-
Notifications
You must be signed in to change notification settings - Fork 2
/
run_plot.sh
executable file
·413 lines (369 loc) · 37.1 KB
/
run_plot.sh
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
#!/bin/bash
# COMPRESSAI_PLOT=(poetry run compressai-plot)
PLOT=(
# "${COMPRESSAI_PLOT[@]}"
poetry run compressai-plot
--aim_repo="$HOME/data/aim/pc-mordor/pcc"
)
COMMON_ARGS=(--x='bpp_loss' --y='acc_top1' --optimal="convex")
# COMMON_ARGS=(--x='bpp_loss' --y='acc_top1')
# Useful queries:
# run.created_at >= datetime(2023, 5, 16)
# run.dataset.train.meta.name == "ModelNet40"
# # FULL_CSV generated by:
# "${PLOT[@]}" "${COMMON_ARGS[@]}" \
# --query='"pcc-cls-only-pointnet" in run.model.name and run.hp.num_channels.g_a == [3,64,64,64,128,1024]' \
# --out_csv=results/plot_rd/modelnet40_bpp_loss_full.csv \
# --out_html=results/plot_rd/modelnet40_bpp_loss_full.html
#
# # LITE_CSV generated by:
# "${PLOT[@]}" "${COMMON_ARGS[@]}" \
# --query='"pcc-cls-only-pointnet" in run.model.name and run.hp.num_channels.g_a == [3,8,8,16,16,32]' \
# --out_csv=results/plot_rd/modelnet40_bpp_loss_lite.csv \
# --out_html=results/plot_rd/modelnet40_bpp_loss_lite.html
#
# # AGGREGATE_NEW_TSV generated using:
# poetry run python scripts/eval_modelnet.py
# # Original lite query:
# "${PLOT[@]}" "${COMMON_ARGS[@]}" \
# --query='"pcc" in run.model.name and run.hp.num_channels.g_a == [3,8,8,16,16,32]' \
# --out_csv=results/plot_rd/modelnet40_bpp_loss_lite.csv \
# --out_html=results/plot_rd/modelnet40_bpp_loss_lite.html
# # run.hp.num_points=1024
# # run.hp.num_channels.g_a=[3,8,64]
# "${PLOT[@]}" "${COMMON_ARGS[@]}" \
# --query='run.hash in "9e12dd9929fe4a02adfa849d d43385b8f039447d85c78171 574386ebe91a4571a88f558f b68f7630627e4238b40bcc2f 61471e16d91f4b24bd2a9c8e dc54ca5b5c0c4a52ba00da0e 023ff4ecb494450ab2c336dd b63dcdbd7ba5459995b2212d d3756ef84fa246f4846c58a0 4c1510780fe94923aaeba1ef 50e59b3e01294da391d6e212 aca683c052df45c5a44d1f48".split()' \
# --out_csv=results/plot_rd/tmp.csv \
# --out_html=results/plot_rd/tmp.html
# # run.hp.num_points=256
# # run.hp.num_channels.g_a=[3,8,64]
# "${PLOT[@]}" "${COMMON_ARGS[@]}" \
# --query='run.hash in "b98114f5eb004aacb8f92ed3 fc219ce93262457595eacb4f 5442605311e5452094be8857 c6a25f7ab07e4ec38a58fea2 05c0f3f2f9c1406bb9c7a052 e044b5af826c4b48a48bc4d8 43fe065ab5a7425fbe0ab1d1 94fb3c03cb7b43ac94a093e2 9bb12d4633e74af9a64266d4 a70440160a53423682c04c31 5545638d22ae407fa0e077a2 c8b1440260b346d58bb6adac".split()' \
# --out_csv=results/plot_rd/tmp.csv \
# --out_html=results/plot_rd/tmp.html
# run.hp.num_points=1024
# run.hp.num_channels.g_a=[3,16]
# "${PLOT[@]}" "${COMMON_ARGS[@]}" \
# --query='run.hash in "9447fba58ce04a1b836621d9 ed15eed869c348f08fcbf15b bc7a1998e7384883a5e174b7 8c4a69f9ac2d40fbacedca7c b472c2d6ac764d78a07b13db 7d369f1007c24b97b488097d 1851701e65d2441695331a6a 30198d1b37e04f2387203d11 9fa0733b39cd47aca81e8ace c8a9ddb4348046d3b5ab6bd4 22ed2eb6149a48edaa51b367 c99577da24c148dbbf0066de ba06b075cdd24ce986f33eb5 ec9f1105d16049799e2696d9 bd2516b456cc49319ea05b2c e6c77d1286004fdea99f9b35".split()' \
# --out_csv=results/plot_rd/tmp.csv \
# --out_html=results/plot_rd/tmp.html
# # run.hp.num_points=512
# # run.hp.num_channels.g_a=[3,16]
# "${PLOT[@]}" "${COMMON_ARGS[@]}" \
# --query='run.hash in "bb7486944652489cad3a2c89 1540cae611c1454bba77432e e411bd7be18f4c22b9484d35 c477b2894a7c44e486a4e935 d7d68e9859764a1e8a5423a0 1fe03b2799524037bdf7efe1 44518ee9d7524f1694ec318d bd49ddcb089a4f45ad1d57fe 61f4bae97485469692ecee06 e586ff8ea57347d1b91d3588 f3774aa69055419a8c6bd830 e190f64a15a94aff96383d1d 4dd085864aac41269c3067fe 2e5976382c7e4a7689893c67".split()' \
# --out_csv=results/plot_rd/tmp.csv \
# --out_html=results/plot_rd/tmp.html
# # run.hp.num_points=256
# # run.hp.num_channels.g_a=[3,16]
# "${PLOT[@]}" "${COMMON_ARGS[@]}" \
# --query='run.hash in "76211de7afa841d7b0a5be9a 76f1dd907ccb4365981d43a7 53be35046367458db7688271 5ceb9a61b1ed4e159e089527 ebcbfa6aa950423daba3f4c0 7f2fb2556240470196fb5486 c7769020447c4aaa882f42f6 d8b9e1b2b0a54ec79a9453ae 34e394c444c944d0a0e2f6e7 d7921331f2404114aaaa4084 7c634070046a421c9c88b97a 786fdabbe75246e29567cae9 de9f3213dde64e3d9c32f5f7 ed4d5236a6f4430b9fbf861d c1e48ed1894f40c7bd9c6157 41e6689071c84bd88997c92e".split()' \
# --out_csv=results/plot_rd/tmp.csv \
# --out_html=results/plot_rd/tmp.html
# # run.hp.num_points=128
# # run.hp.num_channels.g_a=[3,16]
# "${PLOT[@]}" "${COMMON_ARGS[@]}" \
# --query='run.hash in "2e178372769f470998d509f3 c3404cb946924c09b53ee55e bec4f1f91c544ac58200d5f3 8df152cffcf34f9ebeb168d2 dad61bbe762e4942bbb3dc67 5dbb5375d08e4d4bbf34e4be feece5ad2cea4a80a476ce8b e736fff7d5354e1581ca97a5 13a24b77d9334277b98d96e8 c2409b9266fc416cad6873c5 f3d6046d4ff748bf978d24d1 72075040a27a4e988e3d1f75 fa94475428754c85abf4b070 a7c7cd62129841fa85edbc40".split()' \
# --out_csv=results/plot_rd/tmp.csv \
# --out_html=results/plot_rd/tmp.html
# # run.hp.num_points=64
# # run.hp.num_channels.g_a=[3,16]
# "${PLOT[@]}" "${COMMON_ARGS[@]}" \
# --query='run.hash in "b022afc6b77d4c64883f6dbc e0c0739731654d48ab70bfe5 c5d81a5b64eb4ca38edc9106 092281fcda184698a5c8c201 469cb0240d3f48a09acb8844 137ad8fe1ff0475da0e04301 8cd7fe909cbc4e04b82eb867 5b36bfe4f7e84cec806d1487 bbfea59c17294e3c8df6f092 3df0169963fe44bcae64819a 7aefc21e1074420c9320b13b e4208cb581f0465ea01fc9e8 fec053ab1c9745dd8bb0754c 843f9f3a0a5f4b8b86f4ae5d".split()' \
# --out_csv=results/plot_rd/tmp.csv \
# --out_html=results/plot_rd/tmp.html
# # run.hp.num_points=32
# # run.hp.num_channels.g_a=[3,16]
# "${PLOT[@]}" "${COMMON_ARGS[@]}" \
# --query='run.hash in "d04b68cfa61c4230bb222161 b1cece584daa4cd6bb7effc6 7c84e98fa332407899149f72 fce64c08291a4590a0eda6af 6da5e18ee6bc436a9042b80d ac8589442b3d46138faf8497 7e9775c39db145a1b0ad09e2 d61b91494caa4e9dae5ac6ab d8f0b5f4197147ddb381b1fb 611eac082787411691aff1f6 f8a59f922bc94527b40047bb 3cc2a96b171e40d7b228e780 083eb08172b44e4387a68fd6 8d36792b9dfb4fe7b026df8b".split()' \
# --out_csv=results/plot_rd/tmp.csv \
# --out_html=results/plot_rd/tmp.html
# # run.hp.num_points=16
# # run.hp.num_channels.g_a=[3,16]
# "${PLOT[@]}" "${COMMON_ARGS[@]}" \
# --query='run.hash in "8f32f06074b84d869d2f9a1b 75db469d4fc64e178db46836 fe3182982785416cb3287490 87da16e6ef39424493df7c1b fc1dd7a797df4561ace00c7c 08f7a4052005453d80bbb83c 6dd0d1d949dc40dab38f46a7 c028ad01b4b141798fc9e8b8 4e58c04043b64742ae290025 0e7516e8e6544601bbc8d0ae 2d9f1076e28646df9bb9e3bf babe2862f56c4890af2492fc b6942869982b4f5dae455d27 26de322da6de4b438cd43171".split()' \
# --out_csv=results/plot_rd/tmp.csv \
# --out_html=results/plot_rd/tmp.html
# # run.hp.num_points=8
# # run.hp.num_channels.g_a=[3,16]
# "${PLOT[@]}" "${COMMON_ARGS[@]}" \
# --query='run.hash in "8c21b43e3fa54b15a6d00963 c67daf2c81b64b75b8315787 daa4c862966d4431907fee4e 9454a35d132748eea67384a3 086cf7bac0044075b028f2d7 0b9c0fb8fbb649d7ac40cb6b bb5d46037d074eddb4a006ba 4af73ebfeb114109bda1bba8 5e8cbe03878a4887a6e8219f 52cf7fa690bd42a6be7aabed e3a5d5d350774faea80fe4dd bf30415799c84855bb461ffd 3797ad5091ff4c7998e7dc80 22ae8d7f6a1c41b78428df9c".split()' \
# --out_csv=results/plot_rd/tmp.csv \
# --out_html=results/plot_rd/tmp.html
# Write result JSON files, then plot RD curves:
# # porp scripts/save_json_from_aim_query.py
# # porp scripts/plot_rd_mpl_new.py
# Plot point clouds for "micro" codec:
# # porp scripts/plot_point_cloud_mpl.py aadd008012aa40b7bf41d5e7 d657d2773d784935ac8d373f 87f30aee7f0445a4b3a545d8 1f7de7fd3b274e8b89234bfa f54c39d00c334adf903d6a6e 3fe3d84cd7954b75906961d8 29d47d0a657d47e39e900067 915dbd00bcf04aa2896344cb 6cc7b7899ee7411c9f2dbd51 623023c8ab164652bcdb7386 d85c7e59cab247d790a07038 67be12941e4e427380a03e44; echo "micro"
# # Plot point clouds for "lite" codec:
# porp scripts/plot_point_cloud_mpl.py dd702cc73af74bb6a9b61ff3 409c1460ec784662b14dfcbc 63e1dbf94b694d7693dc2cb7 e783b170eda14aa29224c92b 47f31d8b7ef54d039e8e05d3 3616dc2455214fdeb4f4b62d b459932f207b4063976da932 232b3be0e7934c80beb6e389 76a4b98a929244818f28e1ba 06c3bfe845c44c849780114d d6d13bfbb26a4a9798232a22 7e17d0a21ff340fc95f0da58; echo "lite"
# # Plot point clouds for "full" codec:
# porp scripts/plot_point_cloud_mpl.py fe4122df657d40dea6774a47 b67c87a00d5240cc936f6a43 d2880736c02c44b5b540156c 7a1ccbf555ad4d3e896a9a3b 6e0091c06ce044699ecdd206 19f2d7a78f1740f18decd4c9 7c69baf921d84a4482dd1be6 92088f9972e4495aaeefb770 35482eca40bc4d88b3598894 677b6063456b4930a27c2857 2e2eb122c87a4041b7f3f587 5f2ab297622a4a3fa3a1214b; echo "full"
# # Query relevant run_hashes.
#
# poetry run python -c '
# import sys
# import aim
# from compressai_trainer.utils.aim.query import get_runs_dataframe, run_hashes_by_query
# repo = aim.Repo(sys.argv[1])
# query = sys.argv[2]
# run_hashes = run_hashes_by_query(repo, query)
# df = get_runs_dataframe(run_hashes, repo=repo, metrics=["bpp_loss", "acc_top1"], hparams=["criterion.lmbda.cls", "hp.num_points", "hp.num_channels.g_a"])
# df["hp.num_channels.g_a"] = df["hp.num_channels.g_a"].map(tuple)
# df = df.sort_values(["hp.num_channels.g_a", "model.name", "criterion.lmbda.cls"]).reset_index(drop=True)
# df = df[["run_hash", "hp.num_channels.g_a", "model.name", "criterion.lmbda.cls", "bpp_loss", "acc_top1"]]
# print(df.to_string(index=False))
# ' \
# "$HOME/data/aim/pc-mordor/pcc" \
# "run.hp.num_points == 1024"
# porp -c 'import sys; import aim; from compressai_trainer.utils.aim.query import get_runs_dataframe, run_hashes_by_query; repo = aim.Repo(sys.argv[1]); query = sys.argv[2]; run_hashes = run_hashes_by_query(repo, query); df = get_runs_dataframe(run_hashes, repo=repo, metrics=["bpp_loss", "acc_top1"], hparams=["criterion.lmbda.cls", "hp.num_points", "hp.num_channels.g_a"]); df["hp.num_channels.g_a"] = df["hp.num_channels.g_a"].map(tuple); df = df.sort_values(["hp.num_channels.g_a", "model.name", "criterion.lmbda.cls"]).reset_index(drop=True); df = df[["run_hash", "hp.num_channels.g_a", "model.name", "criterion.lmbda.cls", "bpp_loss", "acc_top1"]]; print(df.to_string(index=False))' "$HOME/data/aim/pc-mordor/pcc" "run.hp.num_points == 1024"
# CRITICAL POINT SETS
# Classification-only:
#
# RUN_HASHES=(
# # micro
# # 9447fba58ce04a1b836621d9 # um-pcc-cls-only-pointnet-mini-001 10.620185 0.528363 10
# ed15eed869c348f08fcbf15b # um-pcc-cls-only-pointnet-mini-001 12.647167 0.600081 14
# # bc7a1998e7384883a5e174b7 # um-pcc-cls-only-pointnet-mini-001 16.284716 0.654376 20
# # 8c4a69f9ac2d40fbacedca7c # um-pcc-cls-only-pointnet-mini-001 20.186681 0.707861 28
# b472c2d6ac764d78a07b13db # um-pcc-cls-only-pointnet-mini-001 24.557064 0.742707 40
# # 7d369f1007c24b97b488097d # um-pcc-cls-only-pointnet-mini-001 36.497228 0.775932 80
# 1851701e65d2441695331a6a # um-pcc-cls-only-pointnet-mini-001 48.696047 0.805511 160
# # 30198d1b37e04f2387203d11 # um-pcc-cls-only-pointnet-mini-001 61.583964 0.816856 320
# # 9fa0733b39cd47aca81e8ace # um-pcc-cls-only-pointnet-mini-001 73.351176 0.824554 1000
# # c8a9ddb4348046d3b5ab6bd4 # um-pcc-cls-only-pointnet-mini-001 78.305946 0.825770 4000
# # 22ed2eb6149a48edaa51b367 # um-pcc-cls-only-pointnet-mini-001 78.062356 0.827391 6000
# ec9f1105d16049799e2696d9 # um-pcc-cls-only-pointnet-mini-001 78.978195 0.829417 16000
# # c99577da24c148dbbf0066de # um-pcc-cls-only-pointnet-mini-001 78.144314 0.821313 164000
#
# # lite
# ea428fd42da84dbcbc1f7c34 # um-pcc-cls-only-pointnet 10.983490 0.579822 10
# 5e90b82ce0ab44808345d57b # um-pcc-cls-only-pointnet 12.454252 0.636143 14
# ae645b27cd59429193667fb5 # um-pcc-cls-only-pointnet 13.723202 0.668963 20
# 95c1ea383a584e5aa29d63ce # um-pcc-cls-only-pointnet 18.090725 0.719206 28
# 32d71c950b6243a5b3bc322d # um-pcc-cls-only-pointnet 22.008024 0.762966 40
# b159f611678f405fb942b8e7 # um-pcc-cls-only-pointnet 32.941536 0.794165 80
# 0b27e0776f534cf485cf56e6 # um-pcc-cls-only-pointnet 48.020104 0.820502 160
# 7189f08374f14026a1d782d5 # um-pcc-cls-only-pointnet 65.661915 0.844408 320
# # 73528fbba3624827823566bd # um-pcc-cls-only-pointnet 105.422323 0.845624 1000
# a7dceeefa39747ed90bf9e47 # um-pcc-cls-only-pointnet 137.008761 0.850486 4000
# # cdbb97f57a264f80acf73543 # um-pcc-cls-only-pointnet 157.043000 0.847245 16000
# # 7ab67b4d6fab45a4bc2cf9a3 # um-pcc-cls-only-pointnet 153.931869 0.841977 64000
# # eff4d7bce95c4c1582107a45 # um-pcc-cls-only-pointnet 160.215024 0.848460 256000
#
# # full
# 8a59f52ab3d04649913357c3 # um-pcc-cls-only-pointnet 8.055724 0.451378 10
# 48b76524474e4c1b8206056d # um-pcc-cls-only-pointnet 10.378093 0.593598 14
# 18388dd3556b4af8ba48c038 # um-pcc-cls-only-pointnet 16.711520 0.724878 28
# 55f1acfe0e304065b328202b # um-pcc-cls-only-pointnet 14.498847 0.694895 20
# 5d374263257a4f08b73bf21d # um-pcc-cls-only-pointnet 20.964733 0.765397 40
# 11b05bf4293e4e728caf66a1 # um-pcc-cls-only-pointnet 30.778704 0.810778 80
# c9e04f474c2b4852a4ddaf85 # um-pcc-cls-only-pointnet 46.490746 0.839546 160
# dfd71410fa16468697fb2d47 # um-pcc-cls-only-pointnet 60.244190 0.850081 240
# dc57d83678854c2eaf63ac5e # um-pcc-cls-only-pointnet 74.491030 0.858995 320
# 817ef4ee42c04083a3fbed82 # um-pcc-cls-only-pointnet 108.516576 0.863047 640
# deff251f04694d46ac04b797 # um-pcc-cls-only-pointnet 139.423393 0.876418 1000
# # 3ff11150c09c437e9b415df4 # um-pcc-cls-only-pointnet 208.790239 0.876823 2000
# # 80cc133d0b8847bab5dd7f03 # um-pcc-cls-only-pointnet 317.559084 0.878444 4000
# d701c67e3f03430e930ddadd # um-pcc-cls-only-pointnet 789.780310 0.884927 16000
# # ffe35513aed94664bcffbb57 # um-pcc-cls-only-pointnet 2251.736108 0.882901 64000
# # 05ec3dc03d054eeb80a313c2 # um-pcc-cls-only-pointnet 3730.229713 0.879660 256000
# # 791e48be414f49d5ab9f922b # um-pcc-cls-only-pointnet 4154.949795 0.883712 512000
# # 9e514bff8b574694ace3d18a # um-pcc-cls-only-pointnet 4244.303898 0.875203 1024000
# # d3ecd54998414437a956dc59 # um-pcc-cls-only-pointnet 4328.507145 0.879660 2048000
# # 1571d65ecb63423c975e8ac9 # um-pcc-cls-only-pointnet 4353.796289 0.878849 4096000
# # 37810e597237441caa9e8a33 # um-pcc-cls-only-pointnet 4455.765900 0.877634 16384000
# # b5fb04685dbe4e748f294c88 # um-pcc-cls-only-pointnet 4531.615613 0.881280 65536000
# )
RUN_HASHES=(
# run_hash # hp.num_channels.g_a model.name criterion.lmbda.cls bpp_loss acc_top1
# 9447fba58ce04a1b836621d9 # (3, 16) um-pcc-cls-only-pointnet-mini-001 10 10.620185 0.528363
# ed15eed869c348f08fcbf15b # (3, 16) um-pcc-cls-only-pointnet-mini-001 14 12.647167 0.600081
# bc7a1998e7384883a5e174b7 # (3, 16) um-pcc-cls-only-pointnet-mini-001 20 16.284716 0.654376
# 8c4a69f9ac2d40fbacedca7c # (3, 16) um-pcc-cls-only-pointnet-mini-001 28 20.186681 0.707861
# b472c2d6ac764d78a07b13db # (3, 16) um-pcc-cls-only-pointnet-mini-001 40 24.557064 0.742707
# 7d369f1007c24b97b488097d # (3, 16) um-pcc-cls-only-pointnet-mini-001 80 36.497228 0.775932
# 1851701e65d2441695331a6a # (3, 16) um-pcc-cls-only-pointnet-mini-001 160 48.696047 0.805511
# 30198d1b37e04f2387203d11 # (3, 16) um-pcc-cls-only-pointnet-mini-001 320 61.583964 0.816856
# 9fa0733b39cd47aca81e8ace # (3, 16) um-pcc-cls-only-pointnet-mini-001 1000 73.351176 0.824554
# c8a9ddb4348046d3b5ab6bd4 # (3, 16) um-pcc-cls-only-pointnet-mini-001 4000 78.305946 0.825770
# 22ed2eb6149a48edaa51b367 # (3, 16) um-pcc-cls-only-pointnet-mini-001 6000 78.062356 0.827391
# ba06b075cdd24ce986f33eb5 # (3, 16) um-pcc-cls-only-pointnet-mini-001 8000 79.214763 0.819287
# 38b6b528ab2c408d9e1f6fcd # (3, 16) um-pcc-cls-only-pointnet-mini-001 16000 78.999304 0.830227
# 75e587404be44441bb8b92c5 # (3, 16) um-pcc-cls-only-pointnet-mini-001 16000 77.732358 0.826175
# ec9f1105d16049799e2696d9 # (3, 16) um-pcc-cls-only-pointnet-mini-001 16000 78.978195 0.829417
# bd2516b456cc49319ea05b2c # (3, 16) um-pcc-cls-only-pointnet-mini-001 64000 78.089627 0.818071
# c99577da24c148dbbf0066de # (3, 16) um-pcc-cls-only-pointnet-mini-001 164000 78.144314 0.821313
# e6c77d1286004fdea99f9b35 # (3, 16) um-pcc-cls-only-pointnet-mini-001 256000 78.294346 0.820097
# aadd008012aa40b7bf41d5e7 # (3, 16) um-pcc-multitask-cls-pointnet 10 43.530414 0.733793
d657d2773d784935ac8d373f # (3, 16) um-pcc-multitask-cls-pointnet 14 31.798217 0.684360
# 87f30aee7f0445a4b3a545d8 # (3, 16) um-pcc-multitask-cls-pointnet 20 42.148722 0.745948
1f7de7fd3b274e8b89234bfa # (3, 16) um-pcc-multitask-cls-pointnet 28 20.582290 0.711102
f54c39d00c334adf903d6a6e # (3, 16) um-pcc-multitask-cls-pointnet 40 28.714629 0.738655
# 3fe3d84cd7954b75906961d8 # (3, 16) um-pcc-multitask-cls-pointnet 80 41.712438 0.797002
# 29d47d0a657d47e39e900067 # (3, 16) um-pcc-multitask-cls-pointnet 160 54.002624 0.802269
# 915dbd00bcf04aa2896344cb # (3, 16) um-pcc-multitask-cls-pointnet 320 68.545377 0.815640
6cc7b7899ee7411c9f2dbd51 # (3, 16) um-pcc-multitask-cls-pointnet 1000 76.265409 0.820502
# 623023c8ab164652bcdb7386 # (3, 16) um-pcc-multitask-cls-pointnet 4000 77.950269 0.817666
# d85c7e59cab247d790a07038 # (3, 16) um-pcc-multitask-cls-pointnet 8000 78.470701 0.821718
# 9d1c625213474253989bb599 # (3, 16) um-pcc-multitask-cls-pointnet 16000 76.367574 0.816045 # NOTE What happened here w.r.t. airplane sample model?! WOW! Just became a blob... And yet, the average accuracy...
# 370e286fbaf24ecf85850caf # (3, 16) um-pcc-multitask-cls-pointnet 16000 0.076509 0.820908
# 67be12941e4e427380a03e44 # (3, 16) um-pcc-multitask-cls-pointnet 16000 78.049098 0.821313
# 25db4c30a1fe43ff93fbf0f6 # (3, 16) um-pcc-multitask-cls-pointnet 16000 77.748679 0.824554
# ea428fd42da84dbcbc1f7c34 # (3, 8, 8, 16, 16, 32) um-pcc-cls-only-pointnet 10 10.983490 0.579822
# 5e90b82ce0ab44808345d57b # (3, 8, 8, 16, 16, 32) um-pcc-cls-only-pointnet 14 12.454252 0.636143
# ae645b27cd59429193667fb5 # (3, 8, 8, 16, 16, 32) um-pcc-cls-only-pointnet 20 13.723202 0.668963
# 95c1ea383a584e5aa29d63ce # (3, 8, 8, 16, 16, 32) um-pcc-cls-only-pointnet 28 18.090725 0.719206
# 32d71c950b6243a5b3bc322d # (3, 8, 8, 16, 16, 32) um-pcc-cls-only-pointnet 40 22.008024 0.762966
# b159f611678f405fb942b8e7 # (3, 8, 8, 16, 16, 32) um-pcc-cls-only-pointnet 80 32.941536 0.794165
# 0b27e0776f534cf485cf56e6 # (3, 8, 8, 16, 16, 32) um-pcc-cls-only-pointnet 160 48.020104 0.820502
# 7189f08374f14026a1d782d5 # (3, 8, 8, 16, 16, 32) um-pcc-cls-only-pointnet 320 65.661915 0.844408
# 73528fbba3624827823566bd # (3, 8, 8, 16, 16, 32) um-pcc-cls-only-pointnet 1000 105.422323 0.845624
# a7dceeefa39747ed90bf9e47 # (3, 8, 8, 16, 16, 32) um-pcc-cls-only-pointnet 4000 137.008761 0.850486
# cdbb97f57a264f80acf73543 # (3, 8, 8, 16, 16, 32) um-pcc-cls-only-pointnet 16000 157.043000 0.847245
# 7ab67b4d6fab45a4bc2cf9a3 # (3, 8, 8, 16, 16, 32) um-pcc-cls-only-pointnet 64000 153.931869 0.841977
# eff4d7bce95c4c1582107a45 # (3, 8, 8, 16, 16, 32) um-pcc-cls-only-pointnet 256000 160.215024 0.848460
# dd702cc73af74bb6a9b61ff3 # (3, 8, 8, 16, 16, 32) um-pcc-multitask-cls-pointnet 10 97.743256 0.717180
409c1460ec784662b14dfcbc # (3, 8, 8, 16, 16, 32) um-pcc-multitask-cls-pointnet 14 15.061701 0.619935
63e1dbf94b694d7693dc2cb7 # (3, 8, 8, 16, 16, 32) um-pcc-multitask-cls-pointnet 20 18.964734 0.695705
# e783b170eda14aa29224c92b # (3, 8, 8, 16, 16, 32) um-pcc-multitask-cls-pointnet 28 24.115921 0.705024
# 47f31d8b7ef54d039e8e05d3 # (3, 8, 8, 16, 16, 32) um-pcc-multitask-cls-pointnet 40 26.246635 0.778768
3616dc2455214fdeb4f4b62d # (3, 8, 8, 16, 16, 32) um-pcc-multitask-cls-pointnet 80 39.130638 0.809968
# b459932f207b4063976da932 # (3, 8, 8, 16, 16, 32) um-pcc-multitask-cls-pointnet 160 56.186260 0.832253
# 232b3be0e7934c80beb6e389 # (3, 8, 8, 16, 16, 32) um-pcc-multitask-cls-pointnet 320 73.038072 0.834684
76a4b98a929244818f28e1ba # (3, 8, 8, 16, 16, 32) um-pcc-multitask-cls-pointnet 1000 130.053458 0.841977
# 06c3bfe845c44c849780114d # (3, 8, 8, 16, 16, 32) um-pcc-multitask-cls-pointnet 4000 143.230317 0.841977
# d6d13bfbb26a4a9798232a22 # (3, 8, 8, 16, 16, 32) um-pcc-multitask-cls-pointnet 8000 147.347270 0.842382
# 9ca415935ee64e2fa1dfeb4f # (3, 8, 8, 16, 16, 32) um-pcc-multitask-cls-pointnet 16000 148.434026 0.841572
# 7e17d0a21ff340fc95f0da58 # (3, 8, 8, 16, 16, 32) um-pcc-multitask-cls-pointnet 16000 149.592830 0.848460
# 8a59f52ab3d04649913357c3 # (3, 64, 64, 64, 128, 1024) um-pcc-cls-only-pointnet 10 8.055724 0.451378
# 48b76524474e4c1b8206056d # (3, 64, 64, 64, 128, 1024) um-pcc-cls-only-pointnet 14 10.378093 0.593598
# 55f1acfe0e304065b328202b # (3, 64, 64, 64, 128, 1024) um-pcc-cls-only-pointnet 20 14.498847 0.694895
# 18388dd3556b4af8ba48c038 # (3, 64, 64, 64, 128, 1024) um-pcc-cls-only-pointnet 28 16.711520 0.724878
# 5d374263257a4f08b73bf21d # (3, 64, 64, 64, 128, 1024) um-pcc-cls-only-pointnet 40 20.964733 0.765397
# 11b05bf4293e4e728caf66a1 # (3, 64, 64, 64, 128, 1024) um-pcc-cls-only-pointnet 80 30.778704 0.810778
# c9e04f474c2b4852a4ddaf85 # (3, 64, 64, 64, 128, 1024) um-pcc-cls-only-pointnet 160 46.490746 0.839546
# dfd71410fa16468697fb2d47 # (3, 64, 64, 64, 128, 1024) um-pcc-cls-only-pointnet 240 60.244190 0.850081
# dc57d83678854c2eaf63ac5e # (3, 64, 64, 64, 128, 1024) um-pcc-cls-only-pointnet 320 74.491030 0.858995
# 817ef4ee42c04083a3fbed82 # (3, 64, 64, 64, 128, 1024) um-pcc-cls-only-pointnet 640 108.516576 0.863047
# deff251f04694d46ac04b797 # (3, 64, 64, 64, 128, 1024) um-pcc-cls-only-pointnet 1000 139.423393 0.876418
# 3ff11150c09c437e9b415df4 # (3, 64, 64, 64, 128, 1024) um-pcc-cls-only-pointnet 2000 208.790239 0.876823
# 80cc133d0b8847bab5dd7f03 # (3, 64, 64, 64, 128, 1024) um-pcc-cls-only-pointnet 4000 317.559084 0.878444
# d701c67e3f03430e930ddadd # (3, 64, 64, 64, 128, 1024) um-pcc-cls-only-pointnet 16000 789.780310 0.884927
# ffe35513aed94664bcffbb57 # (3, 64, 64, 64, 128, 1024) um-pcc-cls-only-pointnet 64000 2251.736108 0.882901
# 05ec3dc03d054eeb80a313c2 # (3, 64, 64, 64, 128, 1024) um-pcc-cls-only-pointnet 256000 3730.229713 0.879660
# 791e48be414f49d5ab9f922b # (3, 64, 64, 64, 128, 1024) um-pcc-cls-only-pointnet 512000 4154.949795 0.883712
# 9e514bff8b574694ace3d18a # (3, 64, 64, 64, 128, 1024) um-pcc-cls-only-pointnet 1024000 4244.303898 0.875203
# d3ecd54998414437a956dc59 # (3, 64, 64, 64, 128, 1024) um-pcc-cls-only-pointnet 2048000 4328.507145 0.879660
# 1571d65ecb63423c975e8ac9 # (3, 64, 64, 64, 128, 1024) um-pcc-cls-only-pointnet 4096000 4353.796289 0.878849
# 37810e597237441caa9e8a33 # (3, 64, 64, 64, 128, 1024) um-pcc-cls-only-pointnet 16384000 4455.765900 0.877634
# b5fb04685dbe4e748f294c88 # (3, 64, 64, 64, 128, 1024) um-pcc-cls-only-pointnet 65536000 4531.615613 0.881280
# fe4122df657d40dea6774a47 # (3, 64, 64, 64, 128, 1024) um-pcc-multitask-cls-pointnet 10 11.220641 0.560778
b67c87a00d5240cc936f6a43 # (3, 64, 64, 64, 128, 1024) um-pcc-multitask-cls-pointnet 14 15.212449 0.659238
# d2880736c02c44b5b540156c # (3, 64, 64, 64, 128, 1024) um-pcc-multitask-cls-pointnet 20 17.641623 0.725284 # NOTE also interesting...
7a1ccbf555ad4d3e896a9a3b # (3, 64, 64, 64, 128, 1024) um-pcc-multitask-cls-pointnet 28 21.752992 0.777147
# 6e0091c06ce044699ecdd206 # (3, 64, 64, 64, 128, 1024) um-pcc-multitask-cls-pointnet 40 25.137637 0.788898
# 19f2d7a78f1740f18decd4c9 # (3, 64, 64, 64, 128, 1024) um-pcc-multitask-cls-pointnet 80 35.037599 0.836305
7c69baf921d84a4482dd1be6 # (3, 64, 64, 64, 128, 1024) um-pcc-multitask-cls-pointnet 160 53.775588 0.852512
# 92088f9972e4495aaeefb770 # (3, 64, 64, 64, 128, 1024) um-pcc-multitask-cls-pointnet 320 79.064011 0.865478
# 35482eca40bc4d88b3598894 # (3, 64, 64, 64, 128, 1024) um-pcc-multitask-cls-pointnet 1000 172.812956 0.866288
677b6063456b4930a27c2857 # (3, 64, 64, 64, 128, 1024) um-pcc-multitask-cls-pointnet 4000 347.090900 0.880875
# 2e2eb122c87a4041b7f3f587 # (3, 64, 64, 64, 128, 1024) um-pcc-multitask-cls-pointnet 8000 551.711173 0.880875
# 5f2ab297622a4a3fa3a1214b # (3, 64, 64, 64, 128, 1024) um-pcc-multitask-cls-pointnet 16000 696.513746 0.877634
# bf7e8d76f85f423bb41c7158 # (3, 64, 64, 64, 128, 1024) um-pcc-multitask-cls-pointnet 16000 891.239041 0.874392
# 2b60e7ccebd84e01a121abba # (3, 32) um-pcc-cls-only-pointnet-mini-001 16000 144.800634 0.814830
# 2ad75fda850a46f4926648ff # (3, 33) um-pcc-cls-only-pointnet-mini-001 16000 150.874226 0.831037
# f738dd2f866244189101c050 # (3, 33) um-pcc-cls-only-pointnet-mini-001 16000 160.415925 0.716775
# 04827a25c49242ccbc6c90d0 # (3, 66) um-pcc-cls-only-pointnet-mini-001 16000 307.946254 0.710697
# e33502d57812453aa21b6e79 # (3, 4, 32) um-pcc-cls-only-pointnet-mini-001 16000 86.801987 0.601702
# dd6aeb4cddd14d8192766551 # (3, 8, 8, 32) um-pcc-cls-only-pointnet-mini-001 10 9.296978 0.543760
# df4e180abb2241d4880a6bde # (3, 8, 8, 32) um-pcc-cls-only-pointnet-mini-001 14 12.657459 0.622771
# ca94229306a04e2e8a723711 # (3, 8, 8, 32) um-pcc-cls-only-pointnet-mini-001 20 15.098439 0.674230
# a8a20cb8cc8547d39519d665 # (3, 8, 8, 32) um-pcc-cls-only-pointnet-mini-001 28 18.968165 0.721232
# 988d55d3399c4877bdec3571 # (3, 8, 8, 32) um-pcc-cls-only-pointnet-mini-001 40 22.927232 0.769854
# 59d2aa10d2c343f7ab95b7c7 # (3, 8, 8, 32) um-pcc-cls-only-pointnet-mini-001 80 30.493327 0.782010
# d873d0d114cf4f0db6816177 # (3, 8, 8, 32) um-pcc-cls-only-pointnet-mini-001 160 46.007315 0.800243
# 61e9fd7a8bdd4a219fdba50d # (3, 8, 8, 32) um-pcc-cls-only-pointnet-mini-001 160 152.054280 0.606969
# f2a021eebb3d405a950b79af # (3, 8, 8, 32) um-pcc-cls-only-pointnet-mini-001 160 43.756137 0.816451
# 7acb6b78c3e744d2bb624343 # (3, 8, 8, 32) um-pcc-cls-only-pointnet-mini-001 160 47.387769 0.806321
# 2992f894e8a64822ba0d2188 # (3, 8, 8, 32) um-pcc-cls-only-pointnet-mini-001 160 50.461308 0.821313
# 1c881125f138443493e8ba16 # (3, 8, 8, 32) um-pcc-cls-only-pointnet-mini-001 160 48.572602 0.807942
# c651b7f2e2cd4c59a50de281 # (3, 8, 8, 32) um-pcc-cls-only-pointnet-mini-001 160 47.992585 0.809157
# 9f8724fb326a4ef4bbbc13d2 # (3, 8, 8, 32) um-pcc-cls-only-pointnet-mini-001 320 69.187710 0.825770
# cd8cb67a0e32483483cfc48c # (3, 8, 8, 32) um-pcc-cls-only-pointnet-mini-001 1000 111.212405 0.831442
# 79fbc97a56ad4bd390da424e # (3, 8, 8, 32) um-pcc-cls-only-pointnet-mini-001 4000 136.483891 0.827391
# 1e90542553204cc7b9945b22 # (3, 8, 8, 32) um-pcc-cls-only-pointnet-mini-001 16000 141.078980 0.831037
# befef72bec9b437e87fb9700 # (3, 8, 8, 32) um-pcc-cls-only-pointnet-mini-001 16000 142.281192 0.830632
# 66930b25010a4f009815f88c # (3, 8, 8, 32) um-pcc-cls-only-pointnet-mini-001 16000 141.897184 0.825365
# 192718b578434d4797d713e7 # (3, 8, 8, 32) um-pcc-cls-only-pointnet-mini-001 16000 144.258711 0.835900
# 8b0601cf0fd1468faf47ca46 # (3, 8, 8, 32) um-pcc-cls-only-pointnet-mini-001 16000 147.325375 0.822123
# 36f625472fa448f5830f568f # (3, 8, 8, 32) um-pcc-cls-only-pointnet-mini-001 16000 120.916641 0.831037
# a1064758a50e461690532669 # (3, 8, 8, 64) um-pcc-cls-only-pointnet-mini-001 16000 213.917874 0.831442
# c699ec6f3c2e4d77b1d8d15c # (3, 8, 16, 64) um-pcc-cls-only-pointnet-mini-001 16000 209.282557 0.832658
# 5c6752c4cb9d4acfbee11717 # (3, 8, 32) um-pcc-cls-only-pointnet-mini-001 160 45.781502 0.796596
# b917b156b7f741b88c33d3f5 # (3, 8, 32) um-pcc-cls-only-pointnet-mini-001 16000 113.622243 0.816045
# cce37f2f54584a5c9266cc08 # (3, 8, 32) um-pcc-cls-only-pointnet-mini-001 16000 114.585639 0.816451
# 9e12dd9929fe4a02adfa849d # (3, 8, 64) um-pcc-cls-only-pointnet-mini-001 10 9.905030 0.478930
# d43385b8f039447d85c78171 # (3, 8, 64) um-pcc-cls-only-pointnet-mini-001 14 12.407139 0.567261
# 98c5ac61d374472e83cd2b8e # (3, 8, 64) um-pcc-cls-only-pointnet-mini-001 16 13.248549 0.591167
# 574386ebe91a4571a88f558f # (3, 8, 64) um-pcc-cls-only-pointnet-mini-001 20 15.633077 0.631686
# b68f7630627e4238b40bcc2f # (3, 8, 64) um-pcc-cls-only-pointnet-mini-001 28 19.888383 0.701378
# 61471e16d91f4b24bd2a9c8e # (3, 8, 64) um-pcc-cls-only-pointnet-mini-001 40 23.820020 0.722853
# dc54ca5b5c0c4a52ba00da0e # (3, 8, 64) um-pcc-cls-only-pointnet-mini-001 80 31.886107 0.768639
# f68f2282166b40dc977e20bd # (3, 8, 64) um-pcc-cls-only-pointnet-mini-001 160 46.423074 0.799838
# 023ff4ecb494450ab2c336dd # (3, 8, 64) um-pcc-cls-only-pointnet-mini-001 160 51.443622 0.803079
# b63dcdbd7ba5459995b2212d # (3, 8, 64) um-pcc-cls-only-pointnet-mini-001 320 69.010982 0.826175
# d3756ef84fa246f4846c58a0 # (3, 8, 64) um-pcc-cls-only-pointnet-mini-001 1000 126.716031 0.826580
# 4c1510780fe94923aaeba1ef # (3, 8, 64) um-pcc-cls-only-pointnet-mini-001 4000 211.315149 0.827796
# 50e59b3e01294da391d6e212 # (3, 8, 64) um-pcc-cls-only-pointnet-mini-001 6000 212.664158 0.838331
# e36643a73ce44f98b7306f63 # (3, 8, 64) um-pcc-cls-only-pointnet-mini-001 16000 228.017627 0.831848
# 38c0ebd176874536a7474c0b # (3, 8, 64) um-pcc-cls-only-pointnet-mini-001 16000 149.534167 0.745948
# 538989d7b0524bdf9978620e # (3, 8, 64) um-pcc-cls-only-pointnet-mini-001 16000 223.183946 0.822934
# 2d6366fa08ac4a5090511c25 # (3, 8, 64) um-pcc-cls-only-pointnet-mini-001 16000 228.017627 0.831848
# aca683c052df45c5a44d1f48 # (3, 8, 64) um-pcc-cls-only-pointnet-mini-001 164000 236.782315 0.836305
)
# Condensed version:
RUN_HASHES=(
# run_hash # hp.num_channels.g_a model.name criterion.lmbda.cls bpp_loss acc_top1
d657d2773d784935ac8d373f # (3, 16) um-pcc-multitask-cls-pointnet 14 31.798217 0.684360
1f7de7fd3b274e8b89234bfa # (3, 16) um-pcc-multitask-cls-pointnet 28 20.582290 0.711102
f54c39d00c334adf903d6a6e # (3, 16) um-pcc-multitask-cls-pointnet 40 28.714629 0.738655
6cc7b7899ee7411c9f2dbd51 # (3, 16) um-pcc-multitask-cls-pointnet 1000 76.265409 0.820502
409c1460ec784662b14dfcbc # (3, 8, 8, 16, 16, 32) um-pcc-multitask-cls-pointnet 14 15.061701 0.619935
63e1dbf94b694d7693dc2cb7 # (3, 8, 8, 16, 16, 32) um-pcc-multitask-cls-pointnet 20 18.964734 0.695705
3616dc2455214fdeb4f4b62d # (3, 8, 8, 16, 16, 32) um-pcc-multitask-cls-pointnet 80 39.130638 0.809968
76a4b98a929244818f28e1ba # (3, 8, 8, 16, 16, 32) um-pcc-multitask-cls-pointnet 1000 130.053458 0.841977
b67c87a00d5240cc936f6a43 # (3, 64, 64, 64, 128, 1024) um-pcc-multitask-cls-pointnet 14 15.212449 0.659238
7a1ccbf555ad4d3e896a9a3b # (3, 64, 64, 64, 128, 1024) um-pcc-multitask-cls-pointnet 28 21.752992 0.777147
7c69baf921d84a4482dd1be6 # (3, 64, 64, 64, 128, 1024) um-pcc-multitask-cls-pointnet 160 53.775588 0.852512
677b6063456b4930a27c2857 # (3, 64, 64, 64, 128, 1024) um-pcc-multitask-cls-pointnet 4000 347.090900 0.880875
)
FILENAMES=(
micro_1
micro_2
micro_3
micro_4
lite_1
lite_2
lite_3
lite_4
full_1
full_2
full_3
full_4
)
mkdir -p results/plot_pointcloud/pdf/crit/
mkdir -p results/plot_pointcloud/pdf/rec/
for ((i=0; i < "${#RUN_HASHES[@]}"; i+=1)); do
run_hash="${RUN_HASHES[i]}"
filename="${FILENAMES[i]}"
echo ">>> ${i} ${run_hash} ${filename}"
poetry run python scripts/plot_critical_point_set.py \
--config-path="$HOME/data/runs/pc-mordor/${run_hash}/configs" \
--config-name="config" \
++model.source="config" \
++paths.model_checkpoint='${paths.checkpoints}/runner.last.pth' \
++misc.out_path.critical="results/plot_pointcloud/pdf/crit/${filename}.pdf" \
++misc.out_path.reconstruction="results/plot_pointcloud/pdf/rec/${filename}.pdf"
# ++misc.out_path.critical="results/plot_pointcloud/pdf/crit/${i}_${run_hash}.pdf" \
# ++misc.out_path.reconstruction="results/plot_pointcloud/pdf/rec/${i}_${run_hash}.pdf"
done