Skip to content

Instantly share code, notes, and snippets.

View nilreml's full-sized avatar
💾
Core dump in progress

Merlin nilreml

💾
Core dump in progress
  • Hamburg, Germany
  • 19:47 (UTC +02:00)
View GitHub Profile
# Best practice pyproject.toml:
# https://github.com/airbytehq/airbyte/blob/master/airbyte-lib/pyproject.toml
[tool.coverage.run]
include = ["src/*"]
parallel = true
branch = true
[tool.coverage.paths]
check for bash version 4+ (POSIX-compliant) :
[ -z "${BASH_VERSINFO}" ] || [ -z "${BASH_VERSINFO[0]}" ] || [ ${BASH_VERSINFO[0]} -lt 4 ]
path to script even if symlinked:
SCRIPT_DIR="$( cd -P "$( dirname "$(readlink -f "${BASH_SOURCE[0]}")" )" >/dev/null 2>&1 && pwd )"
@nilreml
nilreml / docker
Last active April 17, 2024 17:33
Fast one-liner in current dir:
docker run -v $PWD:$PWD -w $PWD --net=[host|none] image cmd
Verbose docker info:
docker system df -v
Delete all build cache:
docker builder prune -af
# Monkeypatch pytorch to run GEMM kernel autotuning on non-enterprise CUDA GPUs
# until https://github.com/pytorch/pytorch/issues/109489 is resolved:
# Put these 4 lines at the top of the main script, before any imports
# TODO: remove once https://github.com/pytorch/pytorch/issues/109489 is resolved
from torch._inductor import utils
utils._use_template_for_cuda = lambda x, y: True
Render markdown cells:
from IPython.display import display, Markdown
display(Markdown(...)) # e.g. df.to_markdown()
Pretty print tensors:
import lovely_tensors as lt
lt.lovely(...)
Rich print:
pip install rich