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

* CI with layers=2

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

tests/models/mobilebert/test_modeling_mobilebert.py

@@ -54,7 +54,7 @@ class MobileBertModelTester:
         vocab_size=99,
         hidden_size=64,
         embedding_size=32,
-        num_hidden_layers=5,
+        num_hidden_layers=2,
         num_attention_heads=4,
         intermediate_size=37,
         hidden_act="gelu",

tests/models/mpnet/test_modeling_mpnet.py

@@ -49,7 +49,7 @@ class MPNetModelTester:
         use_labels=True,
         vocab_size=99,
         hidden_size=64,
-        num_hidden_layers=5,
+        num_hidden_layers=2,
         num_attention_heads=4,
         intermediate_size=64,
         hidden_act="gelu",

tests/models/mpt/test_modeling_mpt.py

@@ -54,7 +54,7 @@ class MptModelTester:
         use_mc_token_ids=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/mra/test_modeling_mra.py

@@ -51,7 +51,7 @@ class MraModelTester:
         use_labels=True,
         vocab_size=99,
         hidden_size=16,
-        num_hidden_layers=5,
+        num_hidden_layers=2,
         num_attention_heads=2,
         intermediate_size=36,
         hidden_act="gelu",

tests/models/mvp/test_modeling_mvp.py

@@ -595,7 +595,7 @@ class MvpStandaloneDecoderModelTester:
         use_labels=True,
         decoder_start_token_id=2,
         decoder_ffn_dim=32,
-        decoder_layers=4,
+        decoder_layers=2,
         encoder_attention_heads=4,
         decoder_attention_heads=4,
         max_position_embeddings=30,

tests/models/nezha/test_modeling_nezha.py

@@ -55,7 +55,7 @@ class NezhaModelTester:
         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/nllb_moe/test_modeling_nllb_moe.py

@@ -52,7 +52,7 @@ class NllbMoeModelTester:
         use_labels=False,
         vocab_size=99,
         hidden_size=16,
-        num_hidden_layers=4,
+        num_hidden_layers=2,
         num_attention_heads=4,
         intermediate_size=4,
         hidden_act="relu",

tests/models/nystromformer/test_modeling_nystromformer.py

@@ -51,7 +51,7 @@ class NystromformerModelTester:
         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/openai/test_modeling_openai.py

@@ -49,7 +49,7 @@ class OpenAIGPTModelTester:
         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/opt/test_modeling_opt.py

@@ -70,7 +70,7 @@ class OPTModelTester:
         use_labels=False,
         vocab_size=99,
         hidden_size=16,
-        num_hidden_layers=5,
+        num_hidden_layers=2,
         num_attention_heads=4,
         intermediate_size=4,
         hidden_act="gelu",

tests/models/owlvit/test_modeling_owlvit.py

@@ -62,7 +62,7 @@ class OwlViTVisionModelTester:
         num_channels=3,
         is_training=True,
         hidden_size=32,
-        num_hidden_layers=5,
+        num_hidden_layers=2,
         num_attention_heads=4,
         intermediate_size=37,
         dropout=0.1,

tests/models/pegasus/test_modeling_flax_pegasus.py

@@ -52,7 +52,7 @@ class FlaxPegasusModelTester:
         use_labels=False,
         vocab_size=99,
         hidden_size=32,
-        num_hidden_layers=5,
+        num_hidden_layers=2,
         num_attention_heads=4,
         intermediate_size=37,
         hidden_dropout_prob=0.1,

tests/models/pegasus/test_modeling_pegasus.py

@@ -371,7 +371,7 @@ class PegasusStandaloneDecoderModelTester:
         use_labels=True,
         decoder_start_token_id=2,
         decoder_ffn_dim=32,
-        decoder_layers=4,
+        decoder_layers=2,
         encoder_attention_heads=4,
         decoder_attention_heads=4,
         max_position_embeddings=30,

tests/models/pegasus_x/test_modeling_pegasus_x.py

@@ -670,7 +670,7 @@ class PegasusXStandaloneDecoderModelTester:
         use_labels=True,
         decoder_start_token_id=2,
         decoder_ffn_dim=32,
-        decoder_layers=4,
+        decoder_layers=2,
         encoder_attention_heads=4,
         decoder_attention_heads=4,
         max_position_embeddings=30,

tests/models/pix2struct/test_modeling_pix2struct.py

@@ -71,7 +71,7 @@ class Pix2StructVisionModelTester:
         patch_embed_hidden_size=12,
         projection_dim=32,
         max_patches=64,
-        num_hidden_layers=5,
+        num_hidden_layers=2,
         num_attention_heads=4,
         intermediate_size=37,
         dropout=0.1,
@@ -230,7 +230,7 @@ class Pix2StructTextModelTester:
         vocab_size=99,
         hidden_size=12,
         projection_dim=32,
-        num_hidden_layers=5,
+        num_hidden_layers=2,
         num_attention_heads=4,
         intermediate_size=37,
         dropout=0.1,

tests/models/plbart/test_modeling_plbart.py

@@ -473,7 +473,7 @@ class PLBartStandaloneDecoderModelTester:
         use_labels=True,
         decoder_start_token_id=2,
         decoder_ffn_dim=32,
-        decoder_layers=4,
+        decoder_layers=2,
         encoder_attention_heads=4,
         decoder_attention_heads=4,
         max_position_embeddings=30,

tests/models/prophetnet/test_modeling_prophetnet.py

@@ -55,10 +55,10 @@ class ProphetNetModelTester:
         use_labels=True,
         decoder_start_token_id=0,
         encoder_ffn_dim=32,
-        num_encoder_layers=4,
+        num_encoder_layers=2,
         num_encoder_attention_heads=4,
         decoder_ffn_dim=32,
-        num_decoder_layers=4,
+        num_decoder_layers=2,
         num_decoder_attention_heads=4,
         max_position_embeddings=30,
         is_encoder_decoder=True,
@@ -437,10 +437,10 @@ class ProphetNetModelTester:
                 decoder_attention_mask=decoder_attention_mask,
                 labels=lm_labels,
             )
-        self.parent.assertTrue(torch.allclose(result.loss, torch.tensor(4.5981, device=torch_device), atol=1e-3))
+        self.parent.assertTrue(torch.allclose(result.loss, torch.tensor(4.5892, device=torch_device), atol=1e-3))
 
         expected_logit_slice = torch.tensor(
-            [-0.0648, 0.0790, 0.0360, 0.0089, 0.0039, -0.0639, 0.0131], device=torch_device
+            [-0.0184, 0.0758, -0.0543, -0.0093, 0.0050, -0.0660, -0.1453], device=torch_device
         )
         self.parent.assertTrue(torch.allclose(result.logits[0, :, 1], expected_logit_slice, atol=1e-3))
 
@@ -551,10 +551,10 @@ class ProphetNetStandaloneDecoderModelTester:
         use_labels=True,
         decoder_start_token_id=0,
         encoder_ffn_dim=32,
-        num_encoder_layers=4,
+        num_encoder_layers=2,
         num_encoder_attention_heads=4,
         decoder_ffn_dim=32,
-        num_decoder_layers=4,
+        num_decoder_layers=2,
         num_decoder_attention_heads=4,
         max_position_embeddings=30,
         is_encoder_decoder=False,
@@ -782,10 +782,10 @@ class ProphetNetStandaloneEncoderModelTester:
         use_labels=True,
         decoder_start_token_id=0,
         encoder_ffn_dim=32,
-        num_encoder_layers=4,
+        num_encoder_layers=2,
         num_encoder_attention_heads=4,
         decoder_ffn_dim=32,
-        num_decoder_layers=4,
+        num_decoder_layers=2,
         num_decoder_attention_heads=4,
         max_position_embeddings=30,
         is_encoder_decoder=False,

tests/models/qdqbert/test_modeling_qdqbert.py

@@ -54,7 +54,7 @@ class QDQBertModelTester:
         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/realm/test_modeling_realm.py

@@ -54,7 +54,7 @@ class RealmModelTester:
         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/rembert/test_modeling_rembert.py

@@ -55,7 +55,7 @@ class RemBertModelTester:
         hidden_size=32,
         input_embedding_size=18,
         output_embedding_size=43,
-        num_hidden_layers=5,
+        num_hidden_layers=2,
         num_attention_heads=4,
         intermediate_size=37,
         hidden_act="gelu",