Skip to content

[TRTLLM-8241][feat] Aliasing to comply to LlmArgs #9586

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
LinPoly merged 1 commit into from
Dec 3, 2025
Merged

[TRTLLM-8241][feat] Aliasing to comply to LlmArgs #9586

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
64 changes: 37 additions & 27 deletions tensorrt_llm/commands/serve.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -299,20 +299,28 @@ def convert(self, value: Any, param: Optional["click.Parameter"],
default=BuildConfig.model_fields["max_seq_len"].default,
help="Maximum total length of one request, including prompt and outputs. "
"If unspecified, the value is deduced from the model config.")
@click.option("--tp_size", type=int, default=1, help='Tensor parallelism size.')
@click.option("--pp_size",
@click.option("--tensor_parallel_size",
"--tp_size",
type=int,
default=1,
help='Tensor parallelism size.')
@click.option("--pipeline_parallel_size",
"--pp_size",
type=int,
default=1,
help='Pipeline parallelism size.')
@click.option("--cp_size",
@click.option("--context_parallel_size",
"--cp_size",
type=int,
default=1,
help='Context parallelism size.')
@click.option("--ep_size",
@click.option("--moe_expert_parallel_size",
"--ep_size",
type=int,
default=None,
help="expert parallelism size")
@click.option("--cluster_size",
@click.option("--moe_cluster_parallel_size",
"--cluster_size",
type=int,
default=None,
help="expert cluster parallelism size")
Expand All @@ -321,7 +329,8 @@ def convert(self, value: Any, param: Optional["click.Parameter"],
default=None,
help="Number of GPUs per node. Default to None, and it will be "
"detected automatically.")
@click.option("--kv_cache_free_gpu_memory_fraction",
@click.option("--free_gpu_memory_fraction",
"--kv_cache_free_gpu_memory_fraction",
type=float,
default=0.9,
help="Free GPU memory fraction reserved for KV Cache, "
Expand Down Expand Up @@ -398,20 +407,22 @@ def convert(self, value: Any, param: Optional["click.Parameter"],
default=None,
help="[Experimental] Specify a custom chat template. "
"Can be a file path or one-liner template string")
def serve(
model: str, tokenizer: Optional[str], host: str, port: int,
log_level: str, backend: str, max_beam_width: int, max_batch_size: int,
max_num_tokens: int, max_seq_len: int, tp_size: int, pp_size: int,
cp_size: int, ep_size: Optional[int], cluster_size: Optional[int],
gpus_per_node: Optional[int], kv_cache_free_gpu_memory_fraction: float,
num_postprocess_workers: int, trust_remote_code: bool,
revision: Optional[str], extra_llm_api_options: Optional[str],
reasoning_parser: Optional[str], tool_parser: Optional[str],
metadata_server_config_file: Optional[str], server_role: Optional[str],
fail_fast_on_attention_window_too_large: bool,
otlp_traces_endpoint: Optional[str], enable_chunked_prefill: bool,
disagg_cluster_uri: Optional[str], media_io_kwargs: Optional[str],
custom_module_dirs: list[Path], chat_template: Optional[str]):
def serve(model: str, tokenizer: Optional[str], host: str, port: int,
log_level: str, backend: str, max_beam_width: int,
max_batch_size: int, max_num_tokens: int, max_seq_len: int,
tensor_parallel_size: int, pipeline_parallel_size: int,
context_parallel_size: int, moe_expert_parallel_size: Optional[int],
moe_cluster_parallel_size: Optional[int],
gpus_per_node: Optional[int], free_gpu_memory_fraction: float,
num_postprocess_workers: int, trust_remote_code: bool,
revision: Optional[str], extra_llm_api_options: Optional[str],
reasoning_parser: Optional[str], tool_parser: Optional[str],
metadata_server_config_file: Optional[str],
server_role: Optional[str],
fail_fast_on_attention_window_too_large: bool,
otlp_traces_endpoint: Optional[str], enable_chunked_prefill: bool,
disagg_cluster_uri: Optional[str], media_io_kwargs: Optional[str],
custom_module_dirs: list[Path], chat_template: Optional[str]):
"""Running an OpenAI API compatible server

MODEL: model name | HF checkpoint path | TensorRT engine path
Expand All @@ -425,7 +436,6 @@ def serve(
logger.error(
f"Failed to import custom module from {custom_module_dir}: {e}")
raise e

llm_args, _ = get_llm_args(
model=model,
tokenizer=tokenizer,
Expand All @@ -434,13 +444,13 @@ def serve(
max_batch_size=max_batch_size,
max_num_tokens=max_num_tokens,
max_seq_len=max_seq_len,
tensor_parallel_size=tp_size,
pipeline_parallel_size=pp_size,
context_parallel_size=cp_size,
moe_expert_parallel_size=ep_size,
moe_cluster_parallel_size=cluster_size,
tensor_parallel_size=tensor_parallel_size,
pipeline_parallel_size=pipeline_parallel_size,
context_parallel_size=context_parallel_size,
moe_expert_parallel_size=moe_expert_parallel_size,
moe_cluster_parallel_size=moe_cluster_parallel_size,
gpus_per_node=gpus_per_node,
free_gpu_memory_fraction=kv_cache_free_gpu_memory_fraction,
free_gpu_memory_fraction=free_gpu_memory_fraction,
num_postprocess_workers=num_postprocess_workers,
trust_remote_code=trust_remote_code,
revision=revision,
Expand Down
Loading