From 9eadd2cd6913a0de53b4b0f526157497e8d14381 Mon Sep 17 00:00:00 2001 From: =?utf8?q?Fran=C3=A7ois=20Fleuret?= Date: Sun, 24 Mar 2024 23:11:55 +0100 Subject: [PATCH] Update. --- tasks.py | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/tasks.py b/tasks.py index 56c2b0f..fddcaff 100755 --- a/tasks.py +++ b/tasks.py @@ -1939,7 +1939,7 @@ class Escape(Task): ): # Put the lookahead reward to either 0 or -1 for the # current iteration, sample the next state - s = -1 # (torch.rand(result.size(0), device = result.device) < 0.2).long() + s = -(torch.rand(result.size(0), device=result.device) < 0.2).long() result[:, u - 1] = s + 1 + escape.first_lookahead_rewards_code ar_mask = (t >= u).long() * (t < u + state_len).long() ar(result, ar_mask) @@ -1980,7 +1980,7 @@ class Escape(Task): def produce_results( self, n_epoch, model, result_dir, logger, deterministic_synthesis, nmax=1000 ): - result = self.test_input[:100].clone() + result = self.test_input[:250].clone() # Saving the ground truth -- 2.39.5