TensorRT-LLMs/scripts/generate_config_table.py
Venky c059e6caa1
[TRTC-121] [feat] Add recipe selector UI to complement the recipe database (#10125)
Signed-off-by: Venky Ganesh <23023424+venkywonka@users.noreply.github.com>
2025-12-24 23:56:54 -05:00

262 lines
8.1 KiB
Python

# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
# SPDX-License-Identifier: Apache-2.0
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from __future__ import annotations
import json
import os
import sys
from collections import defaultdict
from dataclasses import asdict, dataclass
from pathlib import Path
SCRIPT_DIR = Path(__file__).parent.resolve()
REPO_ROOT = SCRIPT_DIR.parent
if str(REPO_ROOT) not in sys.path:
sys.path.insert(0, str(REPO_ROOT))
from examples.configs.database.database import ( # noqa: E402
DATABASE_LIST_PATH,
RecipeList,
assign_profile,
)
MODEL_INFO = {
"deepseek-ai/DeepSeek-R1-0528": {
"display_name": "DeepSeek-R1",
"url": "https://huggingface.co/deepseek-ai/DeepSeek-R1-0528",
},
"nvidia/DeepSeek-R1-0528-FP4-v2": {
"display_name": "DeepSeek-R1 (NVFP4)",
"url": "https://huggingface.co/nvidia/DeepSeek-R1-0528-FP4-v2",
},
"openai/gpt-oss-120b": {
"display_name": "gpt-oss-120b",
"url": "https://huggingface.co/openai/gpt-oss-120b",
},
}
@dataclass(frozen=True)
class RecipeRow:
model: str
model_display_name: str
model_url: str
gpu: str
num_gpus: int
isl: int
osl: int
concurrency: int
config_path: str
gpu_display: str
performance_profile: str
command: str
config_filename: str
config_github_url: str
config_raw_url: str
def _model_display_and_url(model: str) -> tuple[str, str]:
if model in MODEL_INFO:
info = MODEL_INFO[model]
return info["display_name"], info["url"]
return model, ""
def build_rows(yaml_path) -> list[RecipeRow]:
recipe_list = RecipeList.from_yaml(Path(yaml_path))
model_groups = defaultdict(lambda: defaultdict(list))
for recipe in recipe_list:
key = (recipe.gpu, recipe.num_gpus, recipe.isl, recipe.osl)
model_groups[recipe.model][key].append(recipe)
rows: list[RecipeRow] = []
sorted_models = sorted(model_groups.keys())
for model in sorted_models:
subgroups = model_groups[model]
sorted_keys = sorted(
subgroups.keys(),
key=lambda k: (str(k[0]), int(k[1] or 0), int(k[2] or 0), int(k[3] or 0)),
)
model_display_name, model_url = _model_display_and_url(model)
for key in sorted_keys:
entries = subgroups[key]
entries.sort(key=lambda x: x.concurrency)
for idx, entry in enumerate(entries):
gpu = entry.gpu
num_gpus = entry.num_gpus
gpu_display = f"{num_gpus}x{gpu}" if num_gpus and num_gpus > 1 else gpu
isl = entry.isl
osl = entry.osl
conc = entry.concurrency
config_path = entry.config_path
profile = assign_profile(len(entries), idx, conc)
command = f"trtllm-serve {model} --config ${{TRTLLM_DIR}}/{config_path}"
config_filename = os.path.basename(config_path)
config_github_url = (
f"https://github.com/NVIDIA/TensorRT-LLM/blob/main/{config_path}"
)
config_raw_url = (
f"https://raw.githubusercontent.com/NVIDIA/TensorRT-LLM/main/{config_path}"
)
rows.append(
RecipeRow(
model=model,
model_display_name=model_display_name,
model_url=model_url,
gpu=gpu,
num_gpus=num_gpus,
isl=isl,
osl=osl,
concurrency=conc,
config_path=config_path,
gpu_display=gpu_display,
performance_profile=profile,
command=command,
config_filename=config_filename,
config_github_url=config_github_url,
config_raw_url=config_raw_url,
)
)
return rows
def generate_rst(yaml_path, output_file=None):
rows = build_rows(yaml_path)
model_groups = defaultdict(list)
for row in rows:
model_groups[row.model].append(row)
lines = []
lines.append(".. start-config-table-note")
lines.append(".. include:: ../_includes/note_sections.rst")
lines.append(" :start-after: .. start-note-traffic-patterns")
lines.append(" :end-before: .. end-note-traffic-patterns")
lines.append(".. end-config-table-note")
lines.append("")
sorted_models = sorted(model_groups.keys())
for model in sorted_models:
lines.append(f".. start-{model}")
lines.append("")
model_display_name, model_url = _model_display_and_url(model)
if model_url:
title_text = f"`{model_display_name} <{model_url}>`_"
else:
title_text = model
lines.append(f".. _{model}:")
lines.append("")
lines.append(title_text)
lines.append("~" * len(title_text))
lines.append("")
lines.append(".. list-table::")
lines.append(" :width: 100%")
lines.append(" :header-rows: 1")
lines.append(" :widths: 12 15 15 13 20 25")
lines.append("")
lines.append(" * - GPU")
lines.append(" - Performance Profile")
lines.append(" - ISL / OSL")
lines.append(" - Concurrency")
lines.append(" - Config")
lines.append(" - Command")
entries = sorted(
model_groups[model],
key=lambda r: (
str(r.gpu),
int(r.num_gpus or 0),
int(r.isl or 0),
int(r.osl or 0),
int(r.concurrency or 0),
),
)
for row in entries:
config_link = f"`{row.config_filename} <{row.config_github_url}>`_"
lines.append(f" * - {row.gpu_display}")
lines.append(f" - {row.performance_profile}")
lines.append(f" - {row.isl} / {row.osl}")
lines.append(f" - {row.concurrency}")
lines.append(f" - {config_link}")
lines.append(f" - ``{row.command}``")
lines.append("")
lines.append(f".. end-{model}")
lines.append("")
output_text = "\n".join(lines)
if output_file:
with open(output_file, "w") as f:
f.write(output_text)
else:
print(output_text)
def generate_json(yaml_path, output_file):
rows = build_rows(yaml_path)
source_path = Path(yaml_path)
try:
source = str(source_path.relative_to(REPO_ROOT))
except ValueError:
source = str(source_path)
models = {}
for row in rows:
if row.model not in models:
models[row.model] = {
"display_name": row.model_display_name,
"url": row.model_url,
}
payload = {
"source": source,
"models": models,
"entries": [asdict(r) for r in rows],
}
with open(output_file, "w") as f:
json.dump(payload, f, indent=2, sort_keys=True)
f.write("\n")
if __name__ == "__main__":
yaml_path = Path(DATABASE_LIST_PATH)
if not yaml_path.exists():
print(f"Error: YAML file not found at {yaml_path}", file=sys.stderr)
sys.exit(1)
output_path = REPO_ROOT / "docs/source/deployment-guide/config_table.rst"
json_output_path = REPO_ROOT / "docs/source/_static/config_db.json"
generate_rst(yaml_path, output_file=output_path)
generate_json(yaml_path, output_file=json_output_path)