Yudhiesh Ravindranath yudhiesh

Ray Serve Unit Test Examples

1. `test_deployment_class.py`

Demonstrates how to test a Ray serve deployment class directly without spinning up a Ray cluster.

The key method: func_or_class, allows accessing the underlying class directly where normally, the class would be revised to a Ray serve deployment type due to the @serve.deployment decorator.

2. `test_deployment_requests.py`

Airflow Slack notifications

Installation

Make sure slackclient v1.3.1 is installed (for apache-airflow 1.10).

pip install -U "apache-airflow[slack,...]"

Install kafka in your local mac machine

run following commands:

brew install kafka
sudo mkdir -p /usr/local/var/run/zookeeper/data
sudo chmod 777 /usr/local/var/run/zookeeper/data
zkServer start

mkdir -p /usr/local/var/lib/kafka-logs

	# train_grpo.py
	#
	# See https://github.com/willccbb/verifiers for ongoing developments
	#
	"""
	citation:

	@misc{brown2025grpodemo,
	title={Granular Format Rewards for Eliciting Mathematical Reasoning Capabilities in Small Language Models},
	author={Brown, William},

	import random
	from metaflow import FlowSpec, step, S3, Flow, Parameter, profile, kubernetes, conda, conda_base

	# change columns according to your schema (or remove column list to load all)
	COLUMNS = ['VendorID', 'tpep_pickup_datetime', 'tpep_dropoff_datetime']

	# group parquet files as 1GB batches
	def shard_data(src, batch_size=1_000_000_000):
	with S3() as s3:
	objs = s3.list_recursive([src])

	import os
	from metaflow import S3

	def put_dir(local_root, s3root):
	root = os.path.abspath(local_root)
	objs = []
	for p, _, files in os.walk(root):
	for f in files:
	path = os.path.join(p, f)
	key = os.path.relpath(path, start=root)

	class LiquidityChanged(TransactionEvent):
	"""A sampled liquidity at any moment."""
	__tablename__ = "liquidity"
	delta0 = sa.Column(Int257, nullable=False, index=False)

	import torch
	import torch.distributed as dist
	import torch.multiprocessing as mp

	from torch.utils.data import IterableDataset, DataLoader


	class DistributedIterableDataset(IterableDataset):
	"""
	Example implementation of an IterableDataset that handles both multiprocessing (num_workers > 0)

	CREATE TABLE test12345 (A Int64) ENGINE = MergeTree ORDER BY A;

	CREATE MATERIALIZED VIEW test12345mv Engine=MergeTree order by A
	AS SELECT A FROM test12345;

	insert into test12345 select number from numbers(10);

	#### maintanance / load stop

	rename table test12345 to test12345_old;

	name: CI

	on: [push]

	jobs:

	test:
	runs-on: ubuntu-latest

	services: