From b2b3f2097ba123f90920bf5f572085dfb74b719d Mon Sep 17 00:00:00 2001 From: kt Date: Thu, 20 Apr 2023 10:27:36 -0400 Subject: [PATCH 1/2] fix docstring for sequential.py --- espaloma/nn/sequential.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/espaloma/nn/sequential.py b/espaloma/nn/sequential.py index f9cb56ee..4f1ef620 100644 --- a/espaloma/nn/sequential.py +++ b/espaloma/nn/sequential.py @@ -81,7 +81,7 @@ class Sequential(torch.nn.Module): A sequence of numbers (for units) and strings (for activation functions) denoting the configuration of the sequential model. - feature_units : int(default=117) + feature_units : int(default=114) The number of input channels. Methods From 821c5f25087d65b0bc9a223f5c2b493cb9d5db96 Mon Sep 17 00:00:00 2001 From: kt Date: Thu, 20 Apr 2023 12:11:19 -0400 Subject: [PATCH 2/2] fix data split issue (#149) --- espaloma/data/dataset.py | 8 ++++++-- 1 file changed, 6 insertions(+), 2 deletions(-) diff --git a/espaloma/data/dataset.py b/espaloma/data/dataset.py index 4e413c86..ba842611 100644 --- a/espaloma/data/dataset.py +++ b/espaloma/data/dataset.py @@ -186,10 +186,14 @@ def split(self, partition): """ n_data = len(self) - partition = [int(n_data * x / sum(partition)) for x in partition] + p_sizes = [] + for i, _partition in enumerate(partition): + p_size = int((n_data - sum(p_sizes)) * _partition / sum(partition[i:])) + p_sizes.append(p_size) + assert sum(p_sizes) == n_data, f"{p_sizes}, {sum(p_sizes)}" ds = [] idx = 0 - for p_size in partition: + for p_size in p_sizes: ds.append(self[idx : idx + p_size]) idx += p_size