lab notebook: first data look & training

.gitignore

@@ -1,3 +1,6 @@
+# Project specific
+/data
+
 # Byte-compiled / optimized / DLL files
 __pycache__/
 *.py[cod]

README.md

@@ -1,2 +1,5 @@
 # birds-classification
-Train model for birds classification & gradio app
+
+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.

training/birds/config.py

@@ -0,0 +1,7 @@
+DATA_STORAGE_PATH = "../data"
+DATASET = "200-bird-species-with-11788-images"
+OWNER = "veeralakrishna"
+
+DATA_PATH = DATA_STORAGE_PATH + "/kaggle/" + DATASET
+
+

training/birds/utils/kaggle.py

@@ -0,0 +1,53 @@
+import os
+from pathlib import Path
+from zipfile import ZipFile
+from kaggle import api
+
+
+def running_on_kaggle() -> bool:
+    """
+    Checks if script is running on kaggle
+    :return: true if the script is running on kaggle, false otherwise
+    """
+    if os.environ.get("KAGGLE_KERNEL_RUN_TYPE", ""):
+        return True
+    else:
+        return False
+
+
+def download_competition_data(competition: str, input_path: str | Path) -> None:
+    """
+    Downloads data from kaggle competition only if input folder is empty
+    :param comptetition: string with the competition name id of kaggle
+    :param input_path: path of the input folder
+    """
+    data_path = Path(input_path)
+    if not data_path.exists():
+        data_path.mkdir(parents=True)
+    if not any(data_path.iterdir()):
+        api.competition_download_cli(competition, path=data_path)
+        with ZipFile(data_path / (competition + ".zip"), "r") as zipObj:
+            # Extract all the contents of zip file in current directory
+            zipObj.extractall(path=data_path)
+        os.remove(data_path / (competition + ".zip"))
+
+        print(os.listdir(data_path))
+
+
+def download_dataset(owner: str, dataset: str, input_path: str | Path) -> None:
+    """
+    Downloads data from kaggle competition only if input folder is empty
+    :param comptetition: string with the competition name id of kaggle
+    :param input_path: path of the input folder
+    """
+    data_path = Path(input_path)
+    if not data_path.exists():
+        data_path.mkdir(parents=True)
+    if not any(data_path.iterdir()):
+        api.dataset_download_files(f"{owner}/{dataset}", path=data_path)
+        with ZipFile(data_path / (dataset + ".zip"), "r") as zipObj:
+            # Extract all the contents of zip file in current directory
+            zipObj.extractall(path=data_path)
+        os.remove(data_path / (dataset + ".zip"))
+
+        print(os.listdir(data_path))

training/environment.yml

@@ -0,0 +1,21 @@
+name: fastai
+channels:
+  - pytorch
+  - nvidia
+  - conda-forge
+  - fastchan
+  - defaults
+dependencies:
+  - python=3.10
+  - pip
+  - cudatoolkit=11.7
+  - pytorch==1.13.1
+  - torchvision
+  - pytorch-cuda=11.7
+  - fastcore=1.5.28
+  - fastai=2.7.11
+  - pip:
+      - ipykernel
+      - ipywidgets
+      - timm==0.6.12
+      - kaggle==1.5.12