Skip to content
15 changes: 15 additions & 0 deletions dockerfile/cuda11.1.1.dockerfile
Original file line number Diff line number Diff line change
Expand Up @@ -24,6 +24,7 @@ RUN apt-get update && \
automake \
bc \
build-essential \
ca-certificates \
curl \
dmidecode \
ffmpeg \
Expand All @@ -46,6 +47,7 @@ RUN apt-get update && \
openssh-client \
openssh-server \
pciutils \
software-properties-common \
sudo \
util-linux \
vim \
Expand All @@ -55,6 +57,19 @@ RUN apt-get update && \
apt-get clean && \
rm -rf /var/lib/apt/lists/* /tmp/* /opt/cmake-3.14.6-Linux-x86_64

# Install Go (used for wandb build)
RUN add-apt-repository -y ppa:longsleep/golang-backports && \
apt-get update && \
apt-get install -y golang-1.24-go
Comment on lines +61 to +63
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

merge it with existing apt-get command

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The 'add-apt-repository' is installed with 'software-properties-common' in the first RUN. So, we cannot combine it with the first apt-get.


# Install Rust (used for wandb build)
RUN curl https://sh.rustup.rs -sSf | sh -s -- -y && \
. /root/.cargo/env && \
rustup update stable && \
rustup default stable

ENV PATH="/usr/lib/go-1.24/bin:/root/.cargo/bin:${PATH}"
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

merge with existing env command

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The previous environment lacks Go and Cargo installations, and the subsequent environment is configured specifically for Docker workflows. Given this separation of concerns, I believe we should maintain the current configuration. Please let me know if you'd like any modifications.


ARG NUM_MAKE_JOBS=

# Install Docker
Expand Down
9 changes: 5 additions & 4 deletions superbench/benchmarks/base.py
Original file line number Diff line number Diff line change
Expand Up @@ -293,10 +293,11 @@ def _process_percentile_result(self, metric, result, reduce_type=None):
if len(result) > 0:
percentile_list = ['50', '90', '95', '99', '99.9']
for percentile in percentile_list:
self._result.add_result(
'{}_{}'.format(metric, percentile),
np.percentile(result, float(percentile), interpolation='nearest'), reduce_type
)
try:
val = np.percentile(result, float(percentile), method='nearest')
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Why do we need update this? Have you met some failure?

Copy link
Contributor Author

@polarG polarG Jan 9, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yes, I met it after the changes for wandb. In PR, this also appeared.

'interpolation' was deprecated in v1.22.0 and removed on 2.4.0. In the python3.12 unit test pipeline, v2.4.1 was installed.

Comment on lines +296 to +297
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

can you open another pr for this?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

If we move this to another PR, the python-unit-test will fail in this PR.

except TypeError:
val = np.percentile(result, float(percentile), interpolation='nearest')
Comment on lines +296 to +299
Copy link

Copilot AI Jan 14, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The try-except approach is too broad for version detection. Since TypeError could occur for other reasons, consider checking NumPy version explicitly at module level or using a more specific exception check. For example, use hasattr(np.percentile, '__wrapped__') or check np.__version__ to determine which parameter to use.

Copilot uses AI. Check for mistakes.
self._result.add_result('{}_{}'.format(metric, percentile), val, reduce_type)

def print_env_info(self):
"""Print environments or dependencies information."""
Expand Down
Loading