SUMIN/HuggingFace_transformer
1
0
Fork 0
Code Issues Pull Requests Actions 7 Packages Projects Releases Wiki Activity

Speed up TF tests by reducing hidden layer counts (#24595)

Browse Source 
* hidden layers, huh, what are they good for (absolutely nothing)

* Some tests break with 1 hidden layer, use 2

* Use 1 hidden layer in a few slow models

* Use num_hidden_layers=2 everywhere

* Slightly higher tol for groupvit

* Slightly higher tol for groupvit
This commit is contained in:
Matt
2023-06-30 16:30:33 +01:00
committed by GitHub
parent 3441ad7d43
commit 134caef31a
47 changed files with 67 additions and 85 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
tests/models/blip/test_modeling_tf_blip.py
View File

@@ -64,7 +64,7 @@ class TFBlipVisionModelTester:
is_training=True,
hidden_size=32,
projection_dim=32,
num_hidden_layers=5,
num_hidden_layers=2,
num_attention_heads=4,
intermediate_size=37,
dropout=0.1,
@@ -207,7 +207,7 @@ class TFBlipTextModelTester:
vocab_size=99,
hidden_size=32,
projection_dim=32,
num_hidden_layers=5,
num_hidden_layers=2,
num_attention_heads=4,
intermediate_size=37,
dropout=0.1,