Further reduce the bigram examples

lukstafi · lukstafi · commit 1606b94bad82 · 2025-08-17T01:25:27.000+02:00
diff --git a/test/training/bigram.expected b/test/training/bigram.expected
@@ -46,94 +46,13 @@ Epoch 10, batch 0, loss=2.4864
 Epoch 10, batch 100, loss=2.2818
 Epoch 10, batch 200, loss=2.6494
 Epoch 10, epoch loss=570.71
-Epoch 11, batch 0, loss=2.482
-Epoch 11, batch 100, loss=2.2799
-Epoch 11, batch 200, loss=2.6456
-Epoch 11, epoch loss=569.88
-Epoch 12, batch 0, loss=2.4784
-Epoch 12, batch 100, loss=2.2785
-Epoch 12, batch 200, loss=2.6423
-Epoch 12, epoch loss=569.18
-Epoch 13, batch 0, loss=2.4753
-Epoch 13, batch 100, loss=2.2773
-Epoch 13, batch 200, loss=2.6395
-Epoch 13, epoch loss=568.58
-Epoch 14, batch 0, loss=2.4726
-Epoch 14, batch 100, loss=2.2763
-Epoch 14, batch 200, loss=2.6371
-Epoch 14, epoch loss=568.07
-Epoch 15, batch 0, loss=2.4703
-Epoch 15, batch 100, loss=2.2755
-Epoch 15, batch 200, loss=2.635
-Epoch 15, epoch loss=567.62
-Epoch 16, batch 0, loss=2.4682
-Epoch 16, batch 100, loss=2.2749
-Epoch 16, batch 200, loss=2.6331
-Epoch 16, epoch loss=567.22
-Epoch 17, batch 0, loss=2.4664
-Epoch 17, batch 100, loss=2.2744
-Epoch 17, batch 200, loss=2.6314
-Epoch 17, epoch loss=566.87
-Epoch 18, batch 0, loss=2.4648
-Epoch 18, batch 100, loss=2.2739
-Epoch 18, batch 200, loss=2.6299
-Epoch 18, epoch loss=566.56
-Epoch 19, batch 0, loss=2.4634
-Epoch 19, batch 100, loss=2.2736
-Epoch 19, batch 200, loss=2.6286
-Epoch 19, epoch loss=566.28
-Epoch 20, batch 0, loss=2.4621
-Epoch 20, batch 100, loss=2.2733
-Epoch 20, batch 200, loss=2.6274
-Epoch 20, epoch loss=566.03
-Epoch 21, batch 0, loss=2.461
-Epoch 21, batch 100, loss=2.273
-Epoch 21, batch 200, loss=2.6263
-Epoch 21, epoch loss=565.81
-Epoch 22, batch 0, loss=2.4599
-Epoch 22, batch 100, loss=2.2728
-Epoch 22, batch 200, loss=2.6254
-Epoch 22, epoch loss=565.6
-Epoch 23, batch 0, loss=2.459
-Epoch 23, batch 100, loss=2.2727
-Epoch 23, batch 200, loss=2.6245
-Epoch 23, epoch loss=565.41
-Epoch 24, batch 0, loss=2.4581
-Epoch 24, batch 100, loss=2.2725
-Epoch 24, batch 200, loss=2.6237
-Epoch 24, epoch loss=565.24
-Epoch 25, batch 0, loss=2.4573
-Epoch 25, batch 100, loss=2.2724
-Epoch 25, batch 200, loss=2.623
-Epoch 25, epoch loss=565.08
-Epoch 26, batch 0, loss=2.4566
-Epoch 26, batch 100, loss=2.2723
-Epoch 26, batch 200, loss=2.6223
-Epoch 26, epoch loss=564.93
-Epoch 27, batch 0, loss=2.456
-Epoch 27, batch 100, loss=2.2722
-Epoch 27, batch 200, loss=2.6217
-Epoch 27, epoch loss=564.8
-Epoch 28, batch 0, loss=2.4553
-Epoch 28, batch 100, loss=2.2722
-Epoch 28, batch 200, loss=2.6211
-Epoch 28, epoch loss=564.67
-Epoch 29, batch 0, loss=2.4548
-Epoch 29, batch 100, loss=2.2721
-Epoch 29, batch 200, loss=2.6207
-Epoch 29, epoch loss=564.55
-Epoch 30, batch 0, loss=2.4543
-Epoch 30, batch 100, loss=2.2721
-Epoch 30, batch 200, loss=2.6202
-Epoch 30, epoch loss=564.44
 vax
 ah
-prerofaers
+pterofaers
 rmanad
-toniorelilievon
-tpdeee
+tonioreliliewwfctpdeee
 tole
-bieynel
+bkaynel
 n
 ahi
 oni
@@ -142,7 +61,8 @@ a
 krn
 c
 r
-labha
+labik
 a
-calopbrilaeiery
-jthpamahie
+calopconaiajary
+jthqamahie
+hler
diff --git a/test/training/bigram.ml b/test/training/bigram.ml
@@ -72,7 +72,7 @@ let () =
 
   let open Operation.At in
   let batch_ref = IDX.find_exn sgd_step.bindings batch_n in
-  for epoch = 0 to 30 do
+  for epoch = 0 to 10 do
     let epoch_loss = ref 0. in
     for batch = 0 to n_batches - 1 do
       batch_ref := batch;
diff --git a/test/training/bigram_mlp.expected b/test/training/bigram_mlp.expected
@@ -7,11 +7,6 @@ Epoch 1, epoch loss=565
 Epoch 2, epoch loss=562
 Epoch 3, epoch loss=561
 Epoch 4, epoch loss=560
-Epoch 5, epoch loss=560
-Epoch 6, epoch loss=559
-Epoch 7, epoch loss=559
-Epoch 8, epoch loss=558
-Epoch 9, epoch loss=558
 nnou
 kyns
 hain
diff --git a/test/training/bigram_mlp.ml b/test/training/bigram_mlp.ml
@@ -44,7 +44,7 @@ let () =
   let n_batches = input_size / batch_size in
   let batch_n, bindings = IDX.get_static_symbol ~static_range:n_batches IDX.empty in
   let step_n, bindings = IDX.get_static_symbol bindings in
-  let epochs = 10 in
+  let epochs = 5 in
   let steps = epochs * n_batches in
 
   let%op input_gram = inputs @| batch_n in
