Team Noah

This commit is contained in:
firestar5683
2026-09-03 23:31:33 -05:00
parent fed4ce6ee0
commit dd7ac353bd
307 changed files with 15168 additions and 6906 deletions
+145 -124
View File
@@ -88,13 +88,13 @@ jobs:
fail-fast: false
matrix:
dev: ['METAL', 'AMD', 'NV']
timeout-minutes: 60
timeout-minutes: 30
defaults:
run:
shell: bash -e -o pipefail {0}
env:
DEV: ${{ matrix.dev }}
HCQ2: ${{ matrix.dev == 'AMD' && '1' || '0' }}
HCQ2: '0'
if: github.repository_owner == 'tinygrad'
steps:
- name: Checkout Code
@@ -102,16 +102,11 @@ jobs:
- name: Setup (AMD)
if: ${{ matrix.dev == 'AMD' }}
run: |
./extra/amdpci/setup_python_cap.sh
./extra/hcq/hcq_smi.py amd rmmod
./extra/hcq/hcq_smi.py amd kill_pids
./extra/hcq/hcq_smi.py amd rmmod --expect
./extra/hcq/hcq_smi.py amd kill_pids --sudoless
- name: Setup (NV)
if: ${{ matrix.dev == 'NV' }}
run: sudo lsof -tQ /dev/nvidia* | { xargs -r sudo kill -9 || true; }
- name: Symlink models and datasets
run: |
mkdir -p weights
ln -s /raid/weights/LLaMA-3 weights/LLaMA-3
run: lsof -tQ /dev/nvidia* | { xargs -r kill -9 || true; }
- name: setup staging db
if: github.ref == 'refs/heads/update_benchmark_staging'
run: |
@@ -121,18 +116,14 @@ jobs:
run: python3 test/external/process_replay/reset.py
- name: Run llama3.2
run: BENCHMARK_LOG=llama32_3b-f16 JITBEAM=2 IGNORE_BEAM_CACHE=1 python3 -m tinygrad.llm -m llama3.2:3b-f16 --benchmark --warmup
- name: Run qwen3.6
# qwen3.6:35b-a3b doesn't fit on mac
- name: Run qwen3.8
# qwen3.8:27b doesn't fit on mac
if: ${{ matrix.dev != 'METAL' }}
run: BENCHMARK_LOG=qwen36_35b-a3b JITBEAM=2 IGNORE_BEAM_CACHE=1 python3 -m tinygrad.llm -m qwen3.6:35b-a3b --benchmark --warmup
run: BENCHMARK_LOG=qwen38_27b JITBEAM=2 IGNORE_BEAM_CACHE=1 python3 -m tinygrad.llm -m qwen3.8:27b --benchmark --warmup
- name: Run olmoe
# just metal for now
if: ${{ matrix.dev == 'METAL' }}
run: BENCHMARK_LOG=olmoe JITBEAM=2 IGNORE_BEAM_CACHE=1 python3 -m tinygrad.llm -m olmoe --benchmark --warmup
- name: Run LLaMA-3 8B on 4 GPUs with BEAM
# only run on machines with multiple gpus
if: ${{ matrix.dev != 'METAL' }}
run: BENCHMARK_LOG=llama3_beam_4gpu JITBEAM=2 IGNORE_BEAM_CACHE=1 CAPTURE_PROCESS_REPLAY=0 python3 examples/llama3.py --size 8B --shard 4 --model weights/LLaMA-3/8B-SF-DPO/ --benchmark --temperature 0
- name: Run process replay tests
uses: ./.github/actions/process-replay
@@ -143,13 +134,13 @@ jobs:
fail-fast: false
matrix:
dev: ['METAL', 'AMD', 'NV']
timeout-minutes: 60
timeout-minutes: 10
defaults:
run:
shell: bash -e -o pipefail {0}
env:
DEV: ${{ matrix.dev }}
HCQ2: ${{ matrix.dev == 'AMD' && '1' || '0' }}
HCQ2: '0'
if: github.repository_owner == 'tinygrad'
steps:
- name: Checkout Code
@@ -157,12 +148,11 @@ jobs:
- name: Setup (AMD)
if: ${{ matrix.dev == 'AMD' }}
run: |
./extra/amdpci/setup_python_cap.sh
./extra/hcq/hcq_smi.py amd rmmod
./extra/hcq/hcq_smi.py amd kill_pids
./extra/hcq/hcq_smi.py amd rmmod --expect
./extra/hcq/hcq_smi.py amd kill_pids --sudoless
- name: Setup (NV)
if: ${{ matrix.dev == 'NV' }}
run: sudo lsof -tQ /dev/nvidia* | { xargs -r sudo kill -9 || true; }
run: lsof -tQ /dev/nvidia* | { xargs -r kill -9 || true; }
- name: setup staging db
if: github.ref == 'refs/heads/update_benchmark_staging'
run: |
@@ -182,10 +172,6 @@ jobs:
# slow on metal
if: ${{ matrix.dev != 'METAL' }}
run: time BENCHMARK_LOG=cifar DEFAULT_FLOAT=HALF STEPS=1000 TARGET_EVAL_ACC_PCT=93.0 python3 examples/hlb_cifar10.py
- name: Run full CIFAR training steps w 6 GPUS
# only run on machines with multiple gpus
if: ${{ matrix.dev != 'METAL' }}
run: time BENCHMARK_LOG=cifar_6gpu CAPTURE_PROCESS_REPLAY=0 DEFAULT_FLOAT=HALF STEPS=350 BS=1536 GPUS=6 TARGET_EVAL_ACC_PCT=93.0 python3 examples/hlb_cifar10.py
- name: Run process replay tests
uses: ./.github/actions/process-replay
@@ -196,13 +182,13 @@ jobs:
fail-fast: false
matrix:
dev: ['AMD', 'NV']
timeout-minutes: 60
timeout-minutes: 5
defaults:
run:
shell: bash -e -o pipefail {0}
env:
DEV: ${{ matrix.dev }}
HCQ2: ${{ matrix.dev == 'AMD' && '1' || '0' }}
HCQ2: '0'
if: github.repository_owner == 'tinygrad'
steps:
- name: Checkout Code
@@ -210,12 +196,11 @@ jobs:
- name: Setup (AMD)
if: ${{ matrix.dev == 'AMD' }}
run: |
./extra/amdpci/setup_python_cap.sh
./extra/hcq/hcq_smi.py amd rmmod
./extra/hcq/hcq_smi.py amd kill_pids
./extra/hcq/hcq_smi.py amd rmmod --expect
./extra/hcq/hcq_smi.py amd kill_pids --sudoless
- name: Setup (NV)
if: ${{ matrix.dev == 'NV' }}
run: sudo lsof -tQ /dev/nvidia* | { xargs -r sudo kill -9 || true; }
run: lsof -tQ /dev/nvidia* | { xargs -r kill -9 || true; }
- name: Symlink models and datasets
run: |
mkdir -p extra/datasets
@@ -227,15 +212,8 @@ jobs:
rm -f /tmp/staging.db /tmp/staging.db-shm /tmp/staging.db-wal
- name: reset process replay
run: test/external/process_replay/reset.py
- name: Run MLPerf resnet eval on training data
run: time BENCHMARK_LOG=resnet_eval MODEL=resnet python3 examples/mlperf/model_eval.py
- name: Run 10 MLPerf ResNet50 training steps (1 gpu)
run: BENCHMARK_LOG=resnet_10steps DEFAULT_FLOAT=HALF BENCHMARK=10 BS=256 GPUS=1 MODEL=resnet python3 examples/mlperf/model_train.py
- name: Run 10 MLPerf ResNet50 training steps (6 gpu)
run: BENCHMARK_LOG=resnet_10steps_6gpu CAPTURE_PROCESS_REPLAY=0 DEFAULT_FLOAT=HALF BENCHMARK=10 BS=1536 GPUS=6 MODEL=resnet python3 examples/mlperf/model_train.py
- name: Run 10 MLPerf Bert training steps (6 gpu)
# TODO: remove BERT_LAYERS once scheduler is fast
run: BENCHMARK_LOG=bert_10steps_6gpu CAPTURE_PROCESS_REPLAY=0 DEFAULT_FLOAT=HALF BENCHMARK=10 BS=72 GPUS=6 BERT_LAYERS=2 MODEL=bert python3 examples/mlperf/model_train.py
- name: Run process replay tests
uses: ./.github/actions/process-replay
@@ -246,13 +224,13 @@ jobs:
fail-fast: false
matrix:
dev: ['METAL', 'AMD', 'NV']
timeout-minutes: 60
timeout-minutes: 15
defaults:
run:
shell: bash -e -o pipefail {0}
env:
DEV: ${{ matrix.dev }}
HCQ2: ${{ matrix.dev == 'AMD' && '1' || '0' }}
HCQ2: '0'
if: github.repository_owner == 'tinygrad'
steps:
- name: Checkout Code
@@ -260,12 +238,11 @@ jobs:
- name: Setup (AMD)
if: ${{ matrix.dev == 'AMD' }}
run: |
./extra/amdpci/setup_python_cap.sh
./extra/hcq/hcq_smi.py amd rmmod
./extra/hcq/hcq_smi.py amd kill_pids
./extra/hcq/hcq_smi.py amd rmmod --expect
./extra/hcq/hcq_smi.py amd kill_pids --sudoless
- name: Setup (NV)
if: ${{ matrix.dev == 'NV' }}
run: sudo lsof -tQ /dev/nvidia* | { xargs -r sudo kill -9 || true; }
run: lsof -tQ /dev/nvidia* | { xargs -r kill -9 || true; }
- name: setup staging db
if: github.ref == 'refs/heads/update_benchmark_staging'
run: |
@@ -285,6 +262,59 @@ jobs:
- name: Run process replay tests
uses: ./.github/actions/process-replay
multigpubenchmark:
name: Multi-GPU Benchmarks (DEV=${{ matrix.dev }})
runs-on: [self-hosted, "${{ matrix.dev == 'AMD' && 'tinybox' || 'tinyboxgreen' }}"]
strategy:
fail-fast: false
matrix:
dev: ['AMD', 'NV']
timeout-minutes: 20
defaults:
run:
shell: bash -e -o pipefail {0}
env:
DEV: ${{ matrix.dev }}
HCQ2: '0'
if: github.repository_owner == 'tinygrad'
steps:
- name: Checkout Code
uses: actions/checkout@v6
- name: Setup (AMD)
if: ${{ matrix.dev == 'AMD' }}
run: |
./extra/hcq/hcq_smi.py amd rmmod --expect
./extra/hcq/hcq_smi.py amd kill_pids --sudoless
- name: Setup (NV)
if: ${{ matrix.dev == 'NV' }}
run: lsof -tQ /dev/nvidia* | { xargs -r kill -9 || true; }
- name: Symlink models and datasets
run: |
mkdir -p weights
mkdir -p extra/datasets
ln -s /raid/weights/LLaMA-3 weights/LLaMA-3
ln -s /raid/datasets/imagenet extra/datasets/imagenet
- name: setup staging db
if: github.ref == 'refs/heads/update_benchmark_staging'
run: |
echo "CACHEDB=/tmp/staging.db" >> $GITHUB_ENV
rm -f /tmp/staging.db /tmp/staging.db-shm /tmp/staging.db-wal
- name: reset process replay
run: python3 test/external/process_replay/reset.py
- name: Run LLaMA-3 8B on 4 GPUs with BEAM
run: BENCHMARK_LOG=llama3_beam_4gpu JITBEAM=2 IGNORE_BEAM_CACHE=1 CAPTURE_PROCESS_REPLAY=0 python3 examples/llama3.py --size 8B --shard 4 --model weights/LLaMA-3/8B-SF-DPO/ --benchmark --temperature 0
- name: Run full CIFAR training steps w 6 GPUS
run: time BENCHMARK_LOG=cifar_6gpu CAPTURE_PROCESS_REPLAY=0 DEFAULT_FLOAT=HALF STEPS=350 BS=1536 GPUS=6 TARGET_EVAL_ACC_PCT=93.0 python3 examples/hlb_cifar10.py
- name: Run MLPerf resnet eval on training data
run: time BENCHMARK_LOG=resnet_eval MODEL=resnet python3 examples/mlperf/model_eval.py
- name: Run 10 MLPerf ResNet50 training steps (6 gpu)
run: BENCHMARK_LOG=resnet_10steps_6gpu CAPTURE_PROCESS_REPLAY=0 DEFAULT_FLOAT=HALF BENCHMARK=10 BS=1536 GPUS=6 MODEL=resnet python3 examples/mlperf/model_train.py
- name: Run 10 MLPerf Bert training steps (6 gpu)
# TODO: remove BERT_LAYERS once scheduler is fast
run: BENCHMARK_LOG=bert_10steps_6gpu CAPTURE_PROCESS_REPLAY=0 DEFAULT_FLOAT=HALF BENCHMARK=10 BS=72 GPUS=6 BERT_LAYERS=2 MODEL=bert python3 examples/mlperf/model_train.py
- name: Run process replay tests
uses: ./.github/actions/process-replay
tests:
name: Tests (DEV=${{ matrix.dev }})
runs-on: [self-hosted, "${{ matrix.dev == 'METAL' && 'macOS' || matrix.dev == 'AMD' && 'tinybox' || 'tinyboxgreen' }}"]
@@ -292,7 +322,7 @@ jobs:
fail-fast: false
matrix:
dev: ['METAL', 'AMD', 'NV']
timeout-minutes: 60
timeout-minutes: 10
defaults:
run:
shell: bash -e -o pipefail {0}
@@ -305,12 +335,11 @@ jobs:
- name: Setup (AMD)
if: ${{ matrix.dev == 'AMD' }}
run: |
./extra/amdpci/setup_python_cap.sh
./extra/hcq/hcq_smi.py amd rmmod
./extra/hcq/hcq_smi.py amd kill_pids
./extra/hcq/hcq_smi.py amd rmmod --expect
./extra/hcq/hcq_smi.py amd kill_pids --sudoless
- name: Setup (NV)
if: ${{ matrix.dev == 'NV' }}
run: sudo lsof -tQ /dev/nvidia* | { xargs -r sudo kill -9 || true; }
run: lsof -tQ /dev/nvidia* | { xargs -r kill -9 || true; }
- name: setup staging db
if: github.ref == 'refs/heads/update_benchmark_staging'
run: |
@@ -374,7 +403,7 @@ jobs:
run: python test/external/external_benchmark_multitensor_allreduce.py
- name: HEVC Decode Benchmark
if: ${{ matrix.dev == 'NV' }}
run: VALIDATE=1 MAX_FRAMES=100 ASSERT_FPS=1400 JITBEAM=1 PYTHONPATH=. python3 extra/hevc/decode.py
run: IGNORE_BEAM_CACHE=1 VALIDATE=1 MAX_FRAMES=100 ASSERT_FPS=1400 JITBEAM=1 PYTHONPATH=. python3 extra/hevc/decode.py
- uses: actions/upload-artifact@v7
if: ${{ matrix.dev != 'AMD' }}
with:
@@ -387,7 +416,7 @@ jobs:
testusbgpu:
name: UsbGPU Benchmark
runs-on: [self-hosted, macOS]
timeout-minutes: 10
timeout-minutes: 3
defaults:
run:
shell: bash -e -o pipefail {0}
@@ -402,32 +431,66 @@ jobs:
rm -f /tmp/staging.db /tmp/staging.db-shm /tmp/staging.db-wal
- name: Kill stale pids
run: |
PYTHONPATH=. ./extra/hcq/hcq_smi.py amd kill_pids
PYTHONPATH=. ./extra/hcq/hcq_smi.py nv kill_pids
# since sudo is required for usbgpu on macos, do not write bytecode, as some of the files are owned by root
./extra/hcq/hcq_smi.py amd kill_pids --sudoless
./extra/hcq/hcq_smi.py nv kill_pids --sudoless
- name: UsbGPU boot time
run: sudo -E PYTHONDONTWRITEBYTECODE=1 PYTHONPATH=. GMMU=0 DEBUG=2 AM_RESET=1 DEV=USB+AMD time python3.11 test/test_tiny.py TestTiny.test_plus
run: GMMU=0 DEBUG=2 AM_RESET=1 DEV=USB+AMD time python3.11 test/test_tiny.py TestTiny.test_plus
- name: UsbGPU tiny tests
run: sudo -E PYTHONDONTWRITEBYTECODE=1 PYTHONPATH=. GMMU=0 DEV=USB+AMD python3.11 test/test_tiny.py
run: GMMU=0 DEV=USB+AMD python3.11 test/test_tiny.py
- name: UsbGPU copy speeds
run: sudo -E PYTHONDONTWRITEBYTECODE=1 PYTHONPATH=. GMMU=0 DEV=USB+AMD python3.11 test/external/external_test_usb_asm24.py TestDevCopySpeeds
#- name: UsbGPU openpilot test
# run: sudo -E PYTHONPATH=. GMMU=0 DEV=USB+AMD GRAPH_ONE_KERNEL=1 python3.11 examples/openpilot/compile3.py https://github.com/commaai/openpilot/raw/9118973ed03c1ae1d40cf69a29507ec2cc78efd7/selfdrive/modeld/models/supercombo.onnx
run: SIZE=64000000 PYTHONPATH=. GMMU=0 DEV=USB+AMD python3.11 test/external/external_test_usb_asm24.py TestDevCopySpeeds
- name: UsbGPU (USB4/TB) install script
run: PYTHONPATH=. sh extra/setup_tinygpu_osx.sh
run: sh extra/setup_tinygpu_osx.sh
- name: UsbGPU (USB4/TB) boot time
run: PYTHONPATH=. DEBUG=3 DEV=PCI+NV:NAK time python3.11 test/test_tiny.py TestTiny.test_plus
run: DEBUG=3 DEV=PCI+NV:NAK time python3.11 test/test_tiny.py TestTiny.test_plus
- name: UsbGPU (USB4/TB) tiny tests
run: PYTHONPATH=. DEV=PCI+NV:NAK python3.11 test/test_tiny.py
run: DEV=PCI+NV:NAK python3.11 test/test_tiny.py
testcommalatest:
name: comma Benchmark (0.11.2)
testcomma:
strategy:
matrix:
dev: ['QCOM', 'QCOM:IR3']
version: ['0.11.0', '0.11.2']
model: ['vision', 'policy', 'supercombo', 'dmonitoring']
# exclude non-existent models
exclude: [{ version: '0.11.0', model: supercombo }, { version: '0.11.2', model: vision }, { version: '0.11.2', model: policy }]
include:
- version: '0.11.0'
model: vision
url: https://github.com/commaai/openpilot/raw/v0.11.0/selfdrive/modeld/models/driving_vision.onnx
timing: 18
- version: '0.11.0'
model: policy
url: https://github.com/commaai/openpilot/raw/v0.11.0/selfdrive/modeld/models/driving_policy.onnx
timing: 3.4
- version: '0.11.0'
model: dmonitoring
url: https://github.com/commaai/openpilot/raw/v0.11.0/selfdrive/modeld/models/dmonitoring_model.onnx
timing: 12.5
- version: '0.11.2'
model: supercombo
url: https://gitlab.com/commaai/openpilot-lfs.git/gitlab-lfs/objects/433f85f956837606ad1f1cbee4aa7e2158ad23c768dea914b20436c97232741b
timing: 28
- dev: QCOM:IR3
version: '0.11.2'
model: supercombo
timing: 29
- version: '0.11.2'
model: dmonitoring
url: https://gitlab.com/commaai/openpilot-lfs.git/gitlab-lfs/objects/3e7b31dfbc0a5234f1baf196513b77fc6af12204b8a8ffe8ee0417e48352f316
timing: 12.5
fail-fast: false
name: openpilot ${{ matrix.version }} compile3 ${{ matrix.model }} (DEV=${{ matrix.dev }})
runs-on: [self-hosted, Linux, comma]
timeout-minutes: 12
timeout-minutes: 5
defaults:
run:
shell: bash -e -o pipefail {0}
if: github.repository_owner == 'tinygrad'
env:
DEV: ${{ matrix.dev }}
ASSERT_MIN_STEP_TIME: ${{ matrix.timing }}
BENCHMARK_LOG: ${{ matrix.dev == 'QCOM:IR3' && 'ir3_' || '' }}openpilot_${{ matrix.version }}_${{ matrix.model }}
steps:
- name: Checkout Code
uses: actions/checkout@v6
@@ -438,45 +501,10 @@ jobs:
rm -f /tmp/staging.db /tmp/staging.db-shm /tmp/staging.db-wal
- name: reset process replay
run: test/external/process_replay/reset.py
- name: openpilot compile3 0.11.2 supercombo
run: BENCHMARK_LOG=openpilot_0_11_2_supercombo PYTHONPATH="." ASSERT_MIN_STEP_TIME=26 DEV=QCOM FLOAT16=1 IMAGE=1 NOLOCALS=1 taskset -c 4-7 python3 examples/openpilot/compile3.py https://gitlab.com/commaai/openpilot-lfs.git/gitlab-lfs/objects/433f85f956837606ad1f1cbee4aa7e2158ad23c768dea914b20436c97232741b
- name: openpilot compile3 0.11.2 supercombo (from pickle)
run: BENCHMARK_LOG=openpilot_0_11_2_supercombo_run_pickle RUN_PICKLE=1 PYTHONPATH="." ASSERT_MIN_STEP_TIME=26 DEV=QCOM taskset -c 4-7 python3 examples/openpilot/compile3.py
- name: IR3 openpilot compile3 0.11.2 supercombo
run: BENCHMARK_LOG=ir3_openpilot_0_11_2_supercombo PYTHONPATH="." ASSERT_MIN_STEP_TIME=41 DEV=QCOM:IR3 FLOAT16=1 IMAGE=1 NOLOCALS=1 taskset -c 4-7 python3 examples/openpilot/compile3.py https://gitlab.com/commaai/openpilot-lfs.git/gitlab-lfs/objects/433f85f956837606ad1f1cbee4aa7e2158ad23c768dea914b20436c97232741b
- name: openpilot compile3 0.11.2 dmonitoring
run: BENCHMARK_LOG=openpilot_0_11_2_dmonitoring PYTHONPATH="." ASSERT_MIN_STEP_TIME=11 DEV=QCOM FLOAT16=1 IMAGE=1 NOLOCALS=1 taskset -c 4-7 python3 examples/openpilot/compile3.py https://gitlab.com/commaai/openpilot-lfs.git/gitlab-lfs/objects/3e7b31dfbc0a5234f1baf196513b77fc6af12204b8a8ffe8ee0417e48352f316
- name: Run process replay tests
uses: ./.github/actions/process-replay
testcommaold:
name: comma Benchmark (0.11.0)
runs-on: [self-hosted, Linux, comma]
timeout-minutes: 12
defaults:
run:
shell: bash -e -o pipefail {0}
if: github.repository_owner == 'tinygrad'
steps:
- name: Checkout Code
uses: actions/checkout@v6
- name: setup staging db
if: github.ref == 'refs/heads/update_benchmark_staging'
run: |
echo "CACHEDB=/tmp/staging.db" >> $GITHUB_ENV
rm -f /tmp/staging.db /tmp/staging.db-shm /tmp/staging.db-wal
- name: reset process replay
run: test/external/process_replay/reset.py
- name: openpilot compile3 0.11.0 driving_vision
run: BENCHMARK_LOG=openpilot_0_11_0_vision PYTHONPATH="." ASSERT_MIN_STEP_TIME=17 DEV=QCOM FLOAT16=1 IMAGE=1 NOLOCALS=1 taskset -c 4-7 python3 examples/openpilot/compile3.py https://github.com/commaai/openpilot/raw/v0.11.0/selfdrive/modeld/models/driving_vision.onnx
- name: openpilot compile3 0.11.0 driving_vision (from pickle)
run: BENCHMARK_LOG=openpilot_0_11_0_vision_run_pickle RUN_PICKLE=1 PYTHONPATH="." ASSERT_MIN_STEP_TIME=17 DEV=QCOM taskset -c 4-7 python3 examples/openpilot/compile3.py
- name: IR3 openpilot compile3 0.11.0 driving_vision
run: BENCHMARK_LOG=ir3_openpilot_0_11_0_vision PYTHONPATH="." ASSERT_MIN_STEP_TIME=18 DEV=QCOM:IR3 FLOAT16=1 IMAGE=1 NOLOCALS=1 taskset -c 4-7 python3 examples/openpilot/compile3.py https://github.com/commaai/openpilot/raw/v0.11.0/selfdrive/modeld/models/driving_vision.onnx
- name: openpilot compile3 0.11.0 driving_policy
run: BENCHMARK_LOG=openpilot_0_11_0_policy PYTHONPATH="." ASSERT_MIN_STEP_TIME=3.2 DEV=QCOM FLOAT16=1 IMAGE=1 NOLOCALS=1 taskset -c 4-7 python3 examples/openpilot/compile3.py https://github.com/commaai/openpilot/raw/v0.11.0/selfdrive/modeld/models/driving_policy.onnx
- name: openpilot compile3 0.11.0 dmonitoring
run: BENCHMARK_LOG=openpilot_0_11_0_dmonitoring PYTHONPATH="." ASSERT_MIN_STEP_TIME=11 DEV=QCOM FLOAT16=1 IMAGE=1 NOLOCALS=1 taskset -c 4-7 python3 examples/openpilot/compile3.py https://github.com/commaai/openpilot/raw/v0.11.0/selfdrive/modeld/models/dmonitoring_model.onnx
- name: compile
run: FLOAT16=1 IMAGE=1 taskset -c 4-7 python3 examples/openpilot/compile3.py ${{ matrix.url }}
- name: run pickle
run: BENCHMARK_LOG="${BENCHMARK_LOG}_run_pickle" RUN_PICKLE=1 taskset -c 4-7 python3 examples/openpilot/compile3.py
- name: Run process replay tests
uses: ./.github/actions/process-replay
@@ -489,15 +517,6 @@ jobs:
shell: bash -e -o pipefail {0}
if: github.repository_owner == 'tinygrad'
steps:
- name: Checkout Code
uses: actions/checkout@v6
- name: setup staging db
if: github.ref == 'refs/heads/update_benchmark_staging'
run: |
echo "CACHEDB=/tmp/staging.db" >> $GITHUB_ENV
rm -f /tmp/staging.db /tmp/staging.db-shm /tmp/staging.db-wal
- name: reset process replay
run: test/external/process_replay/reset.py
- name: Checkout Code
uses: actions/checkout@v6
- name: setup staging db
@@ -521,7 +540,7 @@ jobs:
testcommausbgpubenchmark:
name: UsbGPU Benchmark (comma)
runs-on: [self-hosted, Linux, comma4]
timeout-minutes: 20
timeout-minutes: 10
defaults:
run:
shell: bash -e -o pipefail {0}
@@ -541,7 +560,7 @@ jobs:
- name: openpilot run_pickle big_driving_supercombo
run: BENCHMARK_LOG=usbgpu_openpilot_big_driving_supercombo_run_pickle RUN_PICKLE=1 PICKLE_OOB=1 PYTHONPATH="." GMMU=0 DEV=USB+AMD ASSERT_MIN_STEP_TIME=50 python3 examples/openpilot/compile3.py - openpilot.pkl
- name: Test copy speeds
run: SIZE=64e6 PYTHONPATH=. GMMU=0 DEV=USB+AMD python3 test/external/external_test_usb_asm24.py TestDevCopySpeeds
run: SIZE=64000000 PYTHONPATH=. GMMU=0 DEV=USB+AMD python3 test/external/external_test_usb_asm24.py TestDevCopySpeeds
driverbenchmarks:
name: PCI Driver Benchmark (DEV=${{ matrix.dev }})
@@ -550,7 +569,7 @@ jobs:
fail-fast: false
matrix:
dev: ['AMD', 'NV']
timeout-minutes: 20
timeout-minutes: 5
defaults:
run:
shell: bash -e -o pipefail {0}
@@ -562,9 +581,8 @@ jobs:
uses: actions/checkout@v6
- name: Setup
run: |
./extra/amdpci/setup_python_cap.sh
./extra/hcq/hcq_smi.py ${{ matrix.dev == 'AMD' && 'amd' || 'nv' }} rmmod
./extra/hcq/hcq_smi.py ${{ matrix.dev == 'AMD' && 'amd' || 'nv' }} kill_pids
./extra/hcq/hcq_smi.py ${{ matrix.dev }} rmmod --expect
./extra/hcq/hcq_smi.py ${{ matrix.dev }} kill_pids --sudoless
mkdir -p extra/datasets
ln -s /raid/datasets/imagenet extra/datasets/imagenet
- name: setup staging db
@@ -599,6 +617,9 @@ jobs:
run: |
GRAPH_ONE_KERNEL=1 NSZ=8192 python3 test/speed/external_test_copy_speed.py TestCopySpeed.testCopyDefaulttoCPUJit
GRAPH_ONE_KERNEL=1 NSZ=8192 python3 test/speed/external_test_copy_speed.py TestCopySpeed.testCopyCPUtoDefaultJit
- name: HEVC Decode Benchmark
if: ${{ matrix.dev == 'NV' }}
run: IGNORE_BEAM_CACHE=1 VALIDATE=1 MAX_FRAMES=100 ASSERT_FPS=1400 JITBEAM=1 PYTHONPATH=. python3 extra/hevc/decode.py
- name: Run 10 MLPerf ResNet50 training steps (1 gpu)
if: ${{ matrix.dev == 'NV' }}
run: BENCHMARK_LOG=resnet_10steps MNISTMOCK=1 DEFAULT_FLOAT=HALF BENCHMARK=10 BS=256 GPUS=1 MODEL=resnet python3 examples/mlperf/model_train.py
@@ -621,12 +642,12 @@ jobs:
llvmspeed:
name: LLVM Speed
runs-on: [self-hosted, Linux, tinyboxrandom]
timeout-minutes: 20
timeout-minutes: 10
if: github.repository_owner == 'tinygrad'
steps:
- name: Checkout Code
uses: actions/checkout@v6
- name: Speed Test
run: DEV=CPU:LLVM THREADS=0 python3 test/speed/external_test_speed_v_torch.py
run: DEV=CPU:LLVM python3 test/speed/external_test_speed_v_torch.py
- name: Speed Test (BEAM=2)
run: BEAM=2 DEV=CPU:LLVM THREADS=0 python3 test/speed/external_test_speed_v_torch.py
run: IGNORE_BEAM_CACHE=1 BEAM=2 DEV=CPU:LLVM python3 test/speed/external_test_speed_v_torch.py