formatting updates (#1480)

Author: charlesfrye

13_sandboxes/test_case_generator.py

@@ -4,17 +4,17 @@
 # ---
 
 # # LLM-Generated Unit Test Development
-#
+
 # Unit tests can become tedious to generate and maintain. While LLMs are a useful tool
 # for generating test cases, hallucinations are always possible, making the code
 # potentially unsafe to run locally.
 #
 # In this example, we'll show you how to generate unit tests in a [sample repository](https://github.com/modal-labs/password-analyzer) using an open-source LLM,
 # and then run them in Modal Sandboxes - our sandboxed environment. We'll then open ports
 # on each of our Sandboxes, showing the code diff and new test case coverage.
-#
+
 # # Model Setup
-#
+
 # First, let's pick an LLM to do the heavy lifting. We went with [deepseek-coder-6.7b-instruct](https://huggingface.co/deepseek-ai/deepseek-coder-6.7b-instruct)
 # which offers a good trade-off between size and quality. At just 7B parameters, it ranks 9th on Hugging Face's
 # [Big Code Models Leaderboard](https://huggingface.co/spaces/bigcode/bigcode-models-leaderboard), only beat out primarily by larger models.
@@ -31,6 +31,7 @@
 files_volume = modal.Volume.from_name("files-volume", create_if_missing=True)
 
 # ## Model Dependencies
+
 # We'll package our dependencies into a [Modal Image](https://modal.com/docs/reference/modal.Image). Starting from a
 # container image provided [by the SGLang team via Dockerhub](https://hub.docker.com/r/lmsysorg/sglang/tags),
 # we'll also add a few packages and flags from Hugging Face to facilitate fast model download.
@@ -41,32 +42,25 @@
         "accelerate==1.8.1",
         "hf_transfer==0.1.9",
     )
-    .env(
-        {
-            "HF_HUB_ENABLE_HF_TRANSFER": "1",
-            "HF_HOME": "/cache",
-        }
-    )
+    .env({"HF_HUB_ENABLE_HF_TRANSFER": "1", "HF_HOME": "/cache"})
     .entrypoint([])  # silence noisy logs
 )
 
-app = modal.App(
-    name="sandbox-test-case-generator",
-)
+app = modal.App(name="sandbox-test-case-generator")
 
 
 # ## Model Server
+
 # Let's put it all together to set up our inference server! Using a [modal.Cls](https://modal.com/docs/reference/modal.Cls), we can
 # easily attach an L40S GPU by setting the `gpu` parameter. The `@modal.enter()` decorator creates
 # a `download_model` lifecycle function, which is run once when the container starts. These are executed sequentially,
 # so our SGLang server starts once the weights are downloaded. Finally, the `@modal.web_endpoint()` converts this
 # into a web endpoint that can be invoked for model inference.
+
+
 @app.cls(
     image=server_image,
-    volumes={
-        "/cache": model_volume,
-        "/data": files_volume,
-    },
+    volumes={"/cache": model_volume, "/data": files_volume},
     gpu="L40S",
     timeout=600,
 )
@@ -108,6 +102,7 @@ def serve(self):
 
 
 # ## Model Client
+
 # Now that our server is set up, let's create a client. We [parametrize](https://modal.com/docs/guide/parametrized-functions#using-parametrized-functions-with-lifecycle-functions) our function with
 # the server `url`, which we'll pass in later. We'll add a [`@modal.method()`](https://modal.com/docs/reference/modal.method#modalmethod) decorator
 # to register our `generate()` function as a Modal Function. Finally, we can invoke our OpenAI-compatible server with our prompt
@@ -118,9 +113,7 @@ def serve(self):
     image=modal.Image.debian_slim(python_version="3.12").uv_pip_install(
         "openai==1.97.1"
     ),
-    volumes={
-        "/data": files_volume,
-    },
+    volumes={"/data": files_volume},
 )
 class TestCaseClient:
     url: str = modal.parameter()
@@ -244,11 +237,15 @@ def download_files_to_volume(
 
 
 # # Sandbox Setup
+
 # ## Image
+
 # Now, let's create a secure environment for our generated test cases to run in.
 # We'll define another Modal Image for our [Modal Sandbox](https://modal.com/docs/guide/sandboxes), installing
 # the [Allure Framework](https://github.com/allure-framework) to generate a report, as well as
 # [git](https://git-scm.com/) to clone our [sample repo](https://github.com/modal-labs/password-analyzer).
+
+
 def get_sandbox_image(gh_owner: str, gh_repo_name: str):
     ALLURE_VERSION = "2.34.1"
     MODULE_URL = f"https://github.com/{gh_owner}/{gh_repo_name}"
@@ -270,6 +267,7 @@ def get_sandbox_image(gh_owner: str, gh_repo_name: str):
 
 
 # ## Sandbox Command
+
 # Next, we define our Sandbox command, chaining together a series of commands. We'll show the difference
 # between the two unit test files, re-run unit tests with the LLM-generated files to check test case
 # coverage, and show the results of both in ports 8000 and 8001.
@@ -308,11 +306,14 @@ def run_sandbox(image: modal.Image, file_name: str):
 
 
 # # Put it all together!
+
 # Finally, we can define a [local_entrypoint](https://modal.com/docs/reference/modal.App#local_entrypoint) and chain
 # everything together. We'll create our test case server and download the Github files from our repo to a Modal Volume.
 # We'll use [`map.aio`](https://modal.com/docs/reference/modal.Function#map) to invoke our server asynchronously,
 # generating our unit test files in parallel. Finally, we'll create our Sandboxes to validate our new test cases,
 # similarly using [`sb.wait.aio`](https://modal.com/docs/reference/modal.Sandbox#wait).
+
+
 @app.local_entrypoint()
 async def main(
     gh_owner: str,
@@ -324,7 +325,7 @@ async def main(
     import asyncio
 
     # Start server
-    sg_lang_server = TestCaseServer()
+    sglang_server = TestCaseServer()
 
     # Download files to volume
     input_files = download_files_to_volume.remote(
@@ -335,7 +336,7 @@ async def main(
     )
 
     # Initialize client and generate test files
-    generator = TestCaseClient(url=sg_lang_server.serve.get_web_url())  # type: ignore
+    generator = TestCaseClient(url=sglang_server.serve.get_web_url())  # type: ignore
     output_generator = generator.generate.map.aio(input_files)
     output_files = []
     async for f in output_generator:
@@ -352,7 +353,9 @@ async def main(
 
 
 # # Addenda
-# The below functions are utility functions.
+
+# The functions below are utilities used above.
+
 import subprocess
 import time