Skip to content

Commit eaca6db

Browse files
lukstafilukstafi
committed
Make the moons boundary plots even smaller, and bigram_mlp bigger hidden dim
Signed-off-by: Lukasz Stafiniak <lukstafi@gmail.com>
1 parent 69eb133 commit eaca6db

File tree

8 files changed

+193
-279
lines changed

8 files changed

+193
-279
lines changed

‎test/training/bigram.expected‎

Lines changed: 112 additions & 112 deletions
Original file line numberDiff line numberDiff line change
@@ -2,130 +2,130 @@ Retrieving commandline, environment, or config file variable ocannl_log_level
22
Found 0, in the config file
33
bigrams: 228146
44
input_size: 229000
5-
Epoch 0, batch 0, loss=3.38421
6-
Epoch 0, batch 100, loss=2.70904
7-
Epoch 0, batch 200, loss=2.94577
8-
Epoch 0, epoch loss=680.203
9-
Epoch 1, batch 0, loss=2.78871
10-
Epoch 1, batch 100, loss=2.42573
11-
Epoch 1, batch 200, loss=2.81466
12-
Epoch 1, epoch loss=616.822
13-
Epoch 2, batch 0, loss=2.64955
14-
Epoch 2, batch 100, loss=2.35837
15-
Epoch 2, batch 200, loss=2.75535
16-
Epoch 2, epoch loss=597.547
17-
Epoch 3, batch 0, loss=2.58884
18-
Epoch 3, batch 100, loss=2.32969
19-
Epoch 3, batch 200, loss=2.72171
20-
Epoch 3, epoch loss=588.281
21-
Epoch 4, batch 0, loss=2.55456
22-
Epoch 4, batch 100, loss=2.31366
23-
Epoch 4, batch 200, loss=2.70022
5+
Epoch 0, batch 0, loss=3.3842
6+
Epoch 0, batch 100, loss=2.709
7+
Epoch 0, batch 200, loss=2.9458
8+
Epoch 0, epoch loss=680.2
9+
Epoch 1, batch 0, loss=2.7887
10+
Epoch 1, batch 100, loss=2.4257
11+
Epoch 1, batch 200, loss=2.8147
12+
Epoch 1, epoch loss=616.82
13+
Epoch 2, batch 0, loss=2.6495
14+
Epoch 2, batch 100, loss=2.3584
15+
Epoch 2, batch 200, loss=2.7553
16+
Epoch 2, epoch loss=597.55
17+
Epoch 3, batch 0, loss=2.5888
18+
Epoch 3, batch 100, loss=2.3297
19+
Epoch 3, batch 200, loss=2.7217
20+
Epoch 3, epoch loss=588.28
21+
Epoch 4, batch 0, loss=2.5546
22+
Epoch 4, batch 100, loss=2.3137
23+
Epoch 4, batch 200, loss=2.7002
2424
Epoch 4, epoch loss=582.76
25-
Epoch 5, batch 0, loss=2.53264
26-
Epoch 5, batch 100, loss=2.30338
27-
Epoch 5, batch 200, loss=2.68534
28-
Epoch 5, epoch loss=579.076
29-
Epoch 6, batch 0, loss=2.51753
30-
Epoch 6, batch 100, loss=2.29625
31-
Epoch 6, batch 200, loss=2.67441
32-
Epoch 6, epoch loss=576.442
33-
Epoch 7, batch 0, loss=2.50652
34-
Epoch 7, batch 100, loss=2.29104
35-
Epoch 7, batch 200, loss=2.66601
36-
Epoch 7, epoch loss=574.467
37-
Epoch 8, batch 0, loss=2.49818
38-
Epoch 8, batch 100, loss=2.28714
39-
Epoch 8, batch 200, loss=2.65934
40-
Epoch 8, epoch loss=572.933
41-
Epoch 9, batch 0, loss=2.49164
42-
Epoch 9, batch 100, loss=2.28414
43-
Epoch 9, batch 200, loss=2.65392
44-
Epoch 9, epoch loss=571.708
45-
Epoch 10, batch 0, loss=2.48636
25+
Epoch 5, batch 0, loss=2.5326
26+
Epoch 5, batch 100, loss=2.3034
27+
Epoch 5, batch 200, loss=2.6853
28+
Epoch 5, epoch loss=579.08
29+
Epoch 6, batch 0, loss=2.5175
30+
Epoch 6, batch 100, loss=2.2962
31+
Epoch 6, batch 200, loss=2.6744
32+
Epoch 6, epoch loss=576.44
33+
Epoch 7, batch 0, loss=2.5065
34+
Epoch 7, batch 100, loss=2.291
35+
Epoch 7, batch 200, loss=2.666
36+
Epoch 7, epoch loss=574.47
37+
Epoch 8, batch 0, loss=2.4982
38+
Epoch 8, batch 100, loss=2.2871
39+
Epoch 8, batch 200, loss=2.6593
40+
Epoch 8, epoch loss=572.93
41+
Epoch 9, batch 0, loss=2.4916
42+
Epoch 9, batch 100, loss=2.2841
43+
Epoch 9, batch 200, loss=2.6539
44+
Epoch 9, epoch loss=571.71
45+
Epoch 10, batch 0, loss=2.4864
4646
Epoch 10, batch 100, loss=2.2818
4747
Epoch 10, batch 200, loss=2.6494
48-
Epoch 10, epoch loss=570.709
49-
Epoch 11, batch 0, loss=2.48202
50-
Epoch 11, batch 100, loss=2.27995
48+
Epoch 10, epoch loss=570.71
49+
Epoch 11, batch 0, loss=2.482
50+
Epoch 11, batch 100, loss=2.2799
5151
Epoch 11, batch 200, loss=2.6456
5252
Epoch 11, epoch loss=569.88
53-
Epoch 12, batch 0, loss=2.47838
54-
Epoch 12, batch 100, loss=2.27847
55-
Epoch 12, batch 200, loss=2.64234
53+
Epoch 12, batch 0, loss=2.4784
54+
Epoch 12, batch 100, loss=2.2785
55+
Epoch 12, batch 200, loss=2.6423
5656
Epoch 12, epoch loss=569.18
57-
Epoch 13, batch 0, loss=2.47527
58-
Epoch 13, batch 100, loss=2.27728
59-
Epoch 13, batch 200, loss=2.63953
60-
Epoch 13, epoch loss=568.583
61-
Epoch 14, batch 0, loss=2.47261
62-
Epoch 14, batch 100, loss=2.27631
57+
Epoch 13, batch 0, loss=2.4753
58+
Epoch 13, batch 100, loss=2.2773
59+
Epoch 13, batch 200, loss=2.6395
60+
Epoch 13, epoch loss=568.58
61+
Epoch 14, batch 0, loss=2.4726
62+
Epoch 14, batch 100, loss=2.2763
6363
Epoch 14, batch 200, loss=2.6371
64-
Epoch 14, epoch loss=568.067
65-
Epoch 15, batch 0, loss=2.47029
66-
Epoch 15, batch 100, loss=2.27553
67-
Epoch 15, batch 200, loss=2.63496
68-
Epoch 15, epoch loss=567.618
69-
Epoch 16, batch 0, loss=2.46825
70-
Epoch 16, batch 100, loss=2.27489
71-
Epoch 16, batch 200, loss=2.63307
72-
Epoch 16, epoch loss=567.224
73-
Epoch 17, batch 0, loss=2.46645
74-
Epoch 17, batch 100, loss=2.27436
64+
Epoch 14, epoch loss=568.07
65+
Epoch 15, batch 0, loss=2.4703
66+
Epoch 15, batch 100, loss=2.2755
67+
Epoch 15, batch 200, loss=2.635
68+
Epoch 15, epoch loss=567.62
69+
Epoch 16, batch 0, loss=2.4682
70+
Epoch 16, batch 100, loss=2.2749
71+
Epoch 16, batch 200, loss=2.6331
72+
Epoch 16, epoch loss=567.22
73+
Epoch 17, batch 0, loss=2.4664
74+
Epoch 17, batch 100, loss=2.2744
7575
Epoch 17, batch 200, loss=2.6314
76-
Epoch 17, epoch loss=566.875
77-
Epoch 18, batch 0, loss=2.46485
78-
Epoch 18, batch 100, loss=2.27393
79-
Epoch 18, batch 200, loss=2.62991
80-
Epoch 18, epoch loss=566.563
81-
Epoch 19, batch 0, loss=2.46343
82-
Epoch 19, batch 100, loss=2.27358
83-
Epoch 19, batch 200, loss=2.62859
84-
Epoch 19, epoch loss=566.285
85-
Epoch 20, batch 0, loss=2.46214
86-
Epoch 20, batch 100, loss=2.27329
76+
Epoch 17, epoch loss=566.87
77+
Epoch 18, batch 0, loss=2.4648
78+
Epoch 18, batch 100, loss=2.2739
79+
Epoch 18, batch 200, loss=2.6299
80+
Epoch 18, epoch loss=566.56
81+
Epoch 19, batch 0, loss=2.4634
82+
Epoch 19, batch 100, loss=2.2736
83+
Epoch 19, batch 200, loss=2.6286
84+
Epoch 19, epoch loss=566.28
85+
Epoch 20, batch 0, loss=2.4621
86+
Epoch 20, batch 100, loss=2.2733
8787
Epoch 20, batch 200, loss=2.6274
88-
Epoch 20, epoch loss=566.033
89-
Epoch 21, batch 0, loss=2.46099
90-
Epoch 21, batch 100, loss=2.27304
91-
Epoch 21, batch 200, loss=2.62632
92-
Epoch 21, epoch loss=565.806
93-
Epoch 22, batch 0, loss=2.45995
94-
Epoch 22, batch 100, loss=2.27285
95-
Epoch 22, batch 200, loss=2.62536
96-
Epoch 22, epoch loss=565.599
88+
Epoch 20, epoch loss=566.03
89+
Epoch 21, batch 0, loss=2.461
90+
Epoch 21, batch 100, loss=2.273
91+
Epoch 21, batch 200, loss=2.6263
92+
Epoch 21, epoch loss=565.81
93+
Epoch 22, batch 0, loss=2.4599
94+
Epoch 22, batch 100, loss=2.2728
95+
Epoch 22, batch 200, loss=2.6254
96+
Epoch 22, epoch loss=565.6
9797
Epoch 23, batch 0, loss=2.459
98-
Epoch 23, batch 100, loss=2.27268
99-
Epoch 23, batch 200, loss=2.62447
98+
Epoch 23, batch 100, loss=2.2727
99+
Epoch 23, batch 200, loss=2.6245
100100
Epoch 23, epoch loss=565.41
101-
Epoch 24, batch 0, loss=2.45814
102-
Epoch 24, batch 100, loss=2.27254
103-
Epoch 24, batch 200, loss=2.62367
104-
Epoch 24, epoch loss=565.238
105-
Epoch 25, batch 0, loss=2.45735
106-
Epoch 25, batch 100, loss=2.27242
107-
Epoch 25, batch 200, loss=2.62296
108-
Epoch 25, epoch loss=565.078
109-
Epoch 26, batch 0, loss=2.45662
110-
Epoch 26, batch 100, loss=2.27233
111-
Epoch 26, batch 200, loss=2.62229
112-
Epoch 26, epoch loss=564.932
113-
Epoch 27, batch 0, loss=2.45596
114-
Epoch 27, batch 100, loss=2.27225
115-
Epoch 27, batch 200, loss=2.62169
116-
Epoch 27, epoch loss=564.796
117-
Epoch 28, batch 0, loss=2.45534
118-
Epoch 28, batch 100, loss=2.27219
119-
Epoch 28, batch 200, loss=2.62114
101+
Epoch 24, batch 0, loss=2.4581
102+
Epoch 24, batch 100, loss=2.2725
103+
Epoch 24, batch 200, loss=2.6237
104+
Epoch 24, epoch loss=565.24
105+
Epoch 25, batch 0, loss=2.4573
106+
Epoch 25, batch 100, loss=2.2724
107+
Epoch 25, batch 200, loss=2.623
108+
Epoch 25, epoch loss=565.08
109+
Epoch 26, batch 0, loss=2.4566
110+
Epoch 26, batch 100, loss=2.2723
111+
Epoch 26, batch 200, loss=2.6223
112+
Epoch 26, epoch loss=564.93
113+
Epoch 27, batch 0, loss=2.456
114+
Epoch 27, batch 100, loss=2.2722
115+
Epoch 27, batch 200, loss=2.6217
116+
Epoch 27, epoch loss=564.8
117+
Epoch 28, batch 0, loss=2.4553
118+
Epoch 28, batch 100, loss=2.2722
119+
Epoch 28, batch 200, loss=2.6211
120120
Epoch 28, epoch loss=564.67
121-
Epoch 29, batch 0, loss=2.45478
122-
Epoch 29, batch 100, loss=2.27213
123-
Epoch 29, batch 200, loss=2.62065
124-
Epoch 29, epoch loss=564.553
125-
Epoch 30, batch 0, loss=2.45426
126-
Epoch 30, batch 100, loss=2.27208
127-
Epoch 30, batch 200, loss=2.62018
128-
Epoch 30, epoch loss=564.444
121+
Epoch 29, batch 0, loss=2.4548
122+
Epoch 29, batch 100, loss=2.2721
123+
Epoch 29, batch 200, loss=2.6207
124+
Epoch 29, epoch loss=564.55
125+
Epoch 30, batch 0, loss=2.4543
126+
Epoch 30, batch 100, loss=2.2721
127+
Epoch 30, batch 200, loss=2.6202
128+
Epoch 30, epoch loss=564.44
129129
vax
130130
ah
131131
prerofaers

‎test/training/bigram.ml‎

Lines changed: 2 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -79,9 +79,9 @@ let () =
7979
Train.run sgd_step;
8080
let loss = batch_loss.@[0] in
8181
epoch_loss := !epoch_loss +. loss;
82-
if batch % 100 = 0 then Stdio.printf "Epoch %d, batch %d, loss=%.6g\n%!" epoch batch loss;
82+
if batch % 100 = 0 then Stdio.printf "Epoch %d, batch %d, loss=%.5g\n%!" epoch batch loss;
8383
done;
84-
Stdio.printf "Epoch %d, epoch loss=%.6g\n%!" epoch !epoch_loss
84+
Stdio.printf "Epoch %d, epoch loss=%.5g\n%!" epoch !epoch_loss
8585
done;
8686
(* Train.printf_tree batch_loss; *)
8787

0 commit comments

Comments
 (0)