speedy-utils 1.1.40__tar.gz → 1.1.43__tar.gz

speedy_utils-1.1.43/.githooks/pre-push

@@ -0,0 +1,32 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+if [[ "${SKIP_TAG_HOOK:-}" == "1" ]]; then
+  exit 0
+fi
+
+remote_name="${1:-}"
+remote_url="${2:-}"
+
+# Only tag for GitHub remotes.
+if [[ -z "$remote_url" || "$remote_url" != *"github.com"* ]]; then
+  exit 0
+fi
+
+today="$(date +%y%m%d)"
+
+max_suffix="$(git tag -l "${today}.*" | awk -F. 'NF==2 && $1 ~ /^[0-9]{6}$/ && $2 ~ /^[0-9]+$/ {print $2}' | sort -n | tail -n 1)"
+if [[ -z "${max_suffix}" ]]; then
+  next_suffix=1
+else
+  next_suffix=$((max_suffix + 1))
+fi
+
+tag="${today}.${next_suffix}"
+
+if git rev-parse -q --verify "refs/tags/${tag}" >/dev/null; then
+  exit 0
+fi
+
+git tag "${tag}"
+SKIP_TAG_HOOK=1 git push "${remote_name}" "refs/tags/${tag}"

speedy_utils-1.1.43/.github/prompts/improveParallelErrorHandling.prompt.md

@@ -0,0 +1,64 @@
+---
+name: improveParallelErrorHandling
+description: Enhance error tracebacks in parallel execution with rich formatting and context
+argument-hint: the parallel execution function and backend type
+---
+
+Improve error handling for the specified parallel execution function to provide clean, user-focused tracebacks similar to direct function calls.
+
+## Requirements
+
+1. **Filter Internal Frames**: Remove framework/library internal frames from tracebacks, showing only user code
+2. **Add Context Lines**: Display 3 lines before and after each error location with line numbers
+3. **Include Caller Frame**: Show where the parallel execution function was called, not just where the error occurred
+4. **Rich Formatting**: Use rich library's Panel/formatting for clean, readable output
+5. **Suppress Noise**: Set environment variables or flags to suppress verbose framework error logs
+
+## Implementation Steps
+
+1. **Capture Caller Context**: Use `inspect.currentframe().f_back` to capture where the parallel function was called (filename, line number, function name)
+
+2. **Wrap Error Handling**: Catch framework-specific exceptions (e.g., `RayTaskError`, thread exceptions) in the execution loop
+
+3. **Parse/Extract Original Exception**: Get the underlying user exception from the framework wrapper
+   - Extract exception type, message, and traceback information
+   - Parse from string representation if traceback objects aren't preserved
+
+4. **Filter Frames**: Skip frames matching internal paths:
+   - Framework internals (e.g., `ray/_private`, `concurrent/futures`)
+   - Library worker implementations (e.g., `speedy_utils/multi_worker`)
+   - Site-packages for the framework
+
+5. **Format with Context**:
+   - For each user frame, show: `filepath:lineno in function_name`
+   - Use `linecache.getline()` to retrieve surrounding lines
+   - Highlight the error line with `❱` marker
+   - Number all lines (e.g., `   4 │ code here` or `   5 ❱ error here`)
+
+6. **Display Caller Frame First**: Show where the parallel function was invoked before showing the actual error location
+
+7. **Clean Exit**: Flush output streams before exiting to ensure traceback displays
+
+## Example Output Format
+
+```
+╭─────────────── Traceback (most recent call last) ───────────────╮
+│ /path/to/user/script.py:42 in main                              │
+│                                                                  │
+│   40 │ data = load_data()                                        │
+│   41 │ # Process in parallel                                     │
+│   42 ❱ results = multi_process(process_item, data, workers=8)   │
+│   43 │                                                           │
+│                                                                  │
+│ /path/to/user/module.py:15 in process_item                      │
+│                                                                  │
+│   12 │ def process_item(item):                                   │
+│   13 │     value = item['key']                                   │
+│   14 │     denominator = value - 100                             │
+│   15 ❱     return 1 / denominator                                │
+│   16 │                                                           │
+╰──────────────────────────────────────────────────────────────────╯
+ZeroDivisionError: division by zero
+```
+
+Apply these improvements to the specified parallel execution function, ensuring error messages are as clear as direct function calls while maintaining all performance benefits of parallel execution.

{speedy_utils-1.1.40 → speedy_utils-1.1.43}/.gitignore

@@ -122,3 +122,4 @@ dmypy.json
 .copilot
 .vscode/settings.json
 .codegen
+edu_results.json

speedy_utils-1.1.43/AGENTS.md

@@ -0,0 +1,32 @@
+# Repository Guidelines
+
+## Project Structure & Module Organization
+
+- `src/` contains `speedy_utils`, `llm_utils`, and `vision_utils` packages.
+- `tests/` holds automated tests; `examples/` and `notebooks/` are usage references.
+- `scripts/` and `experiments/` are for tooling and experiments; keep changes scoped.
+- `docs/` contains documentation assets.
+- `pyproject.toml`, `ruff.toml`, and `bumpversion.sh` define tooling and release helpers.
+
+## Build, Test, and Development Commands
+
+- `pip install -e .` installs the package in editable mode.
+- `uv pip install -e .` is a drop-in alternative if you use uv.
+- `python -m pytest` or `pytest tests` runs the test suite.
+- `ruff check .` runs lint rules; `ruff format .` formats code.
+
+## Coding Style & Naming Conventions
+
+- Formatting is aligned with Black-style settings (88 char lines) and Ruff rules in `ruff.toml`.
+- Use `snake_case` for Python modules and functions; class names follow `CamelCase`.
+- Keep public APIs exported from `src/*/__init__.py` small and intentional.
+
+## Testing Guidelines
+
+- Tests live in `tests/` and should be named `test_*.py`.
+- Prefer pytest-style assertions and keep fixtures near the tests that use them.
+
+## Commit & Pull Request Guidelines
+
+- Recent history includes informal messages; prefer concise, descriptive imperatives (e.g., `add cache backend`).
+- PRs should include test results and note any new dependencies or optional extras.

{speedy_utils-1.1.40 → speedy_utils-1.1.43}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: speedy-utils
-Version: 1.1.40
+Version: 1.1.43
 Summary: Fast and easy-to-use package for data science
 Project-URL: Homepage, https://github.com/anhvth/speedy
 Project-URL: Repository, https://github.com/anhvth/speedy
@@ -17,7 +17,7 @@ Classifier: Programming Language :: Python :: 3.11
 Classifier: Programming Language :: Python :: 3.12
 Classifier: Programming Language :: Python :: 3.13
 Classifier: Programming Language :: Python :: 3.14
-Requires-Python: >=3.8
+Requires-Python: >=3.9
 Requires-Dist: aiohttp
 Requires-Dist: bump2version
 Requires-Dist: cachetools
@@ -39,13 +39,15 @@ Requires-Dist: pydantic
 Requires-Dist: pytest
 Requires-Dist: ray
 Requires-Dist: requests
+Requires-Dist: rich>=14.3.1
 Requires-Dist: ruff
 Requires-Dist: scikit-learn
 Requires-Dist: tabulate
 Requires-Dist: tqdm
 Requires-Dist: xxhash
 Provides-Extra: ray
-Requires-Dist: ray>=2.49.1; (python_version >= '3.9') and extra == 'ray'
+Requires-Dist: ray[data,llm]>=2.40.0; extra == 'ray'
+Requires-Dist: vllm>=0.6.3; extra == 'ray'
 Description-Content-Type: text/markdown
 
 # Speedy Utils

speedy_utils-1.1.43/examples/llm_ray_example.py

@@ -0,0 +1,73 @@
+"""
+Example: Using LLMRay for distributed offline batch inference.
+
+This demonstrates how to process large batches of OpenAI-style messages
+across multiple GPUs in a Ray cluster with automatic data parallelism.
+
+Key concepts:
+    - dp (data parallel): Number of model replicas
+    - tp (tensor parallel): GPUs per replica
+    - Total GPUs = dp * tp
+"""
+from llm_utils import LLMRay
+from speedy_utils import dump_json_or_pickle
+
+# --- Example 1: Simple batch generation ---
+print('=== Example 1: Simple batch generation ===')
+
+# Create LLMRay instance
+# - dp=4: 4 model replicas (workers)
+# - tp=2: each replica uses 2 GPUs
+# - Total: 8 GPUs used
+# - If cluster has 16 GPUs across 2 nodes, Ray will distribute automatically
+llm = LLMRay(
+    model_name='Qwen/Qwen3-0.6B',
+    dp=4,
+    tp=2,
+    sampling_params={'temperature': 0.7, 'max_tokens': 128},
+)
+
+# Prepare messages (OpenAI format: list of message lists)
+messages_list = [
+    [{'role': 'user', 'content': 'What is artificial intelligence?'}],
+    [{'role': 'user', 'content': 'Explain quantum computing in simple terms.'}],
+    [{'role': 'user', 'content': 'Write a haiku about programming.'}],
+    [{'role': 'user', 'content': 'What are the benefits of distributed computing?'}],
+] + [[{'role': 'user', 'content': f'Summarize document {i}'}] for i in range(20)]
+
+# Generate responses (automatically distributed across all workers)
+results = llm.generate(messages_list)
+
+# Save results
+dump_json_or_pickle(results, 'llm_ray_results.json')
+
+print(f'\nProcessed {len(results)} messages')
+print(f'\nSample result:\n{results[0]}')
+
+
+# --- Example 2: Multi-turn conversation ---
+print('\n=== Example 2: Multi-turn conversation ===')
+
+# Multi-turn conversations with system prompts
+inputs = [
+    [
+        {'role': 'system', 'content': 'You are a creative writer.'},
+        {'role': 'user', 'content': 'Write a short story about a robot.'},
+    ],
+    [
+        {'role': 'system', 'content': 'You are a math tutor.'},
+        {'role': 'user', 'content': 'What is 2+2?'},
+        {'role': 'assistant', 'content': '2+2 equals 4.'},
+        {'role': 'user', 'content': 'What about 3+3?'},
+    ],
+]
+
+# Process conversations
+results = llm(inputs)  # Can also use __call__ syntax
+
+for i, result in enumerate(results):
+    print(f'\nConversation {i + 1}:')
+    print(f'Generated: {result["generated_text"][:100]}...')
+
+
+print('\n=== All examples completed! ===')

speedy_utils-1.1.43/examples/test_parallel_gpu.py

@@ -0,0 +1,61 @@
+import time
+import random
+import ray
+from vllm import LLM, SamplingParams
+from speedy_utils.multi_worker.parallel_gpu_pool import RayWorkerBase, RayRunner
+import os
+ray.init(ignore_reinit_error=True)
+
+# --- Define Your Worker ---
+class MyEduWorker(RayWorkerBase):
+    def setup(self):
+        print(f"Worker {self.worker_id}: Loading vLLM Engine...")
+        
+        # Initialize vLLM
+        # Note: Set gpu_memory_utilization based on how many workers share a GPU
+        self.model = LLM(
+            model="Qwen/Qwen3-0.6B", 
+            gpu_memory_utilization=0.4, # Adjust based on your GPU pool density
+            trust_remote_code=True, 
+            enforce_eager=True,
+            
+        )
+        
+        # Set default sampling parameters
+        self.sampling_params = SamplingParams(
+            temperature=0.7,
+            top_p=0.9,
+            max_tokens=128
+        )
+        
+    def process_one_item(self, item):
+        # 'item' is the prompt from your all_files list
+        prompt = f"Summarize this file metadata: {item}"
+        
+        # vLLM offline generation
+        outputs = self.model.generate([prompt], self.sampling_params)
+        
+        # Extract the generated text
+        generated_text = outputs[0].outputs[0].text
+        
+        return {
+            "file": item,
+            "response": generated_text.strip(),
+            "worker_id": self.worker_id,
+            "gpu_idx": ray.get_runtime_context().get_assigned_resources().get("GPU", []),
+            "node_id": ray.get_runtime_context().node_id.hex(),
+            "cuda_visible_devices": os.environ.get("CUDA_VISIBLE_DEVICES", "")
+        }
+
+# --- Run It ---
+# Create fake data (prompts or filenames)
+all_files = [f"document_id_{i}" for i in range(20)]
+
+# Set test_mode=False if you want to use real GPUs
+runner = RayRunner(test_mode=False, gpus_per_worker=2)
+results = runner.run(
+    worker_class=MyEduWorker,
+    all_data=all_files
+)
+from speedy_utils import dump_json_or_pickle
+dump_json_or_pickle(results, "edu_results.json")

speedy_utils-1.1.43/notebooks/parallel_gpu_pool.ipynb

@@ -0,0 +1,89 @@
+{
+ "cells": [
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "de8205ba",
+   "metadata": {},
+   "outputs": [
+    {
+     "ename": "ConnectionError",
+     "evalue": "Could not find any running Ray instance. Please specify the one to connect to by setting `--address` flag or `RAY_ADDRESS` environment variable.",
+     "output_type": "error",
+     "traceback": [
+      "\u001b[31m---------------------------------------------------------------------------\u001b[39m",
+      "\u001b[31mConnectionError\u001b[39m                           Traceback (most recent call last)",
+      "\u001b[36mCell\u001b[39m\u001b[36m \u001b[39m\u001b[32mIn[1]\u001b[39m\u001b[32m, line 24\u001b[39m\n\u001b[32m     20\u001b[39m \u001b[38;5;66;03m# --- Run It ---\u001b[39;00m\n\u001b[32m     21\u001b[39m \u001b[38;5;66;03m# Create fake data\u001b[39;00m\n\u001b[32m     22\u001b[39m all_files = [\u001b[33mf\u001b[39m\u001b[33m\"\u001b[39m\u001b[33mfile_\u001b[39m\u001b[38;5;132;01m{\u001b[39;00mi\u001b[38;5;132;01m}\u001b[39;00m\u001b[33m.pdf\u001b[39m\u001b[33m\"\u001b[39m \u001b[38;5;28;01mfor\u001b[39;00m i \u001b[38;5;129;01min\u001b[39;00m \u001b[38;5;28mrange\u001b[39m(\u001b[32m500\u001b[39m)]\n\u001b[32m---> \u001b[39m\u001b[32m24\u001b[39m \u001b[38;5;28;43;01mwith\u001b[39;49;00m\u001b[43m \u001b[49m\u001b[43mGPUCluster\u001b[49m\u001b[43m(\u001b[49m\u001b[43mtest_mode\u001b[49m\u001b[43m=\u001b[49m\u001b[38;5;28;43;01mFalse\u001b[39;49;00m\u001b[43m)\u001b[49m\u001b[43m \u001b[49m\u001b[38;5;28;43;01mas\u001b[39;49;00m\u001b[43m \u001b[49m\u001b[43mcluster\u001b[49m\u001b[43m:\u001b[49m\n\u001b[32m     25\u001b[39m \u001b[43m    \u001b[49m\u001b[43mresults\u001b[49m\u001b[43m \u001b[49m\u001b[43m=\u001b[49m\u001b[43m \u001b[49m\u001b[43mcluster\u001b[49m\u001b[43m.\u001b[49m\u001b[43mrun\u001b[49m\u001b[43m(\u001b[49m\n\u001b[32m     26\u001b[39m \u001b[43m        \u001b[49m\u001b[43mworker_class\u001b[49m\u001b[43m=\u001b[49m\u001b[43mMyEduWorker\u001b[49m\u001b[43m,\u001b[49m\n\u001b[32m     27\u001b[39m \u001b[43m        \u001b[49m\u001b[43mall_data\u001b[49m\u001b[43m=\u001b[49m\u001b[43mall_files\u001b[49m\n\u001b[32m     28\u001b[39m \u001b[43m    \u001b[49m\u001b[43m)\u001b[49m\n\u001b[32m     30\u001b[39m \u001b[38;5;28mprint\u001b[39m(\u001b[33m\"\u001b[39m\u001b[33mProcessing Complete!\u001b[39m\u001b[33m\"\u001b[39m)\n",
+      "\u001b[36mFile \u001b[39m\u001b[32m~/projects/speedy_utils/src/speedy_utils/multi_worker/parallel_gpu_pool.py:42\u001b[39m, in \u001b[36mGPUCluster.__enter__\u001b[39m\u001b[34m(self)\u001b[39m\n\u001b[32m     39\u001b[39m     \u001b[38;5;28;01mreturn\u001b[39;00m \u001b[38;5;28mself\u001b[39m\n\u001b[32m     41\u001b[39m \u001b[38;5;28;01mif\u001b[39;00m \u001b[38;5;129;01mnot\u001b[39;00m ray.is_initialized():\n\u001b[32m---> \u001b[39m\u001b[32m42\u001b[39m     \u001b[43mray\u001b[49m\u001b[43m.\u001b[49m\u001b[43minit\u001b[49m\u001b[43m(\u001b[49m\u001b[43maddress\u001b[49m\u001b[43m=\u001b[49m\u001b[33;43m\"\u001b[39;49m\u001b[33;43mauto\u001b[39;49m\u001b[33;43m\"\u001b[39;49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43mignore_reinit_error\u001b[49m\u001b[43m=\u001b[49m\u001b[38;5;28;43;01mTrue\u001b[39;49;00m\u001b[43m)\u001b[49m\n\u001b[32m     43\u001b[39m     \u001b[38;5;28mself\u001b[39m.is_connected = \u001b[38;5;28;01mTrue\u001b[39;00m\n\u001b[32m     45\u001b[39m resources = ray.cluster_resources()\n",
+      "\u001b[36mFile \u001b[39m\u001b[32m/mnt/data/anhvth8/venvs/Megatron-Bridge-Host/lib/python3.12/site-packages/ray/_private/client_mode_hook.py:104\u001b[39m, in \u001b[36mclient_mode_hook.<locals>.wrapper\u001b[39m\u001b[34m(*args, **kwargs)\u001b[39m\n\u001b[32m    102\u001b[39m     \u001b[38;5;28;01mif\u001b[39;00m func.\u001b[34m__name__\u001b[39m != \u001b[33m\"\u001b[39m\u001b[33minit\u001b[39m\u001b[33m\"\u001b[39m \u001b[38;5;129;01mor\u001b[39;00m is_client_mode_enabled_by_default:\n\u001b[32m    103\u001b[39m         \u001b[38;5;28;01mreturn\u001b[39;00m \u001b[38;5;28mgetattr\u001b[39m(ray, func.\u001b[34m__name__\u001b[39m)(*args, **kwargs)\n\u001b[32m--> \u001b[39m\u001b[32m104\u001b[39m \u001b[38;5;28;01mreturn\u001b[39;00m \u001b[43mfunc\u001b[49m\u001b[43m(\u001b[49m\u001b[43m*\u001b[49m\u001b[43margs\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43m*\u001b[49m\u001b[43m*\u001b[49m\u001b[43mkwargs\u001b[49m\u001b[43m)\u001b[49m\n",
+      "\u001b[36mFile \u001b[39m\u001b[32m/mnt/data/anhvth8/venvs/Megatron-Bridge-Host/lib/python3.12/site-packages/ray/_private/worker.py:1818\u001b[39m, in \u001b[36minit\u001b[39m\u001b[34m(address, num_cpus, num_gpus, resources, labels, object_store_memory, local_mode, ignore_reinit_error, include_dashboard, dashboard_host, dashboard_port, job_config, configure_logging, logging_level, logging_format, logging_config, log_to_driver, namespace, runtime_env, enable_resource_isolation, system_reserved_cpu, system_reserved_memory, **kwargs)\u001b[39m\n\u001b[32m   1815\u001b[39m     job_config.set_py_logging_config(logging_config)\n\u001b[32m   1817\u001b[39m redis_address, gcs_address = \u001b[38;5;28;01mNone\u001b[39;00m, \u001b[38;5;28;01mNone\u001b[39;00m\n\u001b[32m-> \u001b[39m\u001b[32m1818\u001b[39m bootstrap_address = \u001b[43mservices\u001b[49m\u001b[43m.\u001b[49m\u001b[43mcanonicalize_bootstrap_address\u001b[49m\u001b[43m(\u001b[49m\u001b[43maddress\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43m_temp_dir\u001b[49m\u001b[43m)\u001b[49m\n\u001b[32m   1819\u001b[39m \u001b[38;5;28;01mif\u001b[39;00m bootstrap_address \u001b[38;5;129;01mis\u001b[39;00m \u001b[38;5;129;01mnot\u001b[39;00m \u001b[38;5;28;01mNone\u001b[39;00m:\n\u001b[32m   1820\u001b[39m     gcs_address = bootstrap_address\n",
+      "\u001b[36mFile \u001b[39m\u001b[32m/mnt/data/anhvth8/venvs/Megatron-Bridge-Host/lib/python3.12/site-packages/ray/_private/services.py:532\u001b[39m, in \u001b[36mcanonicalize_bootstrap_address\u001b[39m\u001b[34m(addr, temp_dir)\u001b[39m\n\u001b[32m    521\u001b[39m \u001b[38;5;250m\u001b[39m\u001b[33;03m\"\"\"Canonicalizes Ray cluster bootstrap address to host:port.\u001b[39;00m\n\u001b[32m    522\u001b[39m \u001b[33;03mReads address from the environment if needed.\u001b[39;00m\n\u001b[32m    523\u001b[39m \n\u001b[32m   (...)\u001b[39m\u001b[32m    529\u001b[39m \u001b[33;03m    should start a local Ray instance.\u001b[39;00m\n\u001b[32m    530\u001b[39m \u001b[33;03m\"\"\"\u001b[39;00m\n\u001b[32m    531\u001b[39m \u001b[38;5;28;01mif\u001b[39;00m addr \u001b[38;5;129;01mis\u001b[39;00m \u001b[38;5;28;01mNone\u001b[39;00m \u001b[38;5;129;01mor\u001b[39;00m addr == \u001b[33m\"\u001b[39m\u001b[33mauto\u001b[39m\u001b[33m\"\u001b[39m:\n\u001b[32m--> \u001b[39m\u001b[32m532\u001b[39m     addr = \u001b[43mget_ray_address_from_environment\u001b[49m\u001b[43m(\u001b[49m\u001b[43maddr\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43mtemp_dir\u001b[49m\u001b[43m)\u001b[49m\n\u001b[32m    533\u001b[39m \u001b[38;5;28;01mif\u001b[39;00m addr \u001b[38;5;129;01mis\u001b[39;00m \u001b[38;5;28;01mNone\u001b[39;00m \u001b[38;5;129;01mor\u001b[39;00m addr == \u001b[33m\"\u001b[39m\u001b[33mlocal\u001b[39m\u001b[33m\"\u001b[39m:\n\u001b[32m    534\u001b[39m     \u001b[38;5;28;01mreturn\u001b[39;00m \u001b[38;5;28;01mNone\u001b[39;00m\n",
+      "\u001b[36mFile \u001b[39m\u001b[32m/mnt/data/anhvth8/venvs/Megatron-Bridge-Host/lib/python3.12/site-packages/ray/_private/services.py:419\u001b[39m, in \u001b[36mget_ray_address_from_environment\u001b[39m\u001b[34m(addr, temp_dir)\u001b[39m\n\u001b[32m    417\u001b[39m         \u001b[38;5;28;01mreturn\u001b[39;00m \u001b[38;5;28;01mNone\u001b[39;00m\n\u001b[32m    418\u001b[39m     \u001b[38;5;28;01melse\u001b[39;00m:\n\u001b[32m--> \u001b[39m\u001b[32m419\u001b[39m         \u001b[38;5;28;01mraise\u001b[39;00m \u001b[38;5;167;01mConnectionError\u001b[39;00m(\n\u001b[32m    420\u001b[39m             \u001b[33m\"\u001b[39m\u001b[33mCould not find any running Ray instance. \u001b[39m\u001b[33m\"\u001b[39m\n\u001b[32m    421\u001b[39m             \u001b[33m\"\u001b[39m\u001b[33mPlease specify the one to connect to by setting `--address` flag \u001b[39m\u001b[33m\"\u001b[39m\n\u001b[32m    422\u001b[39m             \u001b[33m\"\u001b[39m\u001b[33mor `RAY_ADDRESS` environment variable.\u001b[39m\u001b[33m\"\u001b[39m\n\u001b[32m    423\u001b[39m         )\n\u001b[32m    425\u001b[39m \u001b[38;5;28;01mreturn\u001b[39;00m bootstrap_addr\n",
+      "\u001b[31mConnectionError\u001b[39m: Could not find any running Ray instance. Please specify the one to connect to by setting `--address` flag or `RAY_ADDRESS` environment variable."
+     ]
+    }
+   ],
+   "source": [
+    "import time\n",
+    "import random\n",
+    "# Import the class file we just created\n",
+    "from speedy_utils.multi_worker.parallel_gpu_pool import ParallelGPUPool, GPUCluster\n",
+    "\n",
+    "# --- Define Your Worker ---\n",
+    "class MyEduWorker(ParallelGPUPool):\n",
+    "    def setup(self):\n",
+    "        # Load your heavy model here\n",
+    "        print(f\"Worker {self.worker_id}: Loading Model...\")\n",
+    "        time.sleep(1) # Simulate load\n",
+    "        \n",
+    "    def process_one_item(self, item):\n",
+    "        # Simulate GPU Work\n",
+    "        time.sleep(random.uniform(0.05, 0.2)) \n",
+    "        \n",
+    "        # Return whatever you want (filename, score, etc)\n",
+    "        return f\"{item}_DONE\"\n",
+    "\n",
+    "# --- Run It ---\n",
+    "# Create fake data\n",
+    "all_files = [f\"file_{i}.pdf\" for i in range(500)]\n",
+    "\n",
+    "cluster = GPUCluster(test_mode=False)\n",
+    "results = cluster.run(\n",
+    "    worker_class=MyEduWorker,\n",
+    "    all_data=all_files\n",
+    ")\n",
+    "\n",
+    "print(\"Processing Complete!\")"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "2b67e6b5",
+   "metadata": {},
+   "outputs": [],
+   "source": []
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "Megatron-Bridge-Host (3.12.12)",
+   "language": "python",
+   "name": "python3"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 3
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython3",
+   "version": "3.12.12"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 5
+}

{speedy_utils-1.1.40 → speedy_utils-1.1.43}/pyproject.toml

@@ -1,11 +1,11 @@
 [project]
 name = "speedy-utils"
-version = "1.1.40"
+version = "1.1.43"
 description = "Fast and easy-to-use package for data science"
 authors = [{ name = "AnhVTH", email = "anhvth.226@gmail.com" }]
 readme = "README.md"
 license = { text = "MIT" }
-requires-python = ">=3.8"
+requires-python = ">=3.9"
 dependencies = [
     "numpy",
     "requests",
@@ -33,6 +33,7 @@ dependencies = [
     "ray",
     "aiohttp",
     "pytest",
+    "rich>=14.3.1",
 ]
 classifiers = [
     "Development Status :: 4 - Beta",
@@ -53,7 +54,10 @@ Homepage = "https://github.com/anhvth/speedy"
 Repository = "https://github.com/anhvth/speedy"
 
 [project.optional-dependencies]
-ray = ["ray>=2.49.1; python_version >= '3.9'"]
+ray = [
+    "vllm>=0.6.3",
+    "ray[data,llm]>=2.40.0",
+]
 
 [project.scripts]
 mpython = "speedy_utils.scripts.mpython:main"

speedy_utils-1.1.43/scripts/bug.py

@@ -0,0 +1,34 @@
+# type: ignore
+from speedy_utils import multi_process, multi_thread
+
+
+def do_something(x):
+    if x % 3 == 0:
+        raise ValueError(f'Error at index {x}')
+    return x * 2
+
+
+inputs = range(10)
+
+
+if __name__ == '__main__':
+    print('Testing error_handler="log" with mp backend:')
+    results = multi_process(
+        do_something,
+        inputs,
+        backend='mp',
+        error_handler='log',
+        max_error_files=5,
+    )
+    print(f'Results: {results}')
+    print()
+
+    # print('Testing error_handler="log" with multi_thread:')
+    # results = multi_thread(
+    #     do_something,
+    #     inputs,
+    #     error_handler='log',
+    #     max_error_files=5,
+    # )
+    # print(f'Results: {results}')
+

speedy_utils-1.1.43/scripts/bug_simple.py

@@ -0,0 +1,11 @@
+from speedy_utils import *
+
+
+def do_something(x):
+    x = 10
+    y = 0
+    x/y
+
+
+do_something(1)
+