From: François Fleuret Date: Mon, 1 Jul 2024 10:37:06 +0000 (+0300) Subject: Update. X-Git-Url: https://fleuret.org/cgi-bin/gitweb/gitweb.cgi?a=commitdiff_plain;h=239a52ec7face6fcd4515916e80813702fbdf49b;p=culture.git Update. --- diff --git a/main.py b/main.py index 9d95034..6137834 100755 --- a/main.py +++ b/main.py @@ -383,100 +383,76 @@ def run_tests(model, quizz_machine, deterministic_synthesis): ###################################################################### +def valid_c_quizzes(recorded, criteria): + result = [q[criteria(c)] for q, c in recorded] + return torch.cat(result, dim=0) if len(result) > 0 else torch.tensor([]) + + +###################################################################### + + def create_c_quizzes( models, quizz_machine, nb_for_train=1000, nb_for_test=100, - min_ave_seq_logproba=None, ): - # We will store the generated quizzes for each number of - # correct prediction - recorded = dict([(n, []) for n in range(len(models) + 1)]) + recorded = [] - model_indexes = [] sum_logits, sum_nb_c_quizzes = 0, 0 - def nb_generated(): - return sum([sum([x.size(0) for x in recorded[n]]) for n in recorded.keys()]) - - def nb_validated(): - return sum( - [ - sum([x.size(0) for x in recorded[n]]) - for n in range(args.min_to_validate, args.max_to_validate + 1) - ] - ) - nb_to_create = nb_for_train + nb_for_test - warnings.warn( - f"{args.nb_gpts=} {args.nb_models_for_generation=} {args.min_to_validate=} {args.max_to_validate=}" + # ------------------------------------------------------------ + + standard_validity = lambda nb_correct: torch.logical_and( + nb_correct >= args.min_to_validate, nb_correct <= args.max_to_validate ) - while nb_validated() < nb_to_create: - ( - new_c_quizzes, - nb_correct, - ave_seq_logproba, - ) = quizz_machine.gang_create_c_quizzes( - nb=nb_to_create, - nb_models_for_generation=args.nb_models_for_generation, - models=models, - mode=args.generation_mode, + while valid_c_quizzes(recorded, standard_validity).size(0) < nb_to_create: + model_for_generation = models[torch.randint(len(models), (1,))] + + c_quizzes, ave_seq_logproba = quizz_machine.generate_quizzes( + nb_to_create, + model_for_generation=model_for_generation, reverse_cleanup=args.reverse_cleanup, - min_ave_seq_logproba=min_ave_seq_logproba, - n_epoch=n_epoch, - result_dir=args.result_dir, ) - sum_logits += new_c_quizzes.size(0) * ave_seq_logproba - sum_nb_c_quizzes += new_c_quizzes.size(0) + sum_logits += c_quizzes.size(0) * ave_seq_logproba + sum_nb_c_quizzes += c_quizzes.size(0) + + nb_correct = quizz_machine.comput_correctness(c_quizzes, models) if args.dirty_debug: nb_correct = torch.randint( - len(models) + 1, nb_correct.size(), device=new_c_quizzes.device + len(models) + 1, nb_correct.size(), device=c_quizzes.device ) - for n in range(nb_correct.max() + 1): - recorded[n].append(new_c_quizzes[nb_correct == n].clone()) + recorded.append((c_quizzes, nb_correct)) nv = F.one_hot(nb_correct, num_classes=len(models) + 1).sum(0) nv = " ".join([str(x.item()) for x in nv]) - log_string(f"keep c_quizzes kept {nv} total {nb_validated()} / {nb_to_create}") + nb_validated = valid_c_quizzes(recorded, standard_validity).size(0) - # concatenate and shuffle - for n in recorded.keys(): - if len(recorded[n]) > 0: - q = torch.cat(recorded[n], dim=0) - q = q[torch.randperm(q.size(0), device=q.device)] - recorded[n] = q - else: - del recorded[n] + log_string(f"keep c_quizzes kept {nv} total {nb_validated} / {nb_to_create}") - new_c_quizzes = torch.cat( - [recorded[n] for n in range(args.min_to_validate, args.max_to_validate + 1)], - dim=0, - ) + # ------------------------------------------------------------ - new_c_quizzes = new_c_quizzes[ - torch.randperm(new_c_quizzes.size(0), device=new_c_quizzes.device)[ - : nb_for_train + nb_for_test - ] - ] + new_c_quizzes = valid_c_quizzes(recorded, standard_validity) quizz_machine.store_c_quizzes(new_c_quizzes[:nb_for_train], for_train=True) quizz_machine.store_c_quizzes(new_c_quizzes[nb_for_train:], for_train=False) - for n in recorded.keys(): + for n in range(len(models) + 1): s = ( "_validated" if n >= args.min_to_validate and n <= args.max_to_validate else "" ) + quizz_machine.problem.save_quizzes( - recorded[n][:72], + valid_c_quizzes(recorded, criteria=lambda nb_correct: nb_correct == n)[:72], args.result_dir, f"culture_c_quiz_{n_epoch:04d}_N{n}{s}", ) @@ -511,57 +487,40 @@ log_string(f"nb_parameters {nb_parameters} ({int(nb_parameters/1e6)}M)") ###################################################################### -min_ave_seq_logproba = None - for n_epoch in range(args.nb_epochs): log_string(f"--- epoch {n_epoch} ----------------------------------------") - a = [(model.id, float(model.main_test_accuracy)) for model in models] - a.sort(key=lambda p: p[0]) - s = " ".join([f"{p[1]*100:.02f}%" for p in a]) - log_string(f"current accuracies {s}") - - # select the model with lowest accuracy - models.sort(key=lambda model: model.main_test_accuracy) - model = models[0] + weakest_model = min(models, key=lambda m: float(m.main_test_accuracy)) log_string( - f"training model {model.id} main_test_accuracy {model.main_test_accuracy}" + f"training model {weakest_model.id} main_test_accuracy {weakest_model.main_test_accuracy}" ) # improve it - one_epoch(model, quizz_machine) - - quizz_machine.renew_w_quizzes(args.nb_train_samples // args.nb_gpts) + one_epoch(weakest_model, quizz_machine) log_string( f"train_set_composition w_quizzes {quizz_machine.nb_batch_w_quizzes} c_quizzes {quizz_machine.nb_batch_c_quizzes}" ) # test it - run_tests(model, quizz_machine, deterministic_synthesis=False) + run_tests(weakest_model, quizz_machine, deterministic_synthesis=False) log_string( f"test_set_composition w_quizzes {quizz_machine.nb_batch_w_quizzes} c_quizzes {quizz_machine.nb_batch_c_quizzes}" ) + # replace a fraction of the w_quizzes with a fresh ones + quizz_machine.renew_w_quizzes(args.nb_train_samples // args.nb_gpts) + if min([m.main_test_accuracy for m in models]) >= args.accuracy_to_make_c_quizzes: - ave_seq_logproba = create_c_quizzes( + create_c_quizzes( models, quizz_machine, nb_for_train=nb_new_c_quizzes_for_train, nb_for_test=nb_new_c_quizzes_for_test, - min_ave_seq_logproba=min_ave_seq_logproba, ) - # We keep the first average logits as a reference - # if min_ave_seq_logproba is None: - # min_ave_seq_logproba = ave_seq_logproba - # else: - # log_string( - # f"min_ave_seq_logproba {min_ave_seq_logproba} ave_seq_logproba {ave_seq_logproba}" - # ) - # We update everyone for model in models: run_tests(model, quizz_machine, deterministic_synthesis=False) diff --git a/quizz_machine.py b/quizz_machine.py index 6f7492d..8dc23a5 100755 --- a/quizz_machine.py +++ b/quizz_machine.py @@ -17,43 +17,6 @@ from mygpt import BracketedSequence ###################################################################### - -class Gang(nn.Module): - def __init__(self, models, nb_models_for_generation, mode="groupthink"): - super().__init__() - self.models = nn.ModuleList(models) - self.nb_models_for_generation = nb_models_for_generation - self.mode = mode - - def forward(self, bs): - # If first = 0, we are re-starting an auto-regressive process, - # that's the right moment to randomize who gonna do it - if bs.first == 0: - self.models_to_use = [ - self.models[k] - for k in torch.randperm(len(self.models))[ - : self.nb_models_for_generation - ] - ] - - all_the_logits = torch.cat( - [model(bs).x[None] for model in self.models_to_use], dim=0 - ) - - if self.mode == "groupthink": - y = all_the_logits.mean(dim=0) - elif self.mode == "groupwork": - m = torch.rand(all_the_logits.size(), device=all_the_logits.device) - m = (m.sort(dim=0).indices == 0).long() - y = (y * m).sum(dim=0) - else: - raise ValueError(f"Invalid mode {self.mode}") - - return BracketedSequence(y, bs.first, bs.nb) - - -###################################################################### - # ar_mask is a tensor with 0s and 1s, of same shape as input, with # 1s where tokens should be generated. The others are kept # unchanged. @@ -374,9 +337,7 @@ class QuizzMachine: ############################################################### - def generate_quizzes( - self, nb, model_for_generation, min_ave_seq_logproba, reverse_cleanup=False - ): + def generate_quizzes(self, nb, model_for_generation, reverse_cleanup=False): c_quizzes = torch.empty( nb, self.train_w_quizzes.size(1), device=self.device, dtype=torch.int64 ) @@ -406,7 +367,6 @@ class QuizzMachine: seq_logproba=seq_logproba, temperature=temperature, deterministic_synthesis=False, - # progress_bar_desc="sampling c_quizzes", device=self.device, ) @@ -422,7 +382,6 @@ class QuizzMachine: seq_logproba=seq_logproba, temperature=temperature, deterministic_synthesis=True, - # progress_bar_desc="sampling c_quizzes", device=self.device, ) @@ -436,56 +395,7 @@ class QuizzMachine: seq_logproba=seq_logproba, temperature=temperature, deterministic_synthesis=True, - # progress_bar_desc="sampling c_quizzes", device=self.device, ) return c_quizzes, seq_logproba.mean() - - ###################################################################### - - def create_c_quizzes( - self, - nb, - model_for_generation, - models_for_validation, - min_ave_seq_logproba, - reverse_cleanup, - n_epoch, - result_dir, - ): - c_quizzes, ave_seq_logproba = self.generate_quizzes( - nb, - model_for_generation=model_for_generation, - min_ave_seq_logproba=min_ave_seq_logproba, - reverse_cleanup=reverse_cleanup, - ) - - nb_correct = self.comput_correctness(c_quizzes, models_for_validation) - - return c_quizzes, nb_correct, ave_seq_logproba - - ###################################################################### - - def gang_create_c_quizzes( - self, - nb, - nb_models_for_generation, - models, - mode, - min_ave_seq_logproba, - reverse_cleanup, - n_epoch, - result_dir, - ): - model_for_generation = Gang(models, nb_models_for_generation, mode) - models_for_validation = models - return self.create_c_quizzes( - nb=nb, - model_for_generation=model_for_generation, - models_for_validation=models_for_validation, - min_ave_seq_logproba=min_ave_seq_logproba, - reverse_cleanup=reverse_cleanup, - n_epoch=n_epoch, - result_dir=result_dir, - )