Add Modal training and inpainting model scaffold

modal_inpaint.py

@@ -0,0 +1,131 @@
+#!/usr/bin/env python3
+"""Run Clover Image Tiny inpainting training on Modal.
+
+The default job writes its result to a persistent Modal Volume. This keeps
+training independent from Hub credentials; the trained directory can be
+downloaded and uploaded to the model repository after validation.
+"""
+
+from __future__ import annotations
+
+import os
+import subprocess
+import sys
+from pathlib import Path
+
+import modal
+
+
+APP_NAME = "clover-image-tiny-inpaint"
+OUTPUT_VOLUME_NAME = "clover-image-tiny-inpaint-output"
+OUTPUT_ROOT = Path("/outputs")
+
+image = (
+    modal.Image.debian_slim(python_version="3.11")
+    .pip_install(
+        "accelerate==1.14.0",
+        "datasets==4.8.5",
+        "diffusers==0.39.0",
+        "ftfy==6.3.1",
+        "huggingface_hub==0.36.0",
+        "numpy==2.2.6",
+        "pillow==12.3.0",
+        "safetensors==0.8.0",
+        "torch==2.7.0",
+        "torchvision==0.22.0",
+        "transformers==4.57.6",
+    )
+    .add_local_dir("inpainting", remote_path="/root/inpainting")
+)
+
+output_volume = modal.Volume.from_name(OUTPUT_VOLUME_NAME, create_if_missing=True)
+app = modal.App(
+    APP_NAME,
+    image=image,
+    volumes={str(OUTPUT_ROOT): output_volume},
+)
+
+
+@app.function(gpu="A10G", timeout=4 * 60 * 60)
+def train(
+    *,
+    base_model: str = "neonforestmist/Clover-Image-Tiny",
+    base_revision: str = "63b0e9f6be9c00888ff464f342a9ef052bf76681",
+    dataset: str = "prithivMLmods/Caption3o-Opt",
+    dataset_revision: str = "17e893f785fcd3f5d6fc4a5d65a914b9f7b1ff5b",
+    dataset_split: str = "train",
+    image_column: str = "image",
+    caption_column: str = "caption",
+    max_train_steps: int = 4000,
+    output_name: str = "clover-image-tiny-inpaint",
+    max_train_samples: int | None = None,
+    learning_rate: float = 1e-5,
+    seed: int = 20260810,
+) -> str:
+    """Train once and return the Modal Volume path containing the pipeline."""
+
+    output_dir = OUTPUT_ROOT / output_name
+    if output_dir.exists():
+        raise RuntimeError(f"Output already exists; choose another --output-name: {output_dir}")
+
+    command = [
+        sys.executable,
+        "/root/inpainting/train.py",
+        "--pretrained_model_name_or_path",
+        base_model,
+        "--revision",
+        base_revision,
+        "--dataset_name",
+        dataset,
+        "--dataset_revision",
+        dataset_revision,
+        "--dataset_split",
+        dataset_split,
+        "--image_column",
+        image_column,
+        "--caption_column",
+        caption_column,
+        "--max_train_steps",
+        str(max_train_steps),
+        "--learning_rate",
+        str(learning_rate),
+        "--gradient_accumulation_steps",
+        "4",
+        "--gradient_checkpointing",
+        "--mixed_precision",
+        "fp16",
+        "--seed",
+        str(seed),
+        "--output_dir",
+        str(output_dir),
+    ]
+    if max_train_samples is not None:
+        command.extend(["--max_train_samples", str(max_train_samples)])
+
+    env = os.environ.copy()
+    env.setdefault("HF_HOME", "/root/.cache/huggingface")
+    subprocess.run(command, check=True, env=env)
+    output_volume.commit()
+    return str(output_dir)
+
+
+@app.local_entrypoint()
+def main(
+    smoke: bool = False,
+    steps: int = 4000,
+    output_name: str = "clover-image-tiny-inpaint",
+) -> None:
+    """Launch a bounded smoke job or the full A10G run."""
+
+    if smoke:
+        steps = min(steps, 2)
+        samples = 4
+        output_name = f"{output_name}-smoke"
+    else:
+        samples = None
+    result = train.remote(
+        max_train_steps=steps,
+        max_train_samples=samples,
+        output_name=output_name,
+    )
+    print(f"Training output is available in Modal Volume {OUTPUT_VOLUME_NAME}: {result}")