CI with `num_hidden_layers=2` 🚀🚀🚀 (#25266)

* CI with layers=2

---------
Co-authored-by: ydshieh <ydshieh@users.noreply.github.com>

tests/models/roberta/test_modeling_flax_roberta.py

@@ -46,7 +46,7 @@ class FlaxRobertaModelTester(unittest.TestCase):
         use_labels=True,
         vocab_size=99,
         hidden_size=32,
-        num_hidden_layers=5,
+        num_hidden_layers=2,
         num_attention_heads=4,
         intermediate_size=37,
         hidden_act="gelu",

tests/models/roberta/test_modeling_roberta.py

@@ -58,7 +58,7 @@ class RobertaModelTester:
         use_labels=True,
         vocab_size=99,
         hidden_size=32,
-        num_hidden_layers=5,
+        num_hidden_layers=2,
         num_attention_heads=4,
         intermediate_size=37,
         hidden_act="gelu",

tests/models/roberta_prelayernorm/test_modeling_flax_roberta_prelayernorm.py

@@ -49,7 +49,7 @@ class FlaxRobertaPreLayerNormModelTester(unittest.TestCase):
         use_labels=True,
         vocab_size=99,
         hidden_size=32,
-        num_hidden_layers=5,
+        num_hidden_layers=2,
         num_attention_heads=4,
         intermediate_size=37,
         hidden_act="gelu",

tests/models/roberta_prelayernorm/test_modeling_roberta_prelayernorm.py

@@ -57,7 +57,7 @@ class RobertaPreLayerNormModelTester:
         use_labels=True,
         vocab_size=99,
         hidden_size=32,
-        num_hidden_layers=5,
+        num_hidden_layers=2,
         num_attention_heads=4,
         intermediate_size=37,
         hidden_act="gelu",

tests/models/roc_bert/test_modeling_roc_bert.py

@@ -58,7 +58,7 @@ class RoCBertModelTester:
         pronunciation_embed_dim=32,
         shape_embed_dim=32,
         hidden_size=32,
-        num_hidden_layers=5,
+        num_hidden_layers=2,
         num_attention_heads=4,
         intermediate_size=37,
         hidden_act="gelu",

tests/models/roformer/test_modeling_flax_roformer.py

@@ -47,7 +47,7 @@ class FlaxRoFormerModelTester(unittest.TestCase):
         use_labels=True,
         vocab_size=99,
         hidden_size=32,
-        num_hidden_layers=5,
+        num_hidden_layers=2,
         num_attention_heads=4,
         intermediate_size=37,
         hidden_act="gelu",

tests/models/roformer/test_modeling_roformer.py

@@ -56,7 +56,7 @@ class RoFormerModelTester:
         use_labels=True,
         vocab_size=99,
         hidden_size=32,
-        num_hidden_layers=5,
+        num_hidden_layers=2,
         num_attention_heads=4,
         intermediate_size=37,
         hidden_act="gelu",

tests/models/rwkv/test_modeling_rwkv.py

@@ -52,7 +52,7 @@ class RwkvModelTester:
         use_mc_token_ids=True,
         vocab_size=99,
         hidden_size=32,
-        num_hidden_layers=5,
+        num_hidden_layers=2,
         intermediate_size=37,
         hidden_act="gelu",
         hidden_dropout_prob=0.1,

tests/models/sew/test_modeling_sew.py

@@ -65,7 +65,7 @@ class SEWModelTester:
         num_conv_pos_embeddings=31,
         num_conv_pos_embedding_groups=2,
         squeeze_factor=2,
-        num_hidden_layers=4,
+        num_hidden_layers=2,
         num_attention_heads=2,
         hidden_dropout=0.1,
         intermediate_size=20,

tests/models/sew_d/test_modeling_sew_d.py

@@ -72,7 +72,7 @@ class SEWDModelTester:
         position_biased_input=False,
         pos_att_type=("p2c", "c2p"),
         norm_rel_ebd="layer_norm",
-        num_hidden_layers=4,
+        num_hidden_layers=2,
         num_attention_heads=2,
         hidden_dropout=0.1,
         intermediate_size=20,

tests/models/speech_to_text_2/test_modeling_speech_to_text_2.py

@@ -50,7 +50,7 @@ class Speech2Text2StandaloneDecoderModelTester:
         use_labels=True,
         decoder_start_token_id=2,
         decoder_ffn_dim=32,
-        decoder_layers=4,
+        decoder_layers=2,
         decoder_attention_heads=4,
         max_position_embeddings=30,
         pad_token_id=0,

tests/models/speecht5/test_modeling_speecht5.py

@@ -105,7 +105,7 @@ class SpeechT5ModelTester:
         is_training=False,
         vocab_size=81,
         hidden_size=24,
-        num_hidden_layers=4,
+        num_hidden_layers=2,
         num_attention_heads=2,
         intermediate_size=4,
     ):
@@ -249,7 +249,7 @@ class SpeechT5ForSpeechToTextTester:
         decoder_seq_length=7,
         is_training=False,
         hidden_size=24,
-        num_hidden_layers=4,
+        num_hidden_layers=2,
         num_attention_heads=2,
         intermediate_size=4,
         conv_dim=(32, 32, 32),
@@ -786,7 +786,7 @@ class SpeechT5ForTextToSpeechTester:
         decoder_seq_length=1024,  # speech is longer
         is_training=False,
         hidden_size=24,
-        num_hidden_layers=4,
+        num_hidden_layers=2,
         num_attention_heads=2,
         intermediate_size=4,
         vocab_size=81,
@@ -1031,7 +1031,7 @@ class SpeechT5ForSpeechToSpeechTester:
         decoder_seq_length=1024,
         is_training=False,
         hidden_size=24,
-        num_hidden_layers=4,
+        num_hidden_layers=2,
         num_attention_heads=2,
         intermediate_size=4,
         conv_dim=(32, 32, 32),

tests/models/splinter/test_modeling_splinter.py

@@ -46,7 +46,7 @@ class SplinterModelTester:
         vocab_size=99,
         hidden_size=32,
         question_token_id=1,
-        num_hidden_layers=5,
+        num_hidden_layers=2,
         num_attention_heads=4,
         intermediate_size=37,
         hidden_act="gelu",

tests/models/squeezebert/test_modeling_squeezebert.py

@@ -50,7 +50,7 @@ class SqueezeBertModelTester(object):
         use_labels=True,
         vocab_size=99,
         hidden_size=32,
-        num_hidden_layers=5,
+        num_hidden_layers=2,
         num_attention_heads=4,
         intermediate_size=64,
         hidden_act="gelu",

tests/models/switch_transformers/test_modeling_switch_transformers.py

@@ -58,7 +58,7 @@ class SwitchTransformersModelTester:
         use_attention_mask=True,
         use_labels=True,
         hidden_size=32,
-        num_hidden_layers=5,
+        num_hidden_layers=2,
         num_attention_heads=4,
         d_ff=37,
         relative_attention_num_buckets=8,
@@ -826,7 +826,7 @@ class SwitchTransformersEncoderOnlyModelTester:
         # For common tests
         use_attention_mask=True,
         hidden_size=32,
-        num_hidden_layers=5,
+        num_hidden_layers=2,
         num_attention_heads=4,
         d_ff=37,
         relative_attention_num_buckets=8,

tests/models/t5/test_modeling_flax_t5.py

@@ -70,7 +70,7 @@ class FlaxT5ModelTester:
         use_attention_mask=True,
         use_labels=True,
         hidden_size=32,
-        num_hidden_layers=5,
+        num_hidden_layers=2,
         num_attention_heads=4,
         d_ff=37,
         relative_attention_num_buckets=8,
@@ -477,7 +477,7 @@ class FlaxT5EncoderOnlyModelTester:
         use_attention_mask=True,
         use_labels=True,
         hidden_size=32,
-        num_hidden_layers=5,
+        num_hidden_layers=2,
         num_attention_heads=4,
         d_ff=37,
         relative_attention_num_buckets=8,

tests/models/t5/test_modeling_t5.py

@@ -71,7 +71,7 @@ class T5ModelTester:
         use_attention_mask=True,
         use_labels=True,
         hidden_size=32,
-        num_hidden_layers=5,
+        num_hidden_layers=2,
         num_attention_heads=4,
         d_ff=37,
         relative_attention_num_buckets=8,
@@ -902,7 +902,7 @@ class T5EncoderOnlyModelTester:
         # For common tests
         use_attention_mask=True,
         hidden_size=32,
-        num_hidden_layers=5,
+        num_hidden_layers=2,
         num_attention_heads=4,
         d_ff=37,
         relative_attention_num_buckets=8,

tests/models/tapas/test_modeling_tapas.py

@@ -79,7 +79,7 @@ class TapasModelTester:
         use_labels=True,
         vocab_size=99,
         hidden_size=32,
-        num_hidden_layers=5,
+        num_hidden_layers=2,
         num_attention_heads=4,
         intermediate_size=37,
         hidden_act="gelu",

tests/models/timesformer/test_modeling_timesformer.py

@@ -60,7 +60,7 @@ class TimesformerModelTester:
         is_training=True,
         use_labels=True,
         hidden_size=32,
-        num_hidden_layers=5,
+        num_hidden_layers=2,
         num_attention_heads=4,
         intermediate_size=37,
         hidden_act="gelu",

tests/models/transfo_xl/test_modeling_transfo_xl.py

@@ -52,7 +52,7 @@ class TransfoXLModelTester:
         d_head=8,
         d_inner=128,
         div_val=2,
-        num_hidden_layers=5,
+        num_hidden_layers=2,
         scope=None,
         seed=1,
         eos_token_id=0,