gradio app. models refactor

README.md

@@ -3,3 +3,15 @@
 Train model for birds classification and gradio app
 
 Training is done using fastai, deployment mimics its transforms to publish a gradio app that has no fastai dependencies.
+
+## Train
+
+```bash
+conda env create -f environment.yml
+```
+
+```bash
+conda activate fastai
+cd training
+python -m birds.train
+```

deployment/app.py

@@ -0,0 +1,37 @@
+import torch
+import gradio as gr
+from PIL import Image
+from model import get_model, apply_weights, copy_weight
+from vocab import vocab
+from transforms import resized_crop_pad, gpu_crop
+from torchvision.transforms import Normalize, ToTensor
+
+model = get_model()
+state = torch.load("../models/vit_saved.pth", map_location="cpu")
+apply_weights(model, state, copy_weight)
+
+to_tensor = ToTensor()
+norm = Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])
+
+
+def classify_image(inp):
+    inp = Image.fromarray(inp)
+    transformed_input = resized_crop_pad(inp, (460, 460))
+    transformed_input = to_tensor(transformed_input).unsqueeze(0)
+    transformed_input = gpu_crop(transformed_input, (224, 224))
+    transformed_input = norm(transformed_input)
+    model.eval()
+    with torch.no_grad():
+        pred = model(transformed_input)
+    pred = torch.argmax(pred, dim=1)
+    return vocab[pred]
+
+
+iface = gr.Interface(
+    fn=classify_image,
+    inputs=gr.inputs.Image(),
+    outputs="text",
+    title="Birds Classifier without Fastai",
+    description="A birds classifier over 200 species trained with Fastai"
+    " and deployed with plain pytorch in Gradio.",
+).launch()

deployment/model.py

@@ -46,8 +46,8 @@ def apply_weights(
     application_function: callable,
 ):
     """
-    Takes an input state_dict and applies those weights to the `input_model`, potentially
-    with a modifier function.
+    Takes an input state_dict and applies those weights to the `input_model`,
+    potentially with a modifier function.
 
     Args:
         input_model (`nn.Module`):
@@ -56,7 +56,8 @@ def apply_weights(
             A dictionary of weights, the trained model's `state_dict()`
         application_function (`callable`):
             A function that takes in one parameter and layer name from `input_model`
-            and the `input_weights`. Should apply the weights from the state dict into `input_model`.
+            and the `input_weights`. Should apply the weights from the state dict into
+            `input_model`.
     """
     model_dict = input_model.state_dict()
     for name, parameter in model_dict.items():

deployment/requirements.txt

@@ -1,4 +1,4 @@
-gradio==3.18.0
+gradio==3.20.1
 pillow==9.4.0
 timm==0.6.12
 torch==1.13.1

deployment/vocab.py

@@ -0,0 +1,202 @@
+vocab = [
+    "Acadian_Flycatcher",
+    "American_Crow",
+    "American_Goldfinch",
+    "American_Pipit",
+    "American_Redstart",
+    "American_Three_Toed_Woodpecker",
+    "Anna_Hummingbird",
+    "Artic_Tern",
+    "Baird_Sparrow",
+    "Baltimore_Oriole",
+    "Bank_Swallow",
+    "Barn_Swallow",
+    "Bay_Breasted_Warbler",
+    "Belted_Kingfisher",
+    "Bewick_Wren",
+    "Black_And_White_Warbler",
+    "Black_Billed_Cuckoo",
+    "Black_Capped_Vireo",
+    "Black_Footed_Albatross",
+    "Black_Tern",
+    "Black_Throated_Blue_Warbler",
+    "Black_Throated_Sparrow",
+    "Blue_Grosbeak",
+    "Blue_Headed_Vireo",
+    "Blue_Jay",
+    "Blue_Winged_Warbler",
+    "Boat_Tailed_Grackle",
+    "Bobolink",
+    "Bohemian_Waxwing",
+    "Brandt_Cormorant",
+    "Brewer_Blackbird",
+    "Brewer_Sparrow",
+    "Bronzed_Cowbird",
+    "Brown_Creeper",
+    "Brown_Pelican",
+    "Brown_Thrasher",
+    "Cactus_Wren",
+    "California_Gull",
+    "Canada_Warbler",
+    "Cape_Glossy_Starling",
+    "Cape_May_Warbler",
+    "Cardinal",
+    "Carolina_Wren",
+    "Caspian_Tern",
+    "Cedar_Waxwing",
+    "Cerulean_Warbler",
+    "Chestnut_Sided_Warbler",
+    "Chipping_Sparrow",
+    "Chuck_Will_Widow",
+    "Clark_Nutcracker",
+    "Clay_Colored_Sparrow",
+    "Cliff_Swallow",
+    "Common_Raven",
+    "Common_Tern",
+    "Common_Yellowthroat",
+    "Crested_Auklet",
+    "Dark_Eyed_Junco",
+    "Downy_Woodpecker",
+    "Eared_Grebe",
+    "Eastern_Towhee",
+    "Elegant_Tern",
+    "European_Goldfinch",
+    "Evening_Grosbeak",
+    "Field_Sparrow",
+    "Fish_Crow",
+    "Florida_Jay",
+    "Forsters_Tern",
+    "Fox_Sparrow",
+    "Frigatebird",
+    "Gadwall",
+    "Geococcyx",
+    "Glaucous_Winged_Gull",
+    "Golden_Winged_Warbler",
+    "Grasshopper_Sparrow",
+    "Gray_Catbird",
+    "Gray_Crowned_Rosy_Finch",
+    "Gray_Kingbird",
+    "Great_Crested_Flycatcher",
+    "Great_Grey_Shrike",
+    "Green_Jay",
+    "Green_Kingfisher",
+    "Green_Tailed_Towhee",
+    "Green_Violetear",
+    "Groove_Billed_Ani",
+    "Harris_Sparrow",
+    "Heermann_Gull",
+    "Henslow_Sparrow",
+    "Herring_Gull",
+    "Hooded_Merganser",
+    "Hooded_Oriole",
+    "Hooded_Warbler",
+    "Horned_Grebe",
+    "Horned_Lark",
+    "Horned_Puffin",
+    "House_Sparrow",
+    "House_Wren",
+    "Indigo_Bunting",
+    "Ivory_Gull",
+    "Kentucky_Warbler",
+    "Laysan_Albatross",
+    "Lazuli_Bunting",
+    "Le_Conte_Sparrow",
+    "Least_Auklet",
+    "Least_Flycatcher",
+    "Least_Tern",
+    "Lincoln_Sparrow",
+    "Loggerhead_Shrike",
+    "Long_Tailed_Jaeger",
+    "Louisiana_Waterthrush",
+    "Magnolia_Warbler",
+    "Mallard",
+    "Mangrove_Cuckoo",
+    "Marsh_Wren",
+    "Mockingbird",
+    "Mourning_Warbler",
+    "Myrtle_Warbler",
+    "Nashville_Warbler",
+    "Nelson_Sharp_Tailed_Sparrow",
+    "Nighthawk",
+    "Northern_Flicker",
+    "Northern_Fulmar",
+    "Northern_Waterthrush",
+    "Olive_Sided_Flycatcher",
+    "Orange_Crowned_Warbler",
+    "Orchard_Oriole",
+    "Ovenbird",
+    "Pacific_Loon",
+    "Painted_Bunting",
+    "Palm_Warbler",
+    "Parakeet_Auklet",
+    "Pelagic_Cormorant",
+    "Philadelphia_Vireo",
+    "Pied_Billed_Grebe",
+    "Pied_Kingfisher",
+    "Pigeon_Guillemot",
+    "Pileated_Woodpecker",
+    "Pine_Grosbeak",
+    "Pine_Warbler",
+    "Pomarine_Jaeger",
+    "Prairie_Warbler",
+    "Prothonotary_Warbler",
+    "Purple_Finch",
+    "Red_Bellied_Woodpecker",
+    "Red_Breasted_Merganser",
+    "Red_Cockaded_Woodpecker",
+    "Red_Eyed_Vireo",
+    "Red_Faced_Cormorant",
+    "Red_Headed_Woodpecker",
+    "Red_Legged_Kittiwake",
+    "Red_Winged_Blackbird",
+    "Rhinoceros_Auklet",
+    "Ring_Billed_Gull",
+    "Ringed_Kingfisher",
+    "Rock_Wren",
+    "Rose_Breasted_Grosbeak",
+    "Ruby_Throated_Hummingbird",
+    "Rufous_Hummingbird",
+    "Rusty_Blackbird",
+    "Sage_Thrasher",
+    "Savannah_Sparrow",
+    "Sayornis",
+    "Scarlet_Tanager",
+    "Scissor_Tailed_Flycatcher",
+    "Scott_Oriole",
+    "Seaside_Sparrow",
+    "Shiny_Cowbird",
+    "Slaty_Backed_Gull",
+    "Song_Sparrow",
+    "Sooty_Albatross",
+    "Spotted_Catbird",
+    "Summer_Tanager",
+    "Swainson_Warbler",
+    "Tennessee_Warbler",
+    "Tree_Sparrow",
+    "Tree_Swallow",
+    "Tropical_Kingbird",
+    "Vermilion_Flycatcher",
+    "Vesper_Sparrow",
+    "Warbling_Vireo",
+    "Western_Grebe",
+    "Western_Gull",
+    "Western_Meadowlark",
+    "Western_Wood_Pewee",
+    "Whip_Poor_Will",
+    "White_Breasted_Kingfisher",
+    "White_Breasted_Nuthatch",
+    "White_Crowned_Sparrow",
+    "White_Eyed_Vireo",
+    "White_Necked_Raven",
+    "White_Pelican",
+    "White_Throated_Sparrow",
+    "Wilson_Warbler",
+    "Winter_Wren",
+    "Worm_Eating_Warbler",
+    "Yellow_Bellied_Flycatcher",
+    "Yellow_Billed_Cuckoo",
+    "Yellow_Breasted_Chat",
+    "Yellow_Headed_Blackbird",
+    "Yellow_Throated_Vireo",
+    "Yellow_Warbler",
+]

training/environment.yml → environment.yml

@@ -17,6 +17,7 @@ dependencies:
   - pip:
       - ipykernel
       - ipywidgets
+      - gradio==3.20.1
       - timm==0.6.12
       - kaggle==1.5.12
       - flake8

training/birds/config.py

@@ -1,4 +1,5 @@
 DATA_STORAGE_PATH = "../data"
+MODELS_STORAGE_PATH = "../models"
 DATASET = "200-bird-species-with-11788-images"
 OWNER = "veeralakrishna"
 

training/birds/train.py

@@ -76,5 +76,6 @@ if __name__ == "__main__":
 
     learner.fine_tune(7, base_lr=0.001, freeze_epochs=12)
 
-    learner.export("models/vit_exported")
+    learner.export(Path(config.MODELS_STORAGE_PATH).resolve() / "vit_exported.pkl")
+    learner.model_dir = Path(config.MODELS_STORAGE_PATH).resolve()
     learner.save("vit_saved", with_opt=False)