v0.4.0r2

Remove Incomplete Model Folder

Signed-off-by: QAIHM Team <[email protected]>

qai_hub_models/models/_shared/ffnet/model.py

@@ -6,6 +6,7 @@
 
 import os
 from importlib import reload
+from typing import Type, TypeVar
 
 import torch
 
@@ -58,12 +59,14 @@
     ),
 }
 
+FFNetType = TypeVar("FFNetType", bound="FFNet")
+
 
 class FFNet(CityscapesSegmentor):
     """Exportable FFNet fuss-free Cityscapes segmentation model."""
 
     @classmethod
-    def from_pretrained(cls, variant_name: str) -> FFNet:
+    def from_pretrained(cls: Type[FFNetType], variant_name: str) -> FFNetType:
         model = _load_ffnet_source_model(variant_name)
         model.eval()
 

qai_hub_models/models/_shared/ffnet/test_utils.py

@@ -4,19 +4,22 @@
 # ---------------------------------------------------------------------
 from __future__ import annotations
 
+from typing import Type
+
 import numpy as np
 import torch
 
 from qai_hub_models.models._shared.cityscapes_segmentation.demo import (
     TEST_CITYSCAPES_LIKE_IMAGE_ASSET,
 )
-from qai_hub_models.models._shared.ffnet.model import FFNet, _load_ffnet_source_model
+from qai_hub_models.models._shared.ffnet.model import _load_ffnet_source_model
 from qai_hub_models.utils.asset_loaders import load_image
+from qai_hub_models.utils.base_model import BaseModel
 from qai_hub_models.utils.image_processing import preprocess_PIL_image
 
 
 def run_test_off_target_numerical(
-    ffnet_cls: FFNet, variant_name: str, relax_numerics: bool = False
+    ffnet_cls: Type[BaseModel], variant_name: str, relax_numerics: bool = False
 ):
     """Verify that raw (numeric) outputs of both (qaism and non-qaism) networks are the same."""
     processed_sample_image = preprocess_PIL_image(

qai_hub_models/models/_shared/ffnet_quantized/model.py

@@ -13,6 +13,7 @@
 # isort: on
 
 import os
+from typing import Type, TypeVar
 
 import torch
 from aimet_torch.batch_norm_fold import fold_all_batch_norms
@@ -29,6 +30,8 @@
     os.path.join(os.path.dirname(__file__), "aimet_config.json")
 )
 
+FFNetQuantizableType = TypeVar("FFNetQuantizableType", bound="FFNetQuantizable")
+
 
 class FFNetQuantizable(AIMETQuantizableMixin, FFNet):
     """
@@ -58,10 +61,10 @@ def default_aimet_encodings(cls) -> str:
 
     @classmethod
     def from_pretrained(
-        cls,
+        cls: Type[FFNetQuantizableType],
         variant_name: str,
         aimet_encodings: str | None = "DEFAULT",
-    ) -> "FFNetQuantizable":
+    ) -> FFNetQuantizableType:
         ffnet = FFNet.from_pretrained(variant_name).model
 
         input_shape = FFNetQuantizable.get_input_spec()["image"][0]

qai_hub_models/models/_shared/whisper/model.py

@@ -221,7 +221,7 @@ def get_input_spec(num_blocks: int, attention_dim: int) -> InputSpec:
 
         return specs
 
-    def _get_input_spec_for_model_instance(self) -> InputSpec:
+    def _get_input_spec_for_instance(self) -> InputSpec:
         return self.__class__.get_input_spec(len(self.blocks), self.attention_dim)
 
     @classmethod

qai_hub_models/models/controlnet_quantized/info.yaml

@@ -17,6 +17,7 @@ deploy_license: https://github.com/lllyasviel/ControlNet/blob/main/LICENSE
 source_repo: https://github.com/lllyasviel/ControlNet
 technical_details:
   Input: Text prompt and input image as a reference
+  Conditioning Input: Canny-Edge
   QNN-SDK: '2.19'
   Text Encoder Number of parameters: 340M
   UNet Number of parameters: 865M

qai_hub_models/models/ffnet_122ns_lowres/model.py

@@ -11,7 +11,7 @@
 
 class FFNet122NSLowRes(FFNetLowRes):
     @classmethod
-    def from_pretrained(cls) -> FFNet122NSLowRes:
-        return FFNetLowRes.from_pretrained.__func__(
-            cls, "segmentation_ffnet122NS_CCC_mobile_pre_down"
+    def from_pretrained(cls) -> FFNet122NSLowRes:  # type: ignore
+        return super(FFNet122NSLowRes, cls).from_pretrained(
+            "segmentation_ffnet122NS_CCC_mobile_pre_down"
         )

qai_hub_models/models/ffnet_40s/model.py

@@ -11,5 +11,5 @@
 
 class FFNet40S(FFNet):
     @classmethod
-    def from_pretrained(cls) -> FFNet40S:
-        return FFNet.from_pretrained.__func__(cls, "segmentation_ffnet40S_dBBB_mobile")
+    def from_pretrained(cls) -> FFNet40S:  # type: ignore
+        return super(FFNet40S, cls).from_pretrained("segmentation_ffnet40S_dBBB_mobile")

qai_hub_models/models/ffnet_40s_quantized/model.py

@@ -15,12 +15,11 @@
 
 class FFNet40SQuantizable(FFNetQuantizable, FFNet40S):
     @classmethod
-    def from_pretrained(
+    def from_pretrained(  # type: ignore
         cls,
         aimet_encodings: str | None = "DEFAULT",
     ) -> FFNet40SQuantizable:
-        return FFNetQuantizable.from_pretrained.__func__(
-            cls,
+        return super(FFNet40SQuantizable, cls).from_pretrained(
             "segmentation_ffnet40S_dBBB_mobile",
             aimet_encodings=aimet_encodings,
         )

qai_hub_models/models/ffnet_54s/model.py

@@ -11,5 +11,5 @@
 
 class FFNet54S(FFNet):
     @classmethod
-    def from_pretrained(cls) -> FFNet54S:
-        return FFNet.from_pretrained.__func__(cls, "segmentation_ffnet54S_dBBB_mobile")
+    def from_pretrained(cls) -> FFNet54S:  # type: ignore
+        return super(FFNet54S, cls).from_pretrained("segmentation_ffnet54S_dBBB_mobile")

qai_hub_models/models/ffnet_54s_quantized/model.py

@@ -15,12 +15,12 @@
 
 class FFNet54SQuantizable(FFNetQuantizable, FFNet54S):
     @classmethod
-    def from_pretrained(
+    def from_pretrained(  # type: ignore
         cls,
         aimet_encodings: str | None = "DEFAULT",
     ) -> FFNet54SQuantizable:
-        return FFNetQuantizable.from_pretrained.__func__(
-            cls, "segmentation_ffnet54S_dBBB_mobile", aimet_encodings=aimet_encodings
+        return super(FFNet54SQuantizable, cls).from_pretrained(
+            "segmentation_ffnet54S_dBBB_mobile", aimet_encodings=aimet_encodings
         )
 
     @classmethod

qai_hub_models/models/ffnet_78s/model.py

@@ -11,5 +11,5 @@
 
 class FFNet78S(FFNet):
     @classmethod
-    def from_pretrained(cls) -> FFNet78S:
-        return FFNet.from_pretrained.__func__(cls, "segmentation_ffnet78S_dBBB_mobile")
+    def from_pretrained(cls) -> FFNet78S:  # type: ignore
+        return super(FFNet78S, cls).from_pretrained("segmentation_ffnet78S_dBBB_mobile")

qai_hub_models/models/ffnet_78s_lowres/model.py

@@ -11,7 +11,7 @@
 
 class FFNet78SLowRes(FFNetLowRes):
     @classmethod
-    def from_pretrained(cls) -> FFNet78SLowRes:
-        return FFNetLowRes.from_pretrained.__func__(
-            cls, "segmentation_ffnet78S_BCC_mobile_pre_down"
+    def from_pretrained(cls) -> FFNet78SLowRes:  # type: ignore
+        return super(FFNet78SLowRes, cls).from_pretrained(
+            "segmentation_ffnet78S_BCC_mobile_pre_down"
         )

qai_hub_models/models/ffnet_78s_quantized/model.py

@@ -15,12 +15,12 @@
 
 class FFNet78SQuantizable(FFNetQuantizable, FFNet78S):
     @classmethod
-    def from_pretrained(
+    def from_pretrained(  # type: ignore
         cls,
         aimet_encodings: str | None = "DEFAULT",
     ) -> FFNet78SQuantizable:
-        return FFNetQuantizable.from_pretrained.__func__(
-            cls, "segmentation_ffnet78S_dBBB_mobile", aimet_encodings=aimet_encodings
+        return super(FFNet78SQuantizable, cls).from_pretrained(
+            "segmentation_ffnet78S_dBBB_mobile", aimet_encodings=aimet_encodings
         )
 
     @classmethod

qai_hub_models/models/sam/model.py

@@ -117,7 +117,7 @@ def forward(self, image: torch.Tensor) -> torch.Tensor:
         """
         return self.sam.image_encoder(image)
 
-    def _get_input_spec_for_model_instance(
+    def _get_input_spec_for_instance(
         self,
         batch_size: int = 1,
         num_channels: int = 3,
@@ -227,7 +227,7 @@ def forward(
             image_embeddings, point_coords, point_labels, mask_input, has_mask_input
         )
 
-    def _get_input_spec_for_model_instance(
+    def _get_input_spec_for_instance(
         self,
         num_of_points: int = 1,
     ) -> InputSpec:

qai_hub_models/models/stylegan2/model.py

@@ -95,7 +95,7 @@ def get_input_spec(
             inputs["classes"] = ((batch_size, num_classes), "float32")
         return inputs
 
-    def _get_input_spec_for_model_instance(self, batch_size: int = 1) -> InputSpec:
+    def _get_input_spec_for_instance(self, batch_size: int = 1) -> InputSpec:
         return self.__class__.get_input_spec(
             self.output_size, self.num_classes, batch_size
         )
@@ -104,7 +104,7 @@ def sample_inputs(
         self, input_spec: InputSpec | None = None, seed=None
     ) -> Dict[str, List[np.ndarray]]:
         if not input_spec:
-            input_spec = self._get_input_spec_for_model_instance()
+            input_spec = self._get_input_spec_for_instance()
 
         inputs = {
             "image_noise": [

qai_hub_models/models/trocr/model.py

@@ -256,7 +256,7 @@ def get_input_spec(
 
         return decoder_input_specs
 
-    def _get_input_spec_for_model_instance(self) -> InputSpec:
+    def _get_input_spec_for_instance(self) -> InputSpec:
         return self.__class__.get_input_spec(
             self.decoder_attention_heads,
             self.embeddings_per_head,

qai_hub_models/test/test_async_compile_jobs.py

@@ -5,7 +5,8 @@
 import os
 
 import qai_hub as hub
-import yaml
+
+from qai_hub_models.utils.asset_loaders import load_yaml
 
 
 def test_compile_jobs_success():
@@ -16,8 +17,7 @@ def test_compile_jobs_success():
     """
     if os.stat(os.environ["COMPILE_JOBS_FILE"]).st_size == 0:
         return
-    with open(os.environ["COMPILE_JOBS_FILE"], "r") as f:
-        job_ids = yaml.safe_load(f.read())
+    job_ids = load_yaml(os.environ["COMPILE_JOBS_FILE"])
     failed_jobs = {}
     for name, job_id in job_ids.items():
         result = hub.get_job(job_id).wait()