More fix for tiny model creation (#45228)

* add missing VivitImageProcessor

* fix model_type = "evolla"

* change to google/medasr

* fix auto_docstring tue-mps/videomt-dinov2-small-ytvis2019

* fix 3 tester

* fix test

* fix modular

* fix modular

---------

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

Author: ydshieh

src/transformers/models/auto/image_processing_auto.py

@@ -264,6 +264,7 @@
             ("vit_msn", {"torchvision": "ViTImageProcessor", "pil": "ViTImageProcessorPil"}),
             ("vitmatte", {"torchvision": "VitMatteImageProcessor", "pil": "VitMatteImageProcessorPil"}),
             ("vitpose", {"torchvision": "VitPoseImageProcessor", "pil": "VitPoseImageProcessorPil"}),
+            ("vivit", {"torchvision": "VivitImageProcessor"}),
             ("xclip", {"torchvision": "CLIPImageProcessor", "pil": "CLIPImageProcessorPil"}),
             ("yolos", {"torchvision": "YolosImageProcessor", "pil": "YolosImageProcessorPil"}),
             ("zoedepth", {"torchvision": "ZoeDepthImageProcessor", "pil": "ZoeDepthImageProcessorPil"}),

src/transformers/models/evolla/configuration_evolla.py

@@ -96,7 +96,7 @@ class EvollaConfig(PreTrainedConfig):
     >>> configuration = model.config
     ```"""
 
-    model_type = "EvollaModel"
+    model_type = "evolla"
     sub_configs = {"protein_encoder_config": SaProtConfig}
     default_theta = 500000.0
 

src/transformers/models/lasr/configuration_lasr.py

@@ -62,7 +62,7 @@ class LasrEncoderConfig(PreTrainedConfig):
         ```
 
     This configuration class is based on the LasrEncoder architecture from Google Health AI. You can find more details
-    and pre-trained models at [TODO/TODO](https://huggingface.co/TODO/TODO).
+    and pre-trained models at [google/medasr](https://huggingface.co/google/medasr).
     """
 
     model_type = "lasr_encoder"
@@ -121,7 +121,7 @@ class LasrCTCConfig(PreTrainedConfig):
         >>> configuration = model.config
         ```
     This configuration class is based on the Lasr CTC architecture from Google Health AI. You can find more details
-    and pre-trained models at [TODO/TODO](https://huggingface.co/TODO/TODO).
+    and pre-trained models at [google/medasr](https://huggingface.co/google/medasr).
     """
 
     model_type = "lasr_ctc"

src/transformers/models/lasr/modular_lasr.py

@@ -186,7 +186,7 @@ class LasrEncoderConfig(ParakeetEncoderConfig):
         ```
 
     This configuration class is based on the LasrEncoder architecture from Google Health AI. You can find more details
-    and pre-trained models at [TODO/TODO](https://huggingface.co/TODO/TODO).
+    and pre-trained models at [google/medasr](https://huggingface.co/google/medasr).
     """
 
     hidden_size: int = 512
@@ -231,7 +231,7 @@ class LasrCTCConfig(ParakeetCTCConfig):
         >>> configuration = model.config
         ```
     This configuration class is based on the Lasr CTC architecture from Google Health AI. You can find more details
-    and pre-trained models at [TODO/TODO](https://huggingface.co/TODO/TODO).
+    and pre-trained models at [google/medasr](https://huggingface.co/google/medasr).
     """
 
     vocab_size: int = 512

src/transformers/models/videomt/configuration_videomt.py

@@ -24,7 +24,7 @@
 from ...utils import auto_docstring
 
 
-@auto_docstring(checkpoint="tue-mps/coco_panoptic_videomt_large_640")
+@auto_docstring(checkpoint="tue-mps/videomt-dinov2-small-ytvis2019")
 @strict
 class VideomtConfig(PreTrainedConfig):
     r"""

src/transformers/models/videomt/modular_videomt.py

@@ -15,6 +15,7 @@
 from dataclasses import dataclass
 
 import torch
+from huggingface_hub.dataclasses import strict
 from torch import nn
 
 from ...file_utils import ModelOutput
@@ -36,6 +37,8 @@
 )
 
 
+@auto_docstring(checkpoint="tue-mps/videomt-dinov2-small-ytvis2019")
+@strict
 class VideomtConfig(EomtConfig):
     model_type = "videomt"
 

tests/models/pp_ocrv5_mobile_rec/test_modeling_pp_ocrv5_mobile_rec.py

@@ -51,6 +51,7 @@
 class PPOCRV5MobileRecModelTester:
     def __init__(
         self,
+        parent,
         batch_size=3,
         image_size=[48, 320],
         num_channels=3,
@@ -66,6 +67,7 @@ def __init__(
         num_stages=5,
         attention_dropout=0.0,
     ):
+        self.parent = parent
         self.batch_size = batch_size
         self.num_channels = num_channels
         self.image_size = image_size
@@ -142,8 +144,7 @@ class PPOCRV5MobileRecModelTest(ModelTesterMixin, PipelineTesterMixin, unittest.
     model_split_percents = [0.5, 0.8]
 
     def setUp(self):
-        self.model_tester = PPOCRV5MobileRecModelTester()
-        self.model_tester.parent = self
+        self.model_tester = PPOCRV5MobileRecModelTester(self)
         self.config_tester = ConfigTester(
             self,
             config_class=PPOCRV5MobileRecConfig,

tests/models/pp_ocrv5_server_rec/test_modeling_pp_ocrv5_server_rec.py

@@ -51,6 +51,7 @@
 class PPOCRV5ServerRecModelTester:
     def __init__(
         self,
+        parent,
         batch_size=3,
         image_size=[48, 320],
         num_channels=3,
@@ -65,6 +66,7 @@ def __init__(
         num_attention_heads=2,
         attention_dropout=0.0,
     ):
+        self.parent = parent
         self.batch_size = batch_size
         self.num_channels = num_channels
         self.image_size = image_size
@@ -138,8 +140,7 @@ class PPOCRV5ServerRecModelTest(ModelTesterMixin, PipelineTesterMixin, unittest.
     model_split_percents = [0.5, 0.8]
 
     def setUp(self):
-        self.model_tester = PPOCRV5ServerRecModelTester()
-        self.model_tester.parent = self
+        self.model_tester = PPOCRV5ServerRecModelTester(self)
         self.config_tester = ConfigTester(
             self,
             config_class=PPOCRV5ServerRecConfig,

tests/models/slanext/test_modeling_slanext.py

@@ -53,12 +53,14 @@
 class SLANeXtModelTester:
     def __init__(
         self,
+        parent,
         batch_size=2,
         image_size=512,
         num_channels=3,
         is_training=False,
         vision_config=None,
     ):
+        self.parent = parent
         if vision_config is None:
             vision_config = {
                 "hidden_size": 1,
@@ -106,10 +108,10 @@ class SLANeXtModelTest(ModelTesterMixin, PipelineTesterMixin, unittest.TestCase)
 
     def setUp(self):
         self.model_tester = SLANeXtModelTester(
+            self,
             batch_size=1,
             image_size=512,
         )
-        self.model_tester.parent = self
         self.config_tester = ConfigTester(
             self,
             config_class=SLANeXtConfig,

tests/test_modeling_common.py

@@ -3661,7 +3661,7 @@ def test_sdpa_can_dispatch_on_flash(self):
                     "PaliGemma-like models currently (transformers==4.41.0) requires an attention_mask input"
                 )
             if config.model_type in [
-                "EvollaModel",
+                "evolla",
                 "modernbert",
                 "gemma3",
                 "t5gemma",