338,84500,0.005602981090545654,0.007727970369160175,0.00035302952741244427,15.441930593922734,5231.456842256244
340
+ 339,84750,0.005795659542083741,0.007769661955535412,0.00034928112167471187,15.441614503972232,5247.026075136382
341
+ 340,85000,0.005715486526489258,0.007723961491137743,0.0003455420160715791,15.362668082583696,5262.510528938379
342
+ 341,85250,0.005720527648925782,0.007684564683586359,0.0003418124412486928,15.387402710039169,5278.0254198731855
343
+ 342,85500,0.0061541728973388676,0.007710649166256189,0.0003380926272637956,15.432460107840598,5293.582925484981
344
+ 343,85750,0.005399314403533936,0.007619248703122139,0.0003343828035725363,15.400511227082461,5309.106724246405
345
+ 344,86000,0.005650026321411133,0.007825893349945545,0.00033068319901431465,15.378132698591799,5324.608593989164
346
+ 345,86250,0.00553248929977417,0.007525918539613485,0.00032699404179816675,15.379805096890777,5340.113002915401
347
+ 346,86500,0.005646389484405517,0.007848336361348629,0.0003233155594886872,15.498376204166561,5355.736914824229
348
+ 347,86750,0.00547401762008667,0.007980580441653728,0.0003196479789919917,15.339132950175554,5371.197977529373
349
+ 348,87000,0.005289963722229004,0.008071471005678177,0.0003159915265417215,15.34243176318705,5386.666510739364
350
+ 349,87250,0.005325675010681152,0.007820555940270424,0.00031234642768508697,15.338467356283218,5402.127016474027
351
+ 350,87500,0.005225235939025879,0.007912608794867992,0.0003087129072689556,15.383866761811078,5417.63503638003
352
+ 351,87750,0.005590371608734131,0.007483434863388538,0.000305091189425982,15.26634806022048,5433.025233709253
353
+ 352,88000,0.005161337852478028,0.00787317007780075,0.0003014814975607828,15.328068062197417,5448.477207181044
354
+ 353,88250,0.005458617210388183,0.007810262031853199,0.000297884054336155,15.33421734161675,5463.935352366418
355
+ 354,88500,0.00523027229309082,0.007667079102247953,0.00029429908165934246,15.407118612900376,5479.465802732389
356
+ 355,88750,0.005201905250549316,0.00769444927573204,0.00029072680066834645,15.364633135031909,5495.057323010173
357
+ 356,89000,0.005299112319946289,0.007773960940539837,0.00028716743171828564,15.306229142006487,5510.486709992401
358
+ 357,89250,0.0052646551132202145,0.007597343996167183,0.00028362119436780285,15.381892778910697,5525.991340256296
359
+ 358,89500,0.005037670135498047,0.007695220876485109,0.0002800883073655222,15.4746751380153,5541.591075718403
360
+ 359,89750,0.005043076515197754,0.007447754964232445,0.00027656898863655526,15.395769495051354,5557.109595356975
361
+ 360,90000,0.005075533390045166,0.008333869278430939,0.00027306345526905876,15.357134114019573,5572.591282855254
362
+ 361,90250,0.005188597679138184,0.00756428437307477,0.0002695719235008432,15.193364998791367,5587.90857115807
363
+ 362,90500,0.005013197898864746,0.007487033028155565,0.00026609460870603515,15.347743289079517,5603.382353147026
364
+ 363,90750,0.005086265563964844,0.007981051690876484,0.00026263172538179005,15.267194126732647,5618.772973889019
365
+ 364,91000,0.004895675659179688,0.007413842249661684,0.0002591834871350635,15.29915297916159,5634.198316021357
366
+ 365,91250,0.005015023708343506,0.0074603320099413395,0.00025575010666943295,15.326887825969607,5649.650609885342
367
+ 366,91500,0.004812644481658935,0.007915960624814034,0.00025233179577197843,15.31116531882435,5665.085301011335
368
+ 367,91750,0.0047835888862609866,0.007594904396682978,0.00024892876530021755,15.382083588279784,5680.593488180079
369
+ 368,92000,0.004752264976501465,0.007378905545920134,0.00024554122516909957,15.2792858351022,5695.998095162213
370
+ 369,92250,0.004755453109741211,0.007386049721390009,0.00024216938433805637,15.330494510941207,5711.450620973017
371
+ 370,92500,0.004824925422668457,0.0073403967544436455,0.00023881345079811373,15.229440666269511,5726.8051065690815
372
+ 371,92750,0.004783396244049072,0.007658478803932667,0.00023547363155905962,15.302718072198331,5742.230096494313
373
+ 372,93000,0.004623723030090332,0.007382199168205261,0.0002321501326366768,15.27125365100801,5757.62297965819
374
+ 373,93250,0.004692018032073975,0.007381470408290625,0.00022884315904003435,15.271006076131016,5773.01761520002
375
+ 374,93500,0.0047329959869384765,0.007229887414723635,0.00022555291475884025,15.240451927762479,5788.381262546405
376
+ 375,93750,0.0046414794921875,0.007691189646720886,0.0002222796027508609,15.194331771228462,5803.700250360183
377
+ 376,94000,0.004674319744110107,0.0072512817569077015,0.0002190234249293989,15.262124883942306,5819.087289571296
378
+ 377,94250,0.004569844722747803,0.0071260519325733185,0.00021578458215084075,15.350871060043573,5834.563487325329
379
+ 378,94500,0.0044650864601135255,0.00726814940571785,0.0002125632742022653,15.260630887001753,5849.946197558194
380
+ 379,94750,0.004510135173797608,0.0072937821969389915,0.0002093596997891204,15.233548370655626,5865.303986911196
381
+ 380,95000,0.004436476707458496,0.007137287873774767,0.00020617405652296604,15.211069831624627,5880.642968088388
382
+ 381,95250,0.004378715515136719,0.007306450512260199,0.0002030065409092847,15.259745825082064,5896.028196646366
383
+ 382,95500,0.004391222953796387,0.007107761688530445,0.00019985734833535948,15.19924585102126,5911.3520389641635
384
+ 383,95750,0.004398441314697266,0.007264452520757914,0.00019672667305822294,15.332093956880271,5926.8088303999975
385
+ 384,96000,0.004404749870300293,0.007230805698782206,0.00019361470819267225,15.258687721099705,5942.189670897089
386
+ 385,96250,0.0043243770599365235,0.007465282455086708,0.00019052164569935954,15.313459068071097,5957.62756114034
387
+ 386,96500,0.004461232185363769,0.008062216453254223,0.0001874476763729488,15.389581345021725,5973.146039227024
388
+ 387,96750,0.00446735143661499,0.007260415703058243,0.00018439298983034778,15.368870486039668,5988.640652121976
389
+ 388,97000,0.004264039516448975,0.007164292503148317,0.00018135777449901136,15.249966412782669,6004.01530826604
390
+ 389,97250,0.004296119689941406,0.007225360721349716,0.0001783422176053182,15.254067482892424,6019.39472465124
391
+ 390,97500,0.004193171501159668,0.007219424471259117,0.00017534650516302193,15.205096972174942,6034.724632676225
392
+ 391,97750,0.004295310974121094,0.007236710749566555,0.0001723708219617776,15.211518710013479,6050.061064852402
393
+ 392,98000,0.004269704341888427,0.0072272298857569695,0.00016941535155574105,15.275817543268204,6065.466108444147
394
+ 393,98250,0.004115781307220459,0.007054343819618225,0.000166480276252249,15.3002993860282,6080.892549496144
395
+ 394,98500,0.004123061180114746,0.007193030323833227,0.00016356577710057132,15.192848757840693,6096.209408664145
396
+ 395,98750,0.004167749404907227,0.007100431248545647,0.00016067203388074434,15.178232765756547,6111.510532622226
397
+ 396,99000,0.004078288555145263,0.00711269024759531,0.00015779922509248073,15.275045458227396,6126.909237975255
398
+ 397,99250,0.004126675605773926,0.0072222487069666386,0.00015494752794415812,15.310204732231796,6142.342989180237
399
+ 398,99500,0.004087302684783935,0.007182851433753967,0.00015211711834188992,15.257597389630973,6157.724892765284
400
+ 399,99750,0.003988964319229126,0.007173783145844936,0.00014930817087867258,15.279731107875705,6173.1303216223605
401
+ 400,100000,0.0041365761756896975,0.007092091254889965,0.00014652085882361693,15.293593611102551,6188.550301825162
402
+ 401,100250,0.004090119838714599,0.007169042713940144,0.00014375535411126006,15.305624834261835,6203.981201922055
403
+ 402,100500,0.004079528331756592,0.007136574946343899,0.00014101182733095922,15.281752767041326,6219.386075025424
404
+ 403,100750,0.003936174154281616,0.0071485829539597034,0.00013829044771636922,15.342567735817283,6234.855119644199
405
+ 404,101000,0.003989144802093506,0.007059612311422825,0.000135591383135004,15.28733358392492,6250.267452205997
406
+ 405,101250,0.003960129976272583,0.007125806994736195,0.0001329148000778801,15.242396936751902,6265.720792821143
407
+ 406,101500,0.003849825382232666,0.007024920079857111,0.00013026086364924876,15.235256645362824,6281.081456401385
408
+ 407,101750,0.003880225419998169,0.007111281156539917,0.00012762973755640983,15.331246096175164,6296.534815448336
409
+ 408,102000,0.0038893587589263917,0.007071350235491991,0.00012502158409961455,15.407469442114234,6312.065370327327
410
+ 409,102250,0.0038674988746643067,0.007084503769874573,0.00012243656416205357,15.308510832022876,6327.497708279174
411
+ 410,102500,0.0038533432483673098,0.007110643200576305,0.0001198748371999332,15.374245175626129,6342.994799608365
412
+ 411,102750,0.0038170356750488283,0.007113550789654255,0.00011733656123263903,15.326038193423301,6358.443160447292
413
+ 412,103000,0.0038328659534454346,0.007035889197140932,0.00011482189283298959,15.339555751066655,6373.90684090415
414
+ 413,103250,0.0038169689178466797,0.007043600548058748,0.00011233098711757639,15.223896278068423,6389.251882668119
415
+ 414,103500,0.0037511854171752928,0.0070374528877437115,0.00010986399773719748,15.249274456873536,6404.624586740043
416
+ 415,103750,0.0037399568557739257,0.0070372032932937145,0.00010742107686737832,15.288889884017408,6420.038698968012
417
+ 416,104000,0.0037537775039672852,0.0070139262825250626,0.00010500237519898535,15.320544705260545,6435.483570113312
418
+ 417,104250,0.003681913137435913,0.006954095792025328,0.00010260804192893057,15.279565512668341,6450.8845077659935
419
+ 418,104500,0.003680272102355957,0.006984149105846882,0.00010023822475096836,15.348062830977142,6466.354483556002
420
+ 419,104750,0.0036567020416259765,0.006979385390877724,9.789306984658501e-05,15.293625906575471,6481.773294187151
421
+ 420,105000,0.003651089906692505,0.00707962503656745,9.557272187598207e-05,15.354672602377832,6497.253757372033
422
+ 421,105250,0.0036476335525512694,0.006985151208937168,9.327732396915174e-05,15.288519856054336,6512.667683555279
423
+ 422,105500,0.0036385202407836913,0.007082950323820114,9.100701771704931e-05,15.265169980935752,6528.056436972227
424
+ 423,105750,0.0036674528121948243,0.00717494823038578,8.87619431628588e-05,15.345672369003296,6543.532709156163
425
+ 424,106000,0.0036047823429107666,0.006970803719013929,8.654223879335348e-05,15.353123562876135,6559.010991731193
426
+ 425,106250,0.0036439988613128664,0.006992071866989136,8.43480415303551e-05,15.335134448949248,6574.470956468023
427
+ 426,106500,0.003583061456680298,0.006952441763132811,8.217948672228596e-05,15.406348431948572,6590.000496198423
428
+ 427,106750,0.0035775425434112548,0.0070700692012906075,8.003670813582173e-05,15.294117362238467,6605.416869543027
429
+ 428,107000,0.003559643030166626,0.006997244898229837,7.791983794763897e-05,15.432621175888926,6620.973045435268
430
+ 429,107250,0.003550440311431885,0.007052848115563393,7.582900673626216e-05,15.369667384773493,6636.469194153324
431
+ 430,107500,0.0035558085441589355,0.006972393020987511,7.376434347400921e-05,15.332338734995574,6651.927658826113
432
+ 431,107750,0.0035380637645721436,0.006953365169465542,7.172597551903549e-05,15.374806250911206,6667.424541626126
433
+ 432,108000,0.003519424915313721,0.007004976272583008,6.971402860747794e-05,15.47297346405685,6683.023914709222
434
+ 433,108250,0.003504244089126587,0.0069434489123523235,6.77286268456995e-05,15.410166506655514,6698.557307661977
435
+ 434,108500,0.003494027137756348,0.006956564728170633,6.576989270263249e-05,15.385043501853943,6714.069432265125
436
+ 435,108750,0.003458638668060303,0.006941528525203466,6.383794700222565e-05,15.339944023173302,6729.532902880106
437
+ 436,109000,0.0034623849391937256,0.006932060234248638,6.193290891598997e-05,15.432740794029087,6745.092651287094
438
+ 437,109250,0.003481933116912842,0.006929271388798952,6.005489595564807e-05,15.424642597325146,6760.643576705363
439
+ 438,109500,0.003454143762588501,0.006978271063417196,5.8204023965885666e-05,15.399090165738016,6776.168783689383
440
+ 439,109750,0.0034292802810668944,0.006933268159627914,5.638040711720537e-05,15.380347600206733,6791.673462967388
441
+ 440,110000,0.0034101643562316894,0.006976497825235128,5.458415789888437e-05,15.45775202382356,6807.257777406368
442
+ 441,110250,0.0034113285541534422,0.006938081234693527,5.281538711203586e-05,15.373513997998089,6822.75720337918
443
+ 442,110500,0.0034140992164611817,0.006960998754948378,5.107420386277322e-05,15.421954756136984,6838.304720183369
444
+ 443,110750,0.0034037120342254638,0.006948521826416254,4.9360715555481175e-05,15.505776289850473,6853.938751067035
445
+ 444,111000,0.003383695125579834,0.0069228108040988445,4.767502788618982e-05,15.422310635913163,6869.484382073395
446
+ 445,111250,0.0033709201812744142,0.006932620890438557,4.6017244836054795e-05,15.33273027278483,6884.940096853301
447
+ 446,111500,0.0033601362705230715,0.006942379288375378,4.438746866494362e-05,15.387898024637252,6900.456907832995
448
+ 447,111750,0.0033647541999816894,0.0069233737885952,4.278579990512717e-05,15.417491804808378,6915.999641462229
449
+ 448,112000,0.0033512060642242433,0.006944562308490276,4.1212337355079355e-05,15.384509574621916,6931.507896856405
450
+ 449,112250,0.0033446311950683593,0.00696179224178195,3.9667178073381844e-05,15.324716694653034,6946.9580413582735
451
+ 450,112500,0.0033409602642059328,0.006953740492463112,3.8150417372737465e-05,15.361157259903848,6962.441877358127
452
+ 451,112750,0.0033257577419281005,0.006928788498044014,3.666214881409082e-05,15.336069305893034,6977.90157678118
453
+ 452,113000,0.0033258652687072755,0.006962503306567669,3.520246420085705e-05,15.274141512811184,6993.300778404344
454
+ 453,113250,0.0033145179748535155,0.006937073078006506,3.377145357325864e-05,15.353424463886768,7008.779683215078
455
+ 454,113500,0.003311082601547241,0.006949728354811668,3.236920520277209e-05,15.398238708730787,7024.302596457303
456
+ 455,113750,0.003299025535583496,0.0069467551074922085,3.0995805586681636e-05,15.409362277016044,7039.8384097381495
457
+ 456,114000,0.0032985515594482423,0.006941130850464106,2.9651339442745265e-05,15.378603682853281,7055.342275264207
458
+ 457,114250,0.003288684606552124,0.006937453523278236,2.8335889703967613e-05,15.440451947040856,7070.907414967194
459
+ 458,114500,0.0032871835231781006,0.006931053474545479,2.704953751348483e-05,15.341346244327724,7086.457073679194
460
+ 459,114750,0.0032804808616638185,0.006918028462678194,2.5792362219559353e-05,15.45176373841241,7102.037266775966
461
+ 460,115000,0.0032719173431396483,0.006940564140677452,2.456444137068509e-05,15.35146452812478,7117.525448003318
462
+ 461,115250,0.0032676610946655274,0.0069299256429076195,2.3365850710803953e-05,15.436450499109924,7133.087822506204
463
+ 462,115500,0.0032637984752655028,0.00695067597553134,2.219666417463399e-05,15.466396963689476,7148.678862603381
464
+ 463,115750,0.003261479139328003,0.006932545453310013,2.1056953883107738e-05,15.47854893002659,7164.2827822370455
465
+ 464,116000,0.0032548112869262695,0.006933399476110935,1.9946790138924756e-05,15.451312540099025,7179.85927533824
466
+ 465,116250,0.0032525532245635987,0.006934390868991613,1.886624142221405e-05,15.447840064298362,7195.434051913209
467
+ 466,116500,0.0032485139369964598,0.006923853885382414,1.7815374386310305e-05,15.424379284027964,7210.9879854712635
468
+ 467,116750,0.003242554187774658,0.006934405770152807,1.6794253853642316e-05,15.444013005122542,7226.575940073002
469
+ 468,117000,0.003238537549972534,0.006942023057490587,1.5802942811734407e-05,15.397456561215222,7242.097500276286
470
+ 469,117250,0.003233228921890259,0.0069348253309726715,1.4841502409321174e-05,15.322341090999544,7257.551169830374
471
+ 470,117500,0.003232154607772827,0.006935408338904381,1.3909991952575461e-05,15.35689299274236,7273.0341360191815
472
+ 471,117750,0.003227559804916382,0.006938273087143898,1.3008468901450011e-05,15.412803226150572,7288.577048929408
473
+ 472,118000,0.003223644971847534,0.006939132232218981,1.2136988866133088e-05,15.35723222931847,7304.061702937353
474
+ 473,118250,0.0032220606803894042,0.006938036065548658,1.129560560361851e-05,15.29429430840537,7319.4846314443275
475
+ 474,118500,0.003216599464416504,0.006931551732122898,1.048437101438897e-05,15.452050909865648,7335.061421848368
476
+ 475,118750,0.0032150635719299316,0.006931945215910673,9.70333513921536e-06,15.271277381107211,7350.4622329543345
477
+ 476,119000,0.003213024139404297,0.006938274018466473,8.952546156069267e-06,15.259620094671845,7365.845194449183
478
+ 477,119250,0.0032107532024383545,0.00693887285888195,8.23205037715188e-06,15.244434128981084,7381.215168745257
479
+ 478,119500,0.003208783149719238,0.0069336118176579475,7.5418922460366585e-06,15.35864598210901,7396.70063288603
480
+ 479,119750,0.00320588755607605,0.006942310836166143,6.882114334927984e-06,15.33688904484734,7412.163607998285
481
+ 480,120000,0.003204388380050659,0.0069364141672849655,6.252757342035284e-06,15.26775070792064,7427.553969234228
482
+ 481,120250,0.003202155351638794,0.006940888240933418,5.653860089062413e-06,15.305882797576487,7442.98469470907
483
+ 482,120500,0.0031999690532684326,0.006935384124517441,5.085459518812971e-06,15.321235472802073,7458.432031119242
484
+ 483,120750,0.0031989777088165285,0.006934095174074173,4.547590692911631e-06,15.268839037977159,7473.827275630087
485
+ 484,121000,0.0031970391273498534,0.0069369180127978325,4.040286789641035e-06,15.214677304029465,7489.166890552267
486
+ 485,121250,0.0031950371265411378,0.0069320350885391235,3.56357910189563e-06,15.238867842126638,7504.532315570395
487
+ 486,121500,0.0031941707134246827,0.006937968544661999,3.1174970352511306e-06,15.366881002206355,7520.022151857149
488
+ 487,121750,0.003192492723464966,0.006935844197869301,2.7020681061505403e-06,15.375918770674616,7535.523176422343
489
+ 488,122000,0.0031922211647033693,0.006938366685062647,2.3173179402070666e-06,15.376280087046325,7551.033684147056
490
+ 489,122250,0.00319244647026062,0.0069391424767673016,1.9632702706233246e-06,15.386696572881192,7566.546210127417
491
+ 490,122500,0.0031903953552246092,0.006936934776604176,1.6399469367272608e-06,15.322512584272772,7581.992278079037
492
+ 491,122750,0.0031902809143066406,0.006937052123248577,1.3473678826251473e-06,15.203828105702996,7597.3241410031915
493
+ 492,123000,0.00318953013420105,0.006936695426702499,1.085551155971188e-06,15.204905193764716,7612.654544972349
494
+ 493,123250,0.0031870970726013185,0.00693588238209486,8.545129068542978e-07,15.224784992169589,7628.005436769221
495
+ 494,123500,0.0031874148845672605,0.0069380225613713264,6.542673868019994e-07,15.319725315086544,7643.453755962197
496
+ 495,123750,0.0031860928535461424,0.006936773657798767,4.848269479012145e-07,15.29802629398182,7658.878025745973
497
+ 496,124000,0.0031861863136291506,0.006937854457646608,3.4620204203639405e-07,15.14297747425735,7674.148101664148
498
+ 497,124250,0.0031877269744873048,0.0069373976439237595,2.3840122024459892e-07,15.145579149015248,7689.42139551742
499
+ 498,124500,0.003185642957687378,0.006937212776392698,1.6143113218847369e-07,15.263297004625201,7704.807868059259
500
+ 499,124750,0.0031855576038360596,0.006936921272426844,1.1529652574545007e-07,15.281335772946477,7720.2274958849885
501
+ 500,125000,0.003185694456100464,0.0069372509606182575,1.0000024671545432e-07,15.271698139142245,7735.62425503321
results/learning_curves.png ADDED

Git LFS Details

  • SHA256: a0221f829db21c05ebde6e152e8182fa8518d2424546a917d7c44eb311fb3e12
  • Pointer size: 131 Bytes
  • Size of remote file: 106 kB
results/metrics.json ADDED
@@ -0,0 +1,78 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "mean_relative_l2_raw": 0.006937251414878938,
3
+ "test_samples": 200,
4
+ "actual_training_epochs": 500,
5
+ "optimizer_updates": 125000,
6
+ "paper_table3": 0.007,
7
+ "signed_gap": -6.274858512106234e-05,
8
+ "absolute_gap": 6.274858512106234e-05,
9
+ "relative_gap_percent": -0.8964083588723137,
10
+ "selection": "final_epoch_500_no_test_selection",
11
+ "checkpoint": "/public/home/liuyushuang/zhaomingsi/onescience-cfd-repro-packages/MoNo-light/weight/best_model.pth",
12
+ "checkpoint_sha256": "d20393f9ae89d38f53b8da9e09799db57bef37154ec8c7fddb9a9dd4a6666122",
13
+ "min": 0.0023765875482458245,
14
+ "max": 0.02536537307374934,
15
+ "median": 0.006544329873232485,
16
+ "training_environment": {
17
+ "python": "3.10.16 (main, Dec 11 2024, 16:24:50) [GCC 11.2.0]",
18
+ "torch": "2.4.1",
19
+ "hip": "6.1.25065",
20
+ "cuda": null,
21
+ "device": "BW",
22
+ "device_count": 1,
23
+ "platform": "Linux-5.10.134-17.1.3.sga8.x86_64-x86_64-with-glibc2.28",
24
+ "hostname": "f11r3n13",
25
+ "slurm_job_id": "1100920",
26
+ "trainable_parameters": 1696705,
27
+ "all_parameters": 1734241,
28
+ "precision": "float32",
29
+ "code_hashes": {
30
+ "model/MoNo.py": "7af09c5106dbadc7bac35100c2c852c9739b1fa86a4ab52cc5fe12955b095e72",
31
+ "config/config.yaml": "c333c872f00ea12eb9bb8c61b85df5614b18e17e49cfb76d8f19e7c038657517",
32
+ "scripts/train.py": "a1bee1c36f6361c546b4a9ff166681e99baab2f19842ff480f033449673db958",
33
+ "scripts/common.py": "f4283c26f3bbb3747a6d4a632ac0f499e066c8c411bd5eb73775f0cc01c6e1a9"
34
+ }
35
+ },
36
+ "inference_environment": {
37
+ "python": "3.10.16 (main, Dec 11 2024, 16:24:50) [GCC 11.2.0]",
38
+ "torch": "2.4.1",
39
+ "hip": "6.1.25065",
40
+ "cuda": null,
41
+ "device": "BW",
42
+ "device_count": 1,
43
+ "platform": "Linux-5.10.134-17.1.3.sga8.x86_64-x86_64-with-glibc2.28",
44
+ "hostname": "f11r3n13",
45
+ "slurm_job_id": "1100920"
46
+ },
47
+ "software_versions": {
48
+ "torch": "2.4.1+das.opt2.dtk2504",
49
+ "triton": "3.0.0+das.opt4.dtk2504",
50
+ "onescience": "0.1.2rc1",
51
+ "numpy": "1.26.4",
52
+ "scipy": "1.15.2",
53
+ "PyYAML": "6.0.2",
54
+ "matplotlib": "3.10.1"
55
+ },
56
+ "configured_cluster": {
57
+ "partition": "hx1hdnormal01",
58
+ "nodes": 1,
59
+ "gpus_per_node": 1,
60
+ "cpus_per_task": 8,
61
+ "memory": "30G",
62
+ "time_limit": "24:00:00",
63
+ "gpu_type": "dcu",
64
+ "ntasks_per_node": 1
65
+ },
66
+ "gpu_total_memory_bytes": 68702699520,
67
+ "limitations": [
68
+ "Single DCU/PyTorch 2.4.1 HIP backend versus paper RTX4090/PyTorch2.8 CUDA12.8.",
69
+ "Configured system RAM allocation is 30G, versus up to 600GB reported by the paper; all 1000/200 samples and model dimensions retained.",
70
+ "Installed Triton is 3.0.0 with DTK customizations, versus paper Triton3.4.0; author fused kernels are not used.",
71
+ "Independent implementation, packed QKV/SDPA attention, and RNG consumption differ from explicit author attention; not bitwise reproduction.",
72
+ "Portable PyTorch log-domain Sinkhorn instead of author fused Triton kernels; same finite-step equations and differentiation.",
73
+ "Identical Darcy geometry is computed once and broadcast over batch; arithmetic ordering differs.",
74
+ "Paper does not specify seed uncertainty or a tolerance for the Table3 result.",
75
+ "Final epoch checkpoint named best_model.pth; no test-set model selection.",
76
+ "Units unspecified by supplied data; metric uses stored pressure scale."
77
+ ]
78
+ }
results/per_sample_relative_l2.csv ADDED
@@ -0,0 +1,201 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ sample_id,relative_l2_raw
2
+ 0,0.0138126004386
3
+ 1,0.00999366759716
4
+ 2,0.00505750721051
5
+ 3,0.00410922791988
6
+ 4,0.0130247544897
7
+ 5,0.00913412627994
8
+ 6,0.00760866136037
9
+ 7,0.00437685625986
10
+ 8,0.0128817129573
11
+ 9,0.0120136877571
12
+ 10,0.00399453370847
13
+ 11,0.00779560412112
14
+ 12,0.00445218542823
15
+ 13,0.00376998345676
16
+ 14,0.00476850563405
17
+ 15,0.0126613582054
18
+ 16,0.00494123498385
19
+ 17,0.00452646374066
20
+ 18,0.00796886183484
21
+ 19,0.00269837198393
22
+ 20,0.00579623001242
23
+ 21,0.00470082973048
24
+ 22,0.00579522594628
25
+ 23,0.0110667318381
26
+ 24,0.00711877159268
27
+ 25,0.0110491854063
28
+ 26,0.0107557280705
29
+ 27,0.00905286977391
30
+ 28,0.00912635680864
31
+ 29,0.00681994263552
32
+ 30,0.00653607525011
33
+ 31,0.00683860880962
34
+ 32,0.00438352472066
35
+ 33,0.00735510507046
36
+ 34,0.00710651986291
37
+ 35,0.00991253323097
38
+ 36,0.00970544530158
39
+ 37,0.00711800749442
40
+ 38,0.00745989700356
41
+ 39,0.00732189859014
42
+ 40,0.0033056769119
43
+ 41,0.0068161096994
44
+ 42,0.00562316619202
45
+ 43,0.00420052865681
46
+ 44,0.00273063171145
47
+ 45,0.00459957192056
48
+ 46,0.00960451297734
49
+ 47,0.00522597863039
50
+ 48,0.0105661005488
51
+ 49,0.00678832942927
52
+ 50,0.00756292752274
53
+ 51,0.00791935812366
54
+ 52,0.00765738316387
55
+ 53,0.00935033058748
56
+ 54,0.00545267895447
57
+ 55,0.00477986105489
58
+ 56,0.00548518637493
59
+ 57,0.00369341164178
60
+ 58,0.0052932782858
61
+ 59,0.00499823038384
62
+ 60,0.00471546114126
63
+ 61,0.0118964490131
64
+ 62,0.00413864450502
65
+ 63,0.00884017035999
66
+ 64,0.0130707417001
67
+ 65,0.00460057538493
68
+ 66,0.00423870221818
69
+ 67,0.00898534502389
70
+ 68,0.00535272836915
71
+ 69,0.0040423192719
72
+ 70,0.0102750085854
73
+ 71,0.00660033266588
74
+ 72,0.00530584425223
75
+ 73,0.00338575157397
76
+ 74,0.00470732375604
77
+ 75,0.012045211099
78
+ 76,0.00663286640473
79
+ 77,0.0086903874746
80
+ 78,0.00237658754825
81
+ 79,0.00427263407216
82
+ 80,0.0047036685111
83
+ 81,0.00719275640426
84
+ 82,0.0108054033784
85
+ 83,0.00646236146682
86
+ 84,0.00791578539222
87
+ 85,0.00630664008509
88
+ 86,0.00727419311007
89
+ 87,0.00668073911972
90
+ 88,0.00450902583386
91
+ 89,0.00697214066133
92
+ 90,0.0085116215188
93
+ 91,0.0081286999344
94
+ 92,0.0095798049044
95
+ 93,0.00624657235084
96
+ 94,0.00581032865025
97
+ 95,0.00710803075984
98
+ 96,0.00328809007895
99
+ 97,0.00527473236437
100
+ 98,0.0057717571638
101
+ 99,0.0070211376729
102
+ 100,0.00804533210246
103
+ 101,0.00420762231678
104
+ 102,0.00515386772523
105
+ 103,0.0083418780515
106
+ 104,0.0143090385711
107
+ 105,0.00451963471072
108
+ 106,0.00876046476414
109
+ 107,0.00290578254421
110
+ 108,0.00860362780743
111
+ 109,0.00574500555489
112
+ 110,0.00822511138427
113
+ 111,0.0085787954264
114
+ 112,0.0101081419238
115
+ 113,0.00538206839139
116
+ 114,0.00842403210505
117
+ 115,0.00466886653202
118
+ 116,0.00958678963909
119
+ 117,0.00501527773609
120
+ 118,0.00315401011517
121
+ 119,0.00800852759314
122
+ 120,0.00796792318761
123
+ 121,0.0075437280501
124
+ 122,0.00414416304355
125
+ 123,0.00859304888142
126
+ 124,0.00419853782688
127
+ 125,0.0123323977651
128
+ 126,0.00467593482361
129
+ 127,0.00712798693778
130
+ 128,0.00465241474612
131
+ 129,0.00548711186535
132
+ 130,0.00514275276789
133
+ 131,0.00556534199479
134
+ 132,0.00476738503191
135
+ 133,0.00461674234551
136
+ 134,0.00786071295896
137
+ 135,0.00759764191386
138
+ 136,0.00640229054407
139
+ 137,0.00641014924493
140
+ 138,0.00561892889404
141
+ 139,0.0053896152341
142
+ 140,0.0120917270721
143
+ 141,0.00958822214544
144
+ 142,0.0079730892816
145
+ 143,0.00393242513282
146
+ 144,0.00481926326536
147
+ 145,0.00645239387442
148
+ 146,0.00819815320625
149
+ 147,0.00679767296453
150
+ 148,0.00501972557339
151
+ 149,0.00316405758991
152
+ 150,0.00625714067346
153
+ 151,0.00616670237343
154
+ 152,0.00340059777185
155
+ 153,0.00522902775675
156
+ 154,0.00687998357061
157
+ 155,0.00538438266452
158
+ 156,0.00495841745713
159
+ 157,0.00478002596706
160
+ 158,0.0068462712261
161
+ 159,0.00859655223154
162
+ 160,0.00521924424019
163
+ 161,0.00535899145496
164
+ 162,0.00625913092382
165
+ 163,0.0112882481225
166
+ 164,0.00353276975213
167
+ 165,0.00420763919618
168
+ 166,0.00688317811815
169
+ 167,0.00837558210976
170
+ 168,0.00774839384157
171
+ 169,0.00759722713067
172
+ 170,0.00559909139203
173
+ 171,0.00966010313238
174
+ 172,0.0090295028961
175
+ 173,0.00854307786637
176
+ 174,0.00936618172171
177
+ 175,0.0253653730737
178
+ 176,0.0106234262684
179
+ 177,0.00847254114318
180
+ 178,0.00553991112762
181
+ 179,0.00693319662564
182
+ 180,0.00529715718671
183
+ 181,0.00633835373923
184
+ 182,0.00804425391565
185
+ 183,0.0101535619267
186
+ 184,0.00373579121527
187
+ 185,0.00717421394171
188
+ 186,0.00371881573351
189
+ 187,0.00805076449847
190
+ 188,0.00255481551767
191
+ 189,0.00604306314976
192
+ 190,0.0131715785207
193
+ 191,0.00655258449636
194
+ 192,0.00367518978482
195
+ 193,0.00497217333567
196
+ 194,0.0048732119879
197
+ 195,0.00599320304725
198
+ 196,0.00453774362019
199
+ 197,0.00413766867939
200
+ 198,0.010604448215
201
+ 199,0.0118451464067
results/predictions.npz ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:db9b1e54fc20429a9059c02a622803efc01b65ace323195f9be86d10330380ce
3
+ size 17402518
results/pressure_fields.png ADDED

Git LFS Details

  • SHA256: d8752fb6e9fd17b28fdd19c3f48ba6c96c0b51ae8fac4646036bfdd3e2f6d94a
  • Pointer size: 131 Bytes
  • Size of remote file: 488 kB
results/reproduction_report.md ADDED
@@ -0,0 +1,44 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # MoNo-light Darcy 85×85 reproduction
2
+
3
+ Paper: https://arxiv.org/pdf/2608.09764v1, Table 3 and Appendix F.
4
+
5
+ ## Measured result
6
+
7
+ - Raw-scale per-sample relative L2 mean over all 200 test samples: **0.0069372514**.
8
+ - Paper Table 3: 0.0070; signed difference: **-0.0000627486**; absolute difference: 0.0000627486; relative difference: -0.8964%.
9
+ - Actual completed training: **500 epochs / 125000 optimizer updates**, batch size 4, max LR .001, 100 warmup epochs.
10
+ - Training elapsed time: 2.1488 hours.
11
+ - Weight choice: **final epoch 500**, saved as `weight/best_model.pth`. Test metrics never select or tune weights. No validation split or early stopping.
12
+ - Checkpoint SHA256: `d20393f9ae89d38f53b8da9e09799db57bef37154ec8c7fddb9a9dd4a6666122`.
13
+ - Parameters: 1,696,705 trainable; 1,734,241 including the frozen anchor copy (about 1.73M total).
14
+ - Actual package versions: `{'torch': '2.4.1+das.opt2.dtk2504', 'triton': '3.0.0+das.opt4.dtk2504', 'onescience': '0.1.2rc1', 'numpy': '1.26.4', 'scipy': '1.15.2', 'PyYAML': '6.0.2', 'matplotlib': '3.10.1'}`. Slurm resources: `{'partition': 'hx1hdnormal01', 'nodes': 1, 'gpus_per_node': 1, 'cpus_per_task': 8, 'memory': '30G', 'time_limit': '24:00:00', 'gpu_type': 'dcu', 'ntasks_per_node': 1}`.
15
+
16
+ ## Data and metric
17
+
18
+ Reuse supplied `piececonst_r421_N1024_smooth1.mat[0:1000]` and `piececonst_r421_N1024_smooth2.mat[0:200]`, fields coeff/sol. Slice `[:,::5,::5]` to 85×85. Global per-channel unbiased train statistics only; reuse for test. Normalize input coordinates/coeff and pressure; inverse-transform pressure before loss and evaluation. Metric is arithmetic mean of each sample's norm(pred-target)/norm(target), recomputed independently with float64 NumPy after checkpoint reload.
19
+
20
+ ## Alignment and limitations
21
+
22
+ - Single DCU/PyTorch 2.4.1 HIP backend versus paper RTX4090/PyTorch2.8 CUDA12.8.
23
+ - Configured system RAM allocation is 30G, versus up to 600GB reported by the paper; all 1000/200 samples and model dimensions retained.
24
+ - Installed Triton is 3.0.0 with DTK customizations, versus paper Triton3.4.0; author fused kernels are not used.
25
+ - Independent implementation, packed QKV/SDPA attention, and RNG consumption differ from explicit author attention; not bitwise reproduction.
26
+ - Portable PyTorch log-domain Sinkhorn instead of author fused Triton kernels; same finite-step equations and differentiation.
27
+ - Identical Darcy geometry is computed once and broadcast over batch; arithmetic ordering differs.
28
+ - Paper does not specify seed uncertainty or a tolerance for the Table3 result.
29
+ - Final epoch checkpoint named best_model.pth; no test-set model selection.
30
+ - Units unspecified by supplied data; metric uses stored pressure scale.
31
+
32
+ Implementation sources and paper-versus-supplement distinction are preserved in `.trainer_work/darcy85/source_capture.md`. Official code was consulted read-only for paper-unstated details, including a resume audit; no repository downloaded, cloned, copied, or imported. This session reused the saved paper and training specifications.
33
+
34
+ ## Artifacts
35
+
36
+ - `predictions.npz`: all 200 raw-scale predictions, targets, permeability, coordinates and errors.
37
+ - `per_sample_relative_l2.csv`, `metrics.json`, `history.csv`, `data_manifest.json`, `training_environment.json`.
38
+ - `pressure_fields.png`: fixed test samples 0--3, common pressure color scale within each sample.
39
+ - `test_error_distribution.png`, `learning_curves.png`.
40
+ - Runtime stdout/stderr and Slurm evidence: `../logs/` and `../.onescience/`.
41
+
42
+ ## Execution provenance
43
+
44
+ The prior interrupted session had completed specifications and handoffs but left the model, script, and weight directories empty. This session reused those documents and the original data. Slurm smoke job1100898 passed all six paths. Job1100916 failed at startup before training output or checkpoint creation. The same full-budget experiment continued as job1100920 after excluding that startup-fault node; no formal epochs were discarded or restarted from an existing trained model. See `.onescience/task_state_before_resume.json` and runtime events for the audit trail.
results/smoke_checks.json ADDED
@@ -0,0 +1,33 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "status": "passed",
3
+ "attempt": 1,
4
+ "checks": {
5
+ "forward_pass": "passed",
6
+ "backward_pass": "passed",
7
+ "train_loop_dry_run": "passed: one isolated epoch on 8 real samples; formal budget unchanged",
8
+ "validation_loop_dry_run": "passed",
9
+ "domain_metric_dry_run": "passed: raw relative L2 vs NumPy; OT marginal and finite-step gradcheck",
10
+ "config_consistency": "passed"
11
+ },
12
+ "environment": {
13
+ "python": "3.10.16 (main, Dec 11 2024, 16:24:50) [GCC 11.2.0]",
14
+ "torch": "2.4.1",
15
+ "hip": "6.1.25065",
16
+ "cuda": null,
17
+ "device": "BW",
18
+ "device_count": 1,
19
+ "platform": "Linux-5.10.134-17.1.3.sga8.x86_64-x86_64-with-glibc2.28",
20
+ "hostname": "f11r3n09",
21
+ "slurm_job_id": "1100898"
22
+ },
23
+ "seconds_per_training_step": 0.09329674200853333,
24
+ "estimated_formal_hours": 3.2394702086296294,
25
+ "trainable_parameters": 1696705,
26
+ "losses": [
27
+ 0.5791354179382324,
28
+ 0.5755595564842224
29
+ ],
30
+ "smoke_samples": 8,
31
+ "formal_samples": 1000,
32
+ "formal_epochs": 500
33
+ }
results/test_error_distribution.png ADDED
results/training_environment.json ADDED
@@ -0,0 +1,20 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "python": "3.10.16 (main, Dec 11 2024, 16:24:50) [GCC 11.2.0]",
3
+ "torch": "2.4.1",
4
+ "hip": "6.1.25065",
5
+ "cuda": null,
6
+ "device": "BW",
7
+ "device_count": 1,
8
+ "platform": "Linux-5.10.134-17.1.3.sga8.x86_64-x86_64-with-glibc2.28",
9
+ "hostname": "f11r3n13",
10
+ "slurm_job_id": "1100920",
11
+ "trainable_parameters": 1696705,
12
+ "all_parameters": 1734241,
13
+ "precision": "float32",
14
+ "code_hashes": {
15
+ "model/MoNo.py": "7af09c5106dbadc7bac35100c2c852c9739b1fa86a4ab52cc5fe12955b095e72",
16
+ "config/config.yaml": "c333c872f00ea12eb9bb8c61b85df5614b18e17e49cfb76d8f19e7c038657517",
17
+ "scripts/train.py": "a1bee1c36f6361c546b4a9ff166681e99baab2f19842ff480f033449673db958",
18
+ "scripts/common.py": "f4283c26f3bbb3747a6d4a632ac0f499e066c8c411bd5eb73775f0cc01c6e1a9"
19
+ }
20
+ }
results/training_status.json ADDED
@@ -0,0 +1,18 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "status": "completed",
3
+ "completed_epochs": 500,
4
+ "formal_epochs": 500,
5
+ "global_step": 125000,
6
+ "formal_updates": 125000,
7
+ "checkpoint": "/public/home/liuyushuang/zhaomingsi/onescience-cfd-repro-packages/MoNo-light/weight/last_model.pth",
8
+ "last_epoch": {
9
+ "epoch": 500,
10
+ "updates": 125000,
11
+ "train_relative_l2": 0.003185694456100464,
12
+ "test_relative_l2": 0.0069372509606182575,
13
+ "learning_rate": 1.0000024671545432e-07,
14
+ "epoch_seconds": 15.271698139142245,
15
+ "elapsed_seconds": 7735.62425503321
16
+ },
17
+ "slurm_job_id": "1100920"
18
+ }
results/visualization_quality_checks.json ADDED
@@ -0,0 +1,8 @@
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "status": "passed",
3
+ "pressure_fields": "Reviewed actual permeability/ground truth/prediction/error panels; fixed tight bounding box to retain colorbar labels",
4
+ "learning_curves": "Reviewed actual 500-epoch loss and 100-epoch warmup/OneCycle curve",
5
+ "all_images_nonempty": true,
6
+ "render_only_rerun": "logs/visualization_margin_fix.log",
7
+ "training_and_metrics_unchanged": true
8
+ }
scripts/common.py ADDED
@@ -0,0 +1,167 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ """Data and evaluation contracts shared by the independently written runners."""
2
+ import argparse
3
+ import hashlib
4
+ import json
5
+ import os
6
+ from pathlib import Path
7
+ import platform
8
+ import random
9
+ import sys
10
+
11
+ import numpy as np
12
+ from scipy.io import loadmat, whosmat
13
+ import torch
14
+ import yaml
15
+
16
+ ROOT = Path(__file__).resolve().parents[1]
17
+ sys.path.insert(0, str(ROOT))
18
+ from model.MoNo import MoNoLight
19
+
20
+
21
+ def config_args(description):
22
+ parser = argparse.ArgumentParser(description=description)
23
+ parser.add_argument('--config', default=str(ROOT / 'config/config.yaml'))
24
+ return parser
25
+
26
+
27
+ def read_config(path):
28
+ with open(path) as f:
29
+ cfg = yaml.safe_load(f)
30
+ assert (cfg['data']['train_samples'], cfg['data']['test_samples']) == (1000, 200)
31
+ assert cfg['data']['stride'] == 5 and cfg['data']['grid_size'] == 85
32
+ t = cfg['training']
33
+ assert (t['epochs'], t['batch_size'], t['warmup_epochs'], t['learning_rate']) == (500, 4, 100, .001)
34
+ return cfg
35
+
36
+
37
+ def path(value):
38
+ p = Path(value)
39
+ return p if p.is_absolute() else ROOT / p
40
+
41
+
42
+ def atomic_json(destination, obj):
43
+ destination = Path(destination)
44
+ destination.parent.mkdir(parents=True, exist_ok=True)
45
+ tmp = destination.with_suffix(destination.suffix + '.tmp')
46
+ tmp.write_text(json.dumps(obj, indent=2, ensure_ascii=False, allow_nan=False))
47
+ os.replace(tmp, destination)
48
+
49
+
50
+ def atomic_save(destination, obj):
51
+ destination = Path(destination)
52
+ destination.parent.mkdir(parents=True, exist_ok=True)
53
+ tmp = destination.with_suffix(destination.suffix + '.tmp')
54
+ torch.save(obj, tmp)
55
+ os.replace(tmp, destination)
56
+
57
+
58
+ def seed_all(seed):
59
+ random.seed(seed)
60
+ np.random.seed(seed)
61
+ torch.manual_seed(seed)
62
+ torch.cuda.manual_seed_all(seed)
63
+
64
+
65
+ def require_compute_node():
66
+ assert torch.cuda.is_available(), 'Assigned GPU/DCU unavailable; no CPU fallback'
67
+ torch.set_num_threads(int(os.environ.get('OMP_NUM_THREADS', '8')))
68
+ return torch.device('cuda')
69
+
70
+
71
+ def environment():
72
+ return dict(python=sys.version, torch=torch.__version__, hip=torch.version.hip,
73
+ cuda=torch.version.cuda, device=torch.cuda.get_device_name(),
74
+ device_count=torch.cuda.device_count(), platform=platform.platform(),
75
+ hostname=platform.node(), slurm_job_id=os.environ.get('SLURM_JOB_ID'))
76
+
77
+
78
+ def source_signature(cfg):
79
+ d = cfg['data']
80
+ return [dict(file=str(Path(d['directory']) / d[k]),
81
+ size=(Path(d['directory']) / d[k]).stat().st_size,
82
+ mtime_ns=(Path(d['directory']) / d[k]).stat().st_mtime_ns)
83
+ for k in ['train_file', 'test_file']]
84
+
85
+
86
+ def load_data(cfg):
87
+ d = cfg['data']
88
+ cache = path(d['cache'])
89
+ signature = source_signature(cfg)
90
+ if cache.exists():
91
+ with np.load(cache) as f:
92
+ assert json.loads(str(f['signature'])) == signature, 'Data source changed'
93
+ arrays = {k: f[k].copy() for k in ['train_coeff', 'train_sol', 'test_coeff', 'test_sol']}
94
+ else:
95
+ arrays = {}
96
+ schema = {}
97
+ for split, count in [('train', 1000), ('test', 200)]:
98
+ src = Path(d['directory']) / d[split + '_file']
99
+ schema[split] = whosmat(src)
100
+ print(f'DATA reading {src}, first {count} samples; stride=5', flush=True)
101
+ mat = loadmat(src, variable_names=['coeff', 'sol'])
102
+ for field in ['coeff', 'sol']:
103
+ assert mat[field].shape == (1024, 421, 421), (src, field, mat[field].shape)
104
+ a = np.ascontiguousarray(mat[field][:count, ::5, ::5], dtype=np.float32)
105
+ assert a.shape == (count, 85, 85) and np.isfinite(a).all()
106
+ arrays[split + '_' + field] = a
107
+ del mat
108
+ cache.parent.mkdir(parents=True, exist_ok=True)
109
+ tmp = cache.with_suffix('.tmp.npz')
110
+ np.savez(tmp, **arrays, signature=json.dumps(signature))
111
+ os.replace(tmp, cache)
112
+ atomic_json(path(cfg['results']) / 'data_manifest.json', dict(
113
+ sources=signature, schema=schema, train_samples=1000, test_samples=200,
114
+ split='file1[0:1000], file2[0:200]', downsample='[:,::5,::5]',
115
+ grid=[85,85], coeff='permeability', sol='pressure', units='stored numerical scale; unspecified'))
116
+ lin = np.linspace(0, 1, 85, dtype=np.float32)
117
+ grid = np.stack(np.meshgrid(lin, lin, indexing='xy'), -1).reshape(-1, 2)
118
+ coordinates = torch.from_numpy(grid).unsqueeze(0)
119
+ out = {'coordinates_raw': coordinates}
120
+ for split in ['train', 'test']:
121
+ coeff = torch.from_numpy(arrays[split + '_coeff']).reshape(-1, 7225, 1)
122
+ sol = torch.from_numpy(arrays[split + '_sol']).reshape(-1, 7225, 1)
123
+ out[split + '_coeff'] = coeff
124
+ out[split + '_sol'] = sol
125
+ out[split + '_features'] = torch.cat([coordinates.expand(coeff.shape[0], -1, -1), coeff], -1)
126
+ return out
127
+
128
+
129
+ def compute_stats(data):
130
+ x, y = data['train_features'], data['train_sol']
131
+ return dict(x_mean=x.mean((0,1), keepdim=True),
132
+ x_std=x.std((0,1), keepdim=True) + 1e-8,
133
+ y_mean=y.mean((0,1), keepdim=True),
134
+ y_std=y.std((0,1), keepdim=True) + 1e-8)
135
+
136
+
137
+ def normalized(data, stats, device):
138
+ s = {k: v.to(device) for k, v in stats.items()}
139
+ out = {'coordinates': ((data['coordinates_raw'].to(device) - s['x_mean'][..., :2]) /
140
+ s['x_std'][..., :2])}
141
+ for split in ['train', 'test']:
142
+ out[split + '_x'] = (data[split + '_features'].to(device) - s['x_mean']) / s['x_std']
143
+ out[split + '_y'] = data[split + '_sol'].to(device)
144
+ return out, s
145
+
146
+
147
+ def relative_l2(prediction, target):
148
+ return (prediction - target).flatten(1).norm(dim=1) / target.flatten(1).norm(dim=1)
149
+
150
+
151
+ @torch.no_grad()
152
+ def evaluate(model, data, stats, batch_size=4, return_predictions=False):
153
+ model.eval()
154
+ errors, predictions = [], []
155
+ for i in range(0, data['test_x'].shape[0], batch_size):
156
+ y = model(data['coordinates'], data['test_x'][i:i+batch_size])
157
+ y = y * stats['y_std'] + stats['y_mean']
158
+ errors.append(relative_l2(y, data['test_y'][i:i+batch_size]).cpu())
159
+ if return_predictions:
160
+ predictions.append(y.cpu())
161
+ return torch.cat(errors), (torch.cat(predictions) if return_predictions else None)
162
+
163
+
164
+ def code_hashes():
165
+ return {str(f.relative_to(ROOT)): hashlib.sha256(f.read_bytes()).hexdigest()
166
+ for f in [ROOT/'model/MoNo.py', ROOT/'config/config.yaml',
167
+ ROOT/'scripts/train.py', ROOT/'scripts/common.py']}
scripts/finalize.py ADDED
@@ -0,0 +1,66 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ """Audit completed artifacts and update OneScience state from real execution evidence."""
2
+ import json
3
+ import hashlib
4
+ import os
5
+ from pathlib import Path
6
+
7
+ import numpy as np
8
+
9
+ from common import ROOT, atomic_json
10
+
11
+
12
+ def main():
13
+ results=ROOT/'results'
14
+ metrics=json.loads((results/'metrics.json').read_text())
15
+ for filename,digest in metrics['training_environment']['code_hashes'].items():
16
+ assert hashlib.sha256((ROOT/filename).read_bytes()).hexdigest()==digest, filename
17
+ history=np.genfromtxt(results/'history.csv',delimiter=',',names=True)
18
+ assert len(history)==500 and history['updates'][-1]==125000
19
+ assert np.array_equal(history['epoch'],np.arange(1,501))
20
+ assert np.array_equal(history['updates'],250*np.arange(1,501))
21
+ assert all(np.isfinite(history[k]).all() for k in history.dtype.names)
22
+ with np.load(results/'predictions.npz') as data:
23
+ pred,target=data['prediction'],data['target']
24
+ assert pred.shape==target.shape==(200,85,85)
25
+ assert np.isfinite(pred).all() and np.isfinite(target).all()
26
+ err=np.linalg.norm((pred.astype('float64')-target).reshape(200,-1),axis=1)/np.linalg.norm(target.astype('float64').reshape(200,-1),axis=1)
27
+ assert abs(float(err.mean())-metrics['mean_relative_l2_raw'])<1e-12
28
+ np.testing.assert_allclose(float(err.mean()),history['test_relative_l2'][-1],rtol=1e-4,atol=1e-7)
29
+ for name in ['pressure_fields.png','test_error_distribution.png','learning_curves.png','reproduction_report.md']:
30
+ assert (results/name).stat().st_size>1000
31
+ proof=dict(status='success',job_id=os.environ.get('SLURM_JOB_ID'),
32
+ epochs=500,updates=125000,test_samples=200,
33
+ mean_relative_l2_raw=metrics['mean_relative_l2_raw'],
34
+ signed_gap=metrics['signed_gap'],selection=metrics['selection'],
35
+ checkpoint=str(ROOT/'weight/best_model.pth'),report=str(results/'reproduction_report.md'))
36
+ report_path=results/'reproduction_report.md'
37
+ report=report_path.read_text()
38
+ provenance='''\n## Execution provenance\n\nThe prior interrupted session had completed specifications and handoffs but left the model, script, and weight directories empty. This session reused those documents and the original data. Slurm smoke job1100898 passed all six paths. Job1100916 failed at startup before training output or checkpoint creation. The same full-budget experiment continued as job1100920 after excluding that startup-fault node; no formal epochs were discarded or restarted from an existing trained model. See `.onescience/task_state_before_resume.json` and runtime events for the audit trail.\n'''
39
+ if '## Execution provenance' not in report:
40
+ report_path.write_text(report+provenance)
41
+ for directory,skill in [('.trainer_work/darcy85','onescience-trainer'),('.infer_work','onescience-infer')]:
42
+ work=ROOT/directory
43
+ atomic_json(work/'execution_result.json',dict(skill=skill,status='success',artifacts=proof,
44
+ observation={'validation':'passed; results/artifact_audit.json','limitations':metrics['limitations']}))
45
+ (work/'validation_report.md').write_text((results/'reproduction_report.md').read_text())
46
+ atomic_json(work/'runtime_result.json',dict(execution_channel='local_slurm',execution_mode='slurm',
47
+ preflight_passed=True,submission_state='submitted',execution_state='success',
48
+ job_id=os.environ.get('SLURM_JOB_ID'),local_log_dir=str(ROOT/'logs'),blocking_reason=None))
49
+ manifest_path=work/('trainer_run_manifest.json' if skill=='onescience-trainer' else 'inference_run_manifest.json')
50
+ manifest=json.loads(manifest_path.read_text()) if manifest_path.exists() else {}
51
+ manifest.update(status='success',job_id=os.environ.get('SLURM_JOB_ID'),artifacts=proof)
52
+ manifest['phase_status']={k:'success' for k in manifest.get('phase_status',{'execution':'running','validation':'pending'})}
53
+ atomic_json(manifest_path,manifest)
54
+ state_path=ROOT/'.onescience/task_state.json'
55
+ state=json.loads(state_path.read_text())
56
+ state.update(status='completed',current_phase='completed',active_step=None,blockers=[],final_result=proof)
57
+ for step in state['global_plan']:
58
+ step['status']='success'
59
+ state['events'].append(dict(event_type='task_completed',evidence=str(results/'artifact_audit.json')))
60
+ atomic_json(state_path,state)
61
+ atomic_json(results/'artifact_audit.json',proof)
62
+ print('FINAL_AUDIT',proof,flush=True)
63
+
64
+
65
+ if __name__=='__main__':
66
+ main()
scripts/inference.py ADDED
@@ -0,0 +1,72 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ """Independently reload epoch-500 weights; evaluate all 200 samples at raw scale."""
2
+ import hashlib
3
+ import importlib.metadata
4
+ from copy import deepcopy
5
+
6
+ import numpy as np
7
+ import torch
8
+
9
+ from common import (ROOT, MoNoLight, atomic_json, config_args, environment, evaluate,
10
+ load_data, normalized, path, read_config, require_compute_node)
11
+
12
+
13
+ def main():
14
+ args = config_args(__doc__).parse_args()
15
+ cfg = read_config(args.config)
16
+ device = require_compute_node()
17
+ checkpoint = path(cfg['training']['checkpoint'])
18
+ state = torch.load(checkpoint, map_location='cpu', weights_only=False)
19
+ assert state['epoch'] == 500 and state['global_step'] == 125000
20
+ trained_config = deepcopy(state['config'])
21
+ trained_config['data']['directory'] = cfg['data']['directory']
22
+ assert trained_config == cfg
23
+ model = MoNoLight(**cfg['model']).to(device)
24
+ model.load_state_dict(state['model'], strict=True)
25
+ raw = load_data(cfg)
26
+ data, stats = normalized(raw, state['normalization'], device)
27
+ errors, predictions = evaluate(model, data, stats, return_predictions=True)
28
+ assert predictions.shape == (200, 7225, 1)
29
+ # Independent double precision NumPy recomputation of the requested metric.
30
+ pred = predictions.numpy().reshape(200,85,85)
31
+ target = raw['test_sol'].numpy().reshape(200,85,85)
32
+ coeff = raw['test_coeff'].numpy().reshape(200,85,85)
33
+ per_sample = np.linalg.norm((pred.astype(np.float64)-target).reshape(200,-1),axis=1) / np.linalg.norm(target.astype(np.float64).reshape(200,-1),axis=1)
34
+ np.testing.assert_allclose(errors.numpy(), per_sample, rtol=2e-5, atol=1e-8)
35
+ result_dir = path(cfg['results'])
36
+ np.savez(result_dir/'predictions.npz', prediction=pred, target=target,
37
+ coeff=coeff, coordinates=raw['coordinates_raw'].numpy().reshape(85,85,2),
38
+ relative_l2=per_sample, sample_id=np.arange(200))
39
+ np.savetxt(result_dir/'per_sample_relative_l2.csv',
40
+ np.column_stack((np.arange(200),per_sample)), delimiter=',',
41
+ header='sample_id,relative_l2_raw', comments='', fmt=['%d','%.12g'])
42
+ mean = float(per_sample.mean())
43
+ reference = cfg['paper_relative_l2']
44
+ metrics = dict(mean_relative_l2_raw=mean, test_samples=200,
45
+ actual_training_epochs=state['epoch'], optimizer_updates=state['global_step'],
46
+ paper_table3=reference, signed_gap=mean-reference,
47
+ absolute_gap=abs(mean-reference), relative_gap_percent=100*(mean/reference-1),
48
+ selection=state['selection'], checkpoint=str(checkpoint),
49
+ checkpoint_sha256=hashlib.sha256(checkpoint.read_bytes()).hexdigest(),
50
+ min=float(per_sample.min()), max=float(per_sample.max()),
51
+ median=float(np.median(per_sample)),
52
+ training_environment=state['environment'], inference_environment=environment(),
53
+ software_versions={name:importlib.metadata.version(name) for name in
54
+ ['torch','triton','onescience','numpy','scipy','PyYAML','matplotlib']},
55
+ configured_cluster='SCNet compute node for the original training',
56
+ gpu_total_memory_bytes=torch.cuda.get_device_properties(device).total_memory,
57
+ limitations=[
58
+ 'Single DCU/PyTorch 2.4.1 HIP backend versus paper RTX4090/PyTorch2.8 CUDA12.8.',
59
+ 'Configured system RAM allocation is 30G, versus up to 600GB reported by the paper; all 1000/200 samples and model dimensions retained.',
60
+ 'Installed Triton is 3.0.0 with DTK customizations, versus paper Triton3.4.0; author fused kernels are not used.',
61
+ 'Independent implementation, packed QKV/SDPA attention, and RNG consumption differ from explicit author attention; not bitwise reproduction.',
62
+ 'Portable PyTorch log-domain Sinkhorn instead of author fused Triton kernels; same finite-step equations and differentiation.',
63
+ 'Identical Darcy geometry is computed once and broadcast over batch; arithmetic ordering differs.',
64
+ 'Paper does not specify seed uncertainty or a tolerance for the Table3 result.',
65
+ 'Final epoch checkpoint named best_model.pth; no test-set model selection.',
66
+ 'Units unspecified by supplied data; metric uses stored pressure scale.'])
67
+ atomic_json(result_dir/'metrics.json', metrics)
68
+ print('INFERENCE', metrics, flush=True)
69
+
70
+
71
+ if __name__ == '__main__':
72
+ main()
scripts/result.py ADDED
@@ -0,0 +1,90 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ """Render the fields, errors, learning curves, and audited reproduction report."""
2
+ import json
3
+
4
+ import matplotlib
5
+ matplotlib.use('Agg')
6
+ import matplotlib.pyplot as plt
7
+ import numpy as np
8
+
9
+ from common import config_args, path, read_config
10
+
11
+
12
+ def main():
13
+ args = config_args(__doc__).parse_args()
14
+ cfg = read_config(args.config)
15
+ results = path(cfg['results'])
16
+ metrics = json.loads((results/'metrics.json').read_text())
17
+ with np.load(results/'predictions.npz') as data:
18
+ assert data['prediction'].shape == data['target'].shape == (200,85,85)
19
+ fig, axs = plt.subplots(4,4,figsize=(14,12),constrained_layout=True)
20
+ for i in range(4):
21
+ gt, pred = data['target'][i], data['prediction'][i]
22
+ fields = [data['coeff'][i], gt, pred, np.abs(pred-gt)]
23
+ for j, (field, title) in enumerate(zip(fields,['Permeability','True pressure','Predicted pressure','Absolute error'])):
24
+ kw = dict(vmin=min(gt.min(),pred.min()),vmax=max(gt.max(),pred.max())) if j in [1,2] else {}
25
+ im=axs[i,j].imshow(field,origin='lower',extent=(0,1,0,1),cmap='magma' if j==3 else 'viridis',**kw)
26
+ axs[i,j].set_title(f'Sample {i}: {title}')
27
+ axs[i,j].set_xlabel('x'); axs[i,j].set_ylabel('y')
28
+ fig.colorbar(im,ax=axs[i,j],shrink=.8)
29
+ fig.savefig(results/'pressure_fields.png',dpi=180,bbox_inches='tight')
30
+ plt.close(fig)
31
+ fig,ax=plt.subplots(figsize=(7,4),constrained_layout=True)
32
+ ax.hist(data['relative_l2'],bins=25,color='#286ba5')
33
+ ax.axvline(metrics['mean_relative_l2_raw'],label='Reproduction mean',color='#d44b28')
34
+ ax.axvline(.0070,label='Paper Table 3',color='black',linestyle='--')
35
+ ax.set(xlabel='Raw-scale relative L2 per test sample',ylabel='Count')
36
+ ax.legend(); fig.savefig(results/'test_error_distribution.png',dpi=180,bbox_inches='tight'); plt.close(fig)
37
+ h=np.genfromtxt(results/'history.csv',delimiter=',',names=True)
38
+ assert len(h)==500 and np.array_equal(h['epoch'],np.arange(1,501))
39
+ fig,axs=plt.subplots(1,2,figsize=(11,4),constrained_layout=True)
40
+ axs[0].semilogy(h['epoch'],h['train_relative_l2'],label='Training')
41
+ axs[0].semilogy(h['epoch'],h['test_relative_l2'],label='Test monitoring')
42
+ axs[0].axhline(.007,linestyle='--',color='black',label='Paper Table 3')
43
+ axs[0].set(xlabel='Epoch',ylabel='Raw-scale relative L2'); axs[0].legend()
44
+ axs[1].plot(h['epoch'],h['learning_rate']); axs[1].set(xlabel='Epoch',ylabel='Learning rate')
45
+ fig.savefig(results/'learning_curves.png',dpi=180,bbox_inches='tight'); plt.close(fig)
46
+ limitations='\n'.join('- '+x for x in metrics['limitations'])
47
+ report=f'''# MoNo-light Darcy 85×85 reproduction
48
+
49
+ Paper: https://arxiv.org/pdf/2608.09764v1, Table 3 and Appendix F.
50
+
51
+ ## Measured result
52
+
53
+ - Raw-scale per-sample relative L2 mean over all 200 test samples: **{metrics['mean_relative_l2_raw']:.10f}**.
54
+ - Paper Table 3: 0.0070; signed difference: **{metrics['signed_gap']:+.10f}**; absolute difference: {metrics['absolute_gap']:.10f}; relative difference: {metrics['relative_gap_percent']:+.4f}%.
55
+ - Actual completed training: **500 epochs / 125000 optimizer updates**, batch size 4, max LR .001, 100 warmup epochs.
56
+ - Training elapsed time: {float(h['elapsed_seconds'][-1])/3600:.4f} hours.
57
+ - Weight choice: **final epoch 500**, saved as `weight/best_model.pth`. Test metrics never select or tune weights. No validation split or early stopping.
58
+ - Checkpoint SHA256: `{metrics['checkpoint_sha256']}`.
59
+ - Parameters: {metrics['training_environment']['trainable_parameters']:,} trainable; {metrics['training_environment']['all_parameters']:,} including the frozen anchor copy (about 1.73M total).
60
+ - Actual package versions: `{metrics['software_versions']}`. Slurm resources: `{metrics['configured_cluster']}`.
61
+
62
+ ## Data and metric
63
+
64
+ Reuse supplied `piececonst_r421_N1024_smooth1.mat[0:1000]` and `piececonst_r421_N1024_smooth2.mat[0:200]`, fields coeff/sol. Slice `[:,::5,::5]` to 85×85. Global per-channel unbiased train statistics only; reuse for test. Normalize input coordinates/coeff and pressure; inverse-transform pressure before loss and evaluation. Metric is arithmetic mean of each sample's norm(pred-target)/norm(target), recomputed independently with float64 NumPy after checkpoint reload.
65
+
66
+ ## Alignment and limitations
67
+
68
+ {limitations}
69
+
70
+ Implementation sources and paper-versus-supplement distinction are preserved in `.trainer_work/darcy85/source_capture.md`. Official code was consulted read-only for paper-unstated details, including a resume audit; no repository downloaded, cloned, copied, or imported. This session reused the saved paper and training specifications.
71
+
72
+ ## Artifacts
73
+
74
+ - `predictions.npz`: all 200 raw-scale predictions, targets, permeability, coordinates and errors.
75
+ - `per_sample_relative_l2.csv`, `metrics.json`, `history.csv`, `data_manifest.json`, `training_environment.json`.
76
+ - `pressure_fields.png`: fixed test samples 0--3, common pressure color scale within each sample.
77
+ - `test_error_distribution.png`, `learning_curves.png`.
78
+ - Runtime stdout/stderr and Slurm evidence: `../logs/` and `../.onescience/`.
79
+ '''
80
+ report_path=results/'reproduction_report.md'
81
+ if report_path.exists():
82
+ previous=report_path.read_text()
83
+ if '## Execution provenance' in previous:
84
+ report+='\n## Execution provenance'+previous.split('## Execution provenance',1)[1]
85
+ report_path.write_text(report)
86
+ print(report,flush=True)
87
+
88
+
89
+ if __name__ == '__main__':
90
+ main()
scripts/smoke.py ADDED
@@ -0,0 +1,77 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ """Isolated six-path check on real Darcy samples, never used as formal training."""
2
+ import math
3
+ import time
4
+
5
+ import numpy as np
6
+ import torch
7
+
8
+ from common import (MoNoLight, atomic_json, compute_stats, config_args, environment,
9
+ load_data, normalized, path, read_config, relative_l2,
10
+ require_compute_node, seed_all)
11
+ from model.MoNo import sinkhorn_transport
12
+
13
+
14
+ def main():
15
+ cfg=read_config(config_args(__doc__).parse_args().config)
16
+ device=require_compute_node()
17
+ seed_all(0)
18
+ raw=load_data(cfg)
19
+ data,stats=normalized(raw,compute_stats(raw),device)
20
+ model=MoNoLight(**cfg['model']).to(device)
21
+ checks={}
22
+ model.train()
23
+ t=time.monotonic()
24
+ y=model(data['coordinates'],data['train_x'][:4])
25
+ assert y.shape==(4,7225,1) and torch.isfinite(y).all()
26
+ checks['forward_pass']='passed'
27
+ y=y*stats['y_std']+stats['y_mean']
28
+ loss=relative_l2(y,data['train_y'][:4]).mean()
29
+ loss.backward()
30
+ assert all(p.grad is not None and torch.isfinite(p.grad).all() for p in model.parameters() if p.requires_grad)
31
+ checks['backward_pass']='passed'
32
+ torch.cuda.synchronize()
33
+ print(f'SMOKE full-resolution forward/backward {time.monotonic()-t:.3f}s',flush=True)
34
+ # One isolated epoch on 8 actual training samples, two complete optimizer updates.
35
+ # Formal training is a separate fresh seed-0 model and still uses all 1000 samples.
36
+ opt=torch.optim.AdamW(model.parameters(),lr=1e-5)
37
+ losses=[]
38
+ start=time.monotonic()
39
+ for j in range(2):
40
+ opt.zero_grad(set_to_none=True)
41
+ y=model(data['coordinates'],data['train_x'][4*j:4*j+4])*stats['y_std']+stats['y_mean']
42
+ loss=relative_l2(y,data['train_y'][4*j:4*j+4]).mean()
43
+ loss.backward(); opt.step()
44
+ losses.append(loss.item())
45
+ assert np.isfinite(losses).all()
46
+ checks['train_loop_dry_run']='passed: one isolated epoch on 8 real samples; formal budget unchanged'
47
+ torch.cuda.synchronize()
48
+ seconds=(time.monotonic()-start)/2
49
+ model.eval()
50
+ with torch.no_grad():
51
+ y=model(data['coordinates'],data['test_x'][:4])*stats['y_std']+stats['y_mean']
52
+ metric=relative_l2(y,data['test_y'][:4])
53
+ assert torch.isfinite(metric).all()
54
+ checks['validation_loop_dry_run']='passed'
55
+ yn=y.cpu().numpy().astype(np.float64).reshape(4,-1)
56
+ tn=data['test_y'][:4].cpu().numpy().astype(np.float64).reshape(4,-1)
57
+ np.testing.assert_allclose(metric.cpu().numpy(),np.linalg.norm(yn-tn,axis=1)/np.linalg.norm(tn,axis=1),rtol=1e-5)
58
+ # Independently verify finite Sinkhorn equations and derivatives in float64.
59
+ z=torch.randn(1,17,7,device=device,dtype=torch.float64,requires_grad=True)
60
+ p=sinkhorn_transport(z)
61
+ torch.testing.assert_close(p.sum(-2),torch.full((1,7),1/7,device=device,dtype=torch.float64),atol=1e-10,rtol=1e-10)
62
+ assert torch.autograd.gradcheck(sinkhorn_transport,(z,),fast_mode=True)
63
+ checks['domain_metric_dry_run']='passed: raw relative L2 vs NumPy; OT marginal and finite-step gradcheck'
64
+ assert cfg['model']['width']==96 and tuple(cfg['model']['latent_sizes'])==(512,256,128,64)
65
+ assert tuple(cfg['model']['depths'])==(3,1,1,1)
66
+ assert not any('weight_g' in k for k in model.state_dict())
67
+ checks['config_consistency']='passed'
68
+ result=dict(status='passed',attempt=1,checks=checks,environment=environment(),
69
+ seconds_per_training_step=seconds,estimated_formal_hours=seconds*125000/3600,
70
+ trainable_parameters=sum(p.numel() for p in model.parameters() if p.requires_grad),
71
+ losses=losses,smoke_samples=8,formal_samples=1000,formal_epochs=500)
72
+ atomic_json(path(cfg['results'])/'smoke_checks.json',result)
73
+ print('SMOKE',result,flush=True)
74
+
75
+
76
+ if __name__=='__main__':
77
+ main()
scripts/train.py ADDED
@@ -0,0 +1,129 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ """Full 500-epoch Darcy85 training, with loss/metrics printed as training runs."""
2
+ import csv
3
+ import os
4
+ import random
5
+ import time
6
+
7
+ import numpy as np
8
+ import torch
9
+
10
+ from common import (ROOT, MoNoLight, atomic_json, atomic_save, code_hashes,
11
+ compute_stats, config_args, environment, evaluate, load_data,
12
+ normalized, path, read_config, relative_l2,
13
+ require_compute_node, seed_all)
14
+
15
+
16
+ def main():
17
+ parser = config_args(__doc__)
18
+ parser.add_argument('--resume', action='store_true', help='Resume full optimizer/scheduler/RNG state')
19
+ args = parser.parse_args()
20
+ print('TRAIN_START config loading', flush=True)
21
+ cfg = read_config(args.config)
22
+ print('TRAIN_START compute node and accelerator initialization', flush=True)
23
+ device = require_compute_node()
24
+ seed_all(cfg['seed'])
25
+ print('TRAIN_START data loading', flush=True)
26
+ raw = load_data(cfg)
27
+ stats_cpu = compute_stats(raw)
28
+ data, stats = normalized(raw, stats_cpu, device)
29
+ print('TRAIN_START data normalization complete', flush=True)
30
+ t = cfg['training']
31
+ results = path(cfg['results'])
32
+ results.mkdir(parents=True, exist_ok=True)
33
+ model = MoNoLight(**cfg['model']).to(device)
34
+ optimizer = torch.optim.AdamW(model.parameters(), lr=t['learning_rate'],
35
+ betas=tuple(t['betas']), weight_decay=t['weight_decay'])
36
+ scheduler = torch.optim.lr_scheduler.OneCycleLR(
37
+ optimizer, max_lr=t['learning_rate'], epochs=t['epochs'], steps_per_epoch=250,
38
+ pct_start=t['warmup_epochs']/t['epochs'], div_factor=t['div_factor'],
39
+ final_div_factor=t['final_div_factor'], anneal_strategy='cos')
40
+ first_epoch, global_step, history = 0, 0, []
41
+ elapsed_before = 0.0
42
+ checkpoint = path(t['resume_checkpoint'])
43
+ hashes = code_hashes()
44
+ if args.resume and checkpoint.exists():
45
+ state = torch.load(checkpoint, map_location=device, weights_only=False)
46
+ assert state['config'] == cfg, 'Cannot change formal configuration when resuming'
47
+ assert state['code_hashes'] == hashes, 'Training implementation changed; audit before resuming'
48
+ model.load_state_dict(state['model'])
49
+ optimizer.load_state_dict(state['optimizer'])
50
+ scheduler.load_state_dict(state['scheduler'])
51
+ first_epoch, global_step, history = state['epoch'], state['global_step'], state['history']
52
+ elapsed_before = state['elapsed_seconds']
53
+ # RNG tensors must be copied back to CPU even when optimizer maps to the accelerator.
54
+ torch.set_rng_state(state['rng_torch'].cpu())
55
+ torch.cuda.set_rng_state_all([s.cpu() for s in state['rng_cuda']])
56
+ np.random.set_state(state['rng_numpy'])
57
+ random.setstate(state['rng_python'])
58
+ print(f'RESUME epoch={first_epoch} updates={global_step}; optimizer/scheduler/RNG restored', flush=True)
59
+ elif checkpoint.exists() or path(t['checkpoint']).exists():
60
+ raise RuntimeError('Existing training weights found: use --resume; refusing overwrite/retrain')
61
+ runtime = environment()
62
+ runtime.update(trainable_parameters=sum(p.numel() for p in model.parameters() if p.requires_grad),
63
+ all_parameters=sum(p.numel() for p in model.parameters()),
64
+ precision='float32', code_hashes=hashes)
65
+ atomic_json(results / 'training_environment.json', runtime)
66
+ print('ENVIRONMENT', runtime, flush=True)
67
+ start = time.monotonic()
68
+ columns = ['epoch','updates','train_relative_l2','test_relative_l2','learning_rate','epoch_seconds','elapsed_seconds']
69
+ # The checkpoint is authoritative: discard any uncommitted trailing history from interruption.
70
+ with open(results/'history.csv', 'w', newline='') as f:
71
+ writer = csv.DictWriter(f, fieldnames=columns)
72
+ writer.writeheader()
73
+ writer.writerows(history)
74
+ for epoch in range(first_epoch, t['epochs']):
75
+ epoch_start = time.monotonic()
76
+ model.train()
77
+ permutation = torch.randperm(1000, device=device)
78
+ train_sum = torch.zeros((), device=device)
79
+ for j in range(250):
80
+ ids = permutation[j*4:(j+1)*4]
81
+ optimizer.zero_grad(set_to_none=True)
82
+ output = model(data['coordinates'], data['train_x'][ids])
83
+ output = output * stats['y_std'] + stats['y_mean']
84
+ loss = relative_l2(output, data['train_y'][ids]).mean()
85
+ if not torch.isfinite(loss):
86
+ raise FloatingPointError(f'Nonfinite loss at epoch={epoch+1}, step={j+1}')
87
+ loss.backward()
88
+ norm = torch.nn.utils.clip_grad_norm_(model.parameters(), t['grad_clip'], error_if_nonfinite=True)
89
+ optimizer.step()
90
+ scheduler.step()
91
+ global_step += 1
92
+ train_sum += loss.detach() * 4
93
+ if j == 0 or (j+1) % t['print_every_steps'] == 0:
94
+ print(f'TRAIN epoch={epoch+1}/500 batch={j+1}/250 updates={global_step} '
95
+ f'loss_raw_relative_l2={loss.item():.9f} grad_norm={norm.item():.6g} '
96
+ f'lr={scheduler.get_last_lr()[0]:.10g}', flush=True)
97
+ errors, _ = evaluate(model, data, stats)
98
+ row = dict(epoch=epoch+1, updates=global_step,
99
+ train_relative_l2=train_sum.item()/1000,
100
+ test_relative_l2=errors.mean().item(),
101
+ learning_rate=scheduler.get_last_lr()[0],
102
+ epoch_seconds=time.monotonic()-epoch_start,
103
+ elapsed_seconds=elapsed_before+time.monotonic()-start)
104
+ history.append(row)
105
+ print('EPOCH', row, flush=True)
106
+ state = dict(model=model.state_dict(), optimizer=optimizer.state_dict(),
107
+ scheduler=scheduler.state_dict(), epoch=epoch+1, global_step=global_step,
108
+ history=history, normalization=stats_cpu, config=cfg,
109
+ code_hashes=hashes, environment=runtime,
110
+ elapsed_seconds=row['elapsed_seconds'], rng_torch=torch.get_rng_state(),
111
+ rng_cuda=torch.cuda.get_rng_state_all(), rng_numpy=np.random.get_state(),
112
+ rng_python=random.getstate(), selection=t['selection'])
113
+ atomic_save(checkpoint, state)
114
+ with open(results/'history.csv', 'a', newline='') as f:
115
+ csv.DictWriter(f, fieldnames=columns).writerow(row)
116
+ atomic_json(results/'training_status.json', dict(
117
+ status='completed' if epoch+1 == 500 else 'running',
118
+ completed_epochs=epoch+1, formal_epochs=500, global_step=global_step,
119
+ formal_updates=125000, checkpoint=str(checkpoint), last_epoch=row,
120
+ slurm_job_id=os.environ.get('SLURM_JOB_ID')))
121
+ assert global_step == 125000 and len(history) == 500
122
+ final = torch.load(checkpoint, map_location='cpu', weights_only=False)
123
+ assert final['epoch'] == 500
124
+ atomic_save(path(t['checkpoint']), final)
125
+ print(f'COMPLETE 500 epochs; 125000 updates; final epoch saved to {path(t["checkpoint"])}', flush=True)
126
+
127
+
128
+ if __name__ == '__main__':
129
+ main()
weight/best_model.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d20393f9ae89d38f53b8da9e09799db57bef37154ec8c7fddb9a9dd4a6666122
3
+ size 20828886
weight/last_model.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:43bcf48d1a95103a42b92222d7cff0e6e602bafe56aa48c6e0a0d85275daa4fa
3
+ size 20838614