Add the cuTile compiler dependency for model preparation (#68)

Install the cuTile compiler toolchain for model preparation

Author: eugr

docs/model-preparation.md

@@ -69,6 +69,9 @@ The JPEG decoder builds its bundled libjpeg-turbo source. Preparation requires
 the CUDA toolkit, including `nvcc` and CUPTI; configure `[paths] cuda` when it
 is not supplied by the Python environment.
 
+The `prepare` extra installs `cuda-tile[tileiras]`, including the `tileiras`
+compiler toolchain required by cuTile kernels during model preparation.
+
 The DeepSeek Rust crates share a pinned revision of
 [`local-inference-lab/deepseek-recipe`](https://github.com/local-inference-lab/deepseek-recipe)
 in `helios/Cargo.toml`. Its `flare` branch exposes normalized-message byte

pyproject.toml

@@ -24,6 +24,7 @@ prepare = [
     "transformers[torch,vision,chat_template] @ git+https://github.com/huggingface/transformers.git@1d9654e78a23fb167ef539cbd699dfbd3e4830cf",
     "flashinfer-python[b12x] @ git+https://github.com/local-inference-lab/flashinfer.git@1e7f9e3b19a9e69b81a9f758a51a19d74aed50c0",
     "cuda-python==13.3.1",
+    "cuda-tile[tileiras]==1.6.0",
     "safetensors==0.8.0",
     "tokenizers==0.23.2",
     "huggingface-hub==1.33.0",