From: François Fleuret <francois@fleuret.org>
Date: Mon, 29 Jul 2024 10:38:48 +0000 (+0200)
Subject: Update.
X-Git-Url: https://fleuret.org/cgi-bin/gitweb/gitweb.cgi?a=commitdiff_plain;h=eb1399cc7de466cf1b62bd24c067f661c0127047;p=culture.git

Update.
---

diff --git a/main.py b/main.py
index 0d4cd10..fc703f7 100755
--- a/main.py
+++ b/main.py
@@ -87,11 +87,11 @@ parser.add_argument("--gpus", type=str, default="all")
 
 parser.add_argument("--nb_gpts", type=int, default=5)
 
-parser.add_argument("--max_fail_to_validate", type=int, default=1)
+parser.add_argument("--max_fail_to_validate", type=int, default=2)
 
 parser.add_argument("--accuracy_to_make_c_quizzes", type=float, default=0.98)
 
-parser.add_argument("--proba_understands", type=float, default=0.99)
+parser.add_argument("--proba_understands", type=float, default=0.9)
 
 parser.add_argument("--proba_not_understands", type=float, default=0.5)
 
@@ -454,7 +454,7 @@ def one_epoch(model, quiz_machine, local_device=main_device):
 
 def model_transformer_hot(model):
     # model.temperature = args.temperature_hot
-    model.set_noise_injection(1.0, ("ffw", 2))
+    model.set_noise_injection(0.5, ("ffw", args.nb_blocks // 2))
 
 
 def model_transformer_cold(model):
@@ -549,9 +549,11 @@ def record_new_c_quizzes(models, quiz_machine, nb_for_train=1000, nb_for_test=10
         # We use the model that has generated the fewest quizzes to
         # balance the number of quizzes per model overall
 
-        model_for_generation = sorted(
-            models, key=lambda m: nb_validated_per_model[m.id]
-        )[0]
+        # model_for_generation = sorted(
+        # models, key=lambda m: nb_validated_per_model[m.id]
+        # )[0]
+
+        model_for_generation = models[torch.randint(len(models), (1,)).item()]
 
         # We generate quizzes with a procedure that injects some
         # structured noise
@@ -860,6 +862,44 @@ log_string(
 
 ######################################################################
 
+#!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
+if False:
+    for model in models:
+        for b in range(args.nb_blocks):
+            for o in [0.5, 1.0, 2.0]:
+
+                def model_transformer_hot(model):
+                    # model.temperature = args.temperature_hot
+                    model.set_noise_injection(o, ("ffw", b))
+
+                def model_transformer_cold(model):
+                    pass
+
+                # model.temperature = args.temperature_cold
+
+                c_quizzes_procedure = [
+                    # (("f_B", "f_A", "A", "B"), (1, 0, 0, 0), model_transformer_hot),
+                    # (("f_B", "f_A", "A", "B"), (0, 1, 1, 1), model_transformer_cold),
+                    # (("A", "f_A", "B", "f_B"), (0, 0, 0, 1), model_transformer_cold),
+                    (("A", "f_A", "B", "f_B"), (1, 1, 1, 1), model_transformer_hot),
+                ]
+
+                c_quizzes = quiz_machine.generate_c_quizzes(
+                    128, model_for_generation=model, procedure=c_quizzes_procedure
+                )
+
+                filename = f"generated_{b:02d}_{o:.02f}_{model.id:02d}.png"
+                print(filename)
+
+                quiz_machine.problem.save_quizzes_as_image(
+                    args.result_dir,
+                    filename,
+                    c_quizzes,
+                )
+        exit(0)
+
+#!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
+
 if args.dirty_debug:
     args.accuracy_to_make_c_quizzes = 0.0
     args.nb_gpts = 2