Skip to content

[None][chore] remove CLI support for mamba cache dtype setting #7119

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
merged 3 commits into from
Aug 25, 2025
Merged

[None][chore] remove CLI support for mamba cache dtype setting #7119

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
9bd4ef2
remove CLI support for mamba cache dtype setting
shaharmor98 Aug 21, 2025
39e3471
Merge branch 'main' into bugfix/remove-mamba-cache-dtype-from-cli
shaharmor98 Aug 24, 2025
e8cb1a2
Merge branch 'main' into bugfix/remove-mamba-cache-dtype-from-cli
shaharmor98 Aug 25, 2025
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
6 changes: 0 additions & 6 deletions tensorrt_llm/bench/benchmark/low_latency.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -56,12 +56,6 @@
default=.90,
help="The percentage of memory to use for KV Cache after model load.",
)
@optgroup.option(
"--mamba_ssm_cache_dtype",
type=click.Choice(["auto", "float16", "bfloat16", "float32"]),
default="auto",
help="Data type for Mamba SSM cache. If 'auto', inferred from model config.",
)
@optgroup.option(
"--max_seq_len",
type=int,
Expand Down
6 changes: 0 additions & 6 deletions tensorrt_llm/bench/benchmark/throughput.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -103,12 +103,6 @@
default=.90,
help="The percentage of memory to use for KV Cache after model load.",
)
@optgroup.option(
"--mamba_ssm_cache_dtype",
type=click.Choice(["auto", "float16", "bfloat16", "float32"]),
default="auto",
help="Data type for Mamba SSM cache. If 'auto', inferred from model config.",
)
@optgroup.group(
"Engine Input Configuration",
help="Input configuration for driving the engine.",
Expand Down
32 changes: 11 additions & 21 deletions tensorrt_llm/commands/serve.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -83,7 +83,6 @@ def get_llm_args(model: str,
moe_expert_parallel_size: Optional[int] = None,
gpus_per_node: Optional[int] = None,
free_gpu_memory_fraction: Optional[float] = None,
mamba_ssm_cache_dtype: str = "auto",
num_postprocess_workers: int = 0,
trust_remote_code: bool = False,
reasoning_parser: Optional[str] = None,
Expand All @@ -99,8 +98,7 @@ def get_llm_args(model: str,
max_beam_width=max_beam_width,
max_seq_len=max_seq_len)
kv_cache_config = KvCacheConfig(
free_gpu_memory_fraction=free_gpu_memory_fraction,
mamba_ssm_cache_dtype=mamba_ssm_cache_dtype)
free_gpu_memory_fraction=free_gpu_memory_fraction, )

dynamic_batch_config = DynamicBatchConfig(
enable_batch_size_tuning=True,
Expand Down Expand Up @@ -266,12 +264,6 @@ def launch_mm_encoder_server(
default=0.9,
help="Free GPU memory fraction reserved for KV Cache, "
"after allocating model weights and buffers.")
@click.option(
"--mamba_ssm_cache_dtype",
type=click.Choice(["auto", "float16", "bfloat16", "float32"]),
default="auto",
help="Data type for Mamba SSM cache. If 'auto', inferred from model config."
)
@click.option(
"--num_postprocess_workers",
type=int,
Expand Down Expand Up @@ -312,17 +304,16 @@ def launch_mm_encoder_server(
help=
"Exit with runtime error when attention window is too large to fit even a single sequence in the KV cache."
)
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, ep_size: Optional[int],
cluster_size: Optional[int], gpus_per_node: Optional[int],
kv_cache_free_gpu_memory_fraction: float, mamba_ssm_cache_dtype: str,
num_postprocess_workers: int, trust_remote_code: bool,
extra_llm_api_options: Optional[str], reasoning_parser: Optional[str],
metadata_server_config_file: Optional[str],
server_role: Optional[str],
fail_fast_on_attention_window_too_large: bool):
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,
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,
extra_llm_api_options: Optional[str], reasoning_parser: Optional[str],
metadata_server_config_file: Optional[str], server_role: Optional[str],
fail_fast_on_attention_window_too_large: bool):
"""Running an OpenAI API compatible server

MODEL: model name | HF checkpoint path | TensorRT engine path
Expand All @@ -343,7 +334,6 @@ def serve(model: str, tokenizer: Optional[str], host: str, port: int,
moe_cluster_parallel_size=cluster_size,
gpus_per_node=gpus_per_node,
free_gpu_memory_fraction=kv_cache_free_gpu_memory_fraction,
mamba_ssm_cache_dtype=mamba_ssm_cache_dtype,
num_postprocess_workers=num_postprocess_workers,
trust_remote_code=trust_remote_code,
reasoning_parser=reasoning_parser,
Expand Down