github/tinygrad
1
1
Fork 0
mirror of https://github.com/tinygrad/tinygrad.git synced 2026-01-10 07:28:15 -05:00
Code Issues Packages Projects Releases Wiki Activity

use BERT_LAYERS=2 in bert init (#9769)

Browse Source 
save 5 minut scheduling in setup so we can fit more search
This commit is contained in:
chenyu
2025-04-07 07:46:37 -04:00
committed by GitHub
parent 891322fd51
commit 3069ebfad1
2 changed files with 2 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
examples/mlperf/training_submission_v5.0/tinycorp/benchmarks/bert/implementations/tinybox_green/run_and_time.sh
View File

@@ -17,7 +17,7 @@ DATETIME=$(date "+%m%d%H%M")
LOGFILE="bert_green_${DATETIME}_${SEED}.log"
# init
BENCHMARK=10 INITMLPERF=1 BEAM_LOG_SURPASS_MAX=1 python3 examples/mlperf/model_train.py | tee $LOGFILE
BENCHMARK=10 INITMLPERF=1 BERT_LAYERS=2 python3 examples/mlperf/model_train.py | tee $LOGFILE
# run
PARALLEL=0 RUNMLPERF=1 python3 examples/mlperf/model_train.py | tee -a $LOGFILE

2
examples/mlperf/training_submission_v5.0/tinycorp/benchmarks/bert/implementations/tinybox_red/run_and_time.sh
View File

@@ -18,7 +18,7 @@ LOGFILE="bert_red_${DATETIME}_${SEED}.log"
# init
sudo rmmod amdgpu || true
BENCHMARK=10 INITMLPERF=1 BEAM_LOG_SURPASS_MAX=1 python3 examples/mlperf/model_train.py | tee $LOGFILE
BENCHMARK=10 INITMLPERF=1 BERT_LAYERS=2 python3 examples/mlperf/model_train.py | tee $LOGFILE
# run
# TODO: AMD driver hangs during init, but is 5% faster per step in real run.