[GuideLLM Refactor] Data pipelines rework and multimodal support (#384)

…icated combinations

## Summary

<!--
Include a short paragraph of the changes introduced in this PR.
If this PR requires additional context or rationale, explain why
the changes are necessary.
-->

## Details

<!--
Provide a detailed list of all changes introduced in this pull request.
-->
- [ ]

## Test Plan

<!--
List the steps needed to test this PR.
-->
-

## Related Issues

<!--
Link any relevant issues that this PR addresses.
-->
- Resolves #

---

- [ ] "I certify that all code in this PR is my own, except as noted
below."

## Use of AI

- [ ] Includes AI-assisted code completion
- [ ] Includes code generated by an AI application
- [ ] Includes AI-generated tests (NOTE: AI written tests should have a
docstring that includes `## WRITTEN BY AI ##`)

Author: markurtz

pylock.toml

@@ -4227,4 +4227,4 @@ strategy = ["inherit_metadata", "static_urls"]
 requires_python = "~=3.12"
 
 [[tool.pdm.targets]]
-requires_python = ">=3.10.0,<3.12"
+requires_python = ">=3.10.0,<3.12"

pyproject.toml

@@ -13,6 +13,13 @@ include = ["*"]
 [tool.pdm]
 distribution = true
 
+[[tool.pdm.source]]
+name = "torch"
+type = "find_links"
+#url = "https://download.pytorch.org/whl/cpu/torch_stable.html"
+url = "https://download.pytorch.org/whl/cpu/torch/"
+include_packages = ["torch"]
+
 
 # ************************************************
 # ********** Project Metadata **********
@@ -54,29 +61,25 @@ dependencies = [
     "httpx[http2]<1.0.0",
     "loguru",
     "msgpack",
-    "numpy",
+    "numpy<2.0.0",
     "pillow",
     "protobuf",
     "pydantic>=2.11.7",
     "pydantic-settings>=2.0.0",
+    "pydub",
     "pyyaml>=6.0.0",
     "rich",
     "sanic",
     "transformers",
     "uvloop>=0.18",
+    "librosa>=0.11.0",
+    "torch",
 ]
 
 [project.optional-dependencies]
-perf = [
-    "orjson",
-    "msgpack",
-    "msgspec",
-    "uvloop",
-]
-recommended = [
-    "tiktoken>=0.11.0",  # For OpenAI tokenizer
-    "blobfile>=3.1.0",   # For OpenAI tokenizer
-]
+perf = ["orjson", "msgpack", "msgspec", "uvloop"]
+openai = ["tiktoken>=0.11.0", "blobfile>=3.1.0"]
+recommended = ["guidellm[perf,openai]"]
 dev = [
     # build
     "build>=1.0.0",
@@ -118,7 +121,7 @@ dev = [
 ]
 
 [dependency-groups]
-dev = [ "guidellm[dev]" ]
+dev = ["guidellm[dev]"]
 
 [project.urls]
 homepage = "https://github.com/vllm-project/guidellm"

src/guidellm/__init__.py

@@ -7,6 +7,8 @@
 import logging
 import os
 
+from datasets import config
+
 with (
     open(os.devnull, "w") as devnull,  # noqa: PTH123
     contextlib.redirect_stderr(devnull),
@@ -19,6 +21,7 @@
     os.environ["TOKENIZERS_PARALLELISM"] = "false"  # Silence warnings for tokenizers
     hf_logging.set_verbosity_error()
     logging.getLogger("transformers").setLevel(logging.ERROR)
+    config.USE_AUDIO_DECODE = False
 
 from .logger import configure_logger, logger
 from .settings import (