diff options
| author | Roman Grundkiewicz <rgrundkiewicz@gmail.com> | 2020-11-13 20:40:28 +0300 |
|---|---|---|
| committer | Roman Grundkiewicz <rgrundkiewicz@gmail.com> | 2020-11-13 20:40:28 +0300 |
| commit | 8d5756fe53745978f8c5c9c51c25d52357d03095 (patch) | |
| tree | 802eb77d745e05dcf6db000c3415a03129424e35 /tests | |
| parent | dee40c7b413120536179c6612cafe46160b24aa4 (diff) | |
Use clip-norm 0 in training/multi-gpu
Diffstat (limited to 'tests')
4 files changed, 14 insertions, 14 deletions
diff --git a/tests/training/multi-gpu/sync_sgd_1gpu.expected b/tests/training/multi-gpu/sync_sgd_1gpu.expected index 73c9503..b826610 100644 --- a/tests/training/multi-gpu/sync_sgd_1gpu.expected +++ b/tests/training/multi-gpu/sync_sgd_1gpu.expected @@ -1,4 +1,4 @@ -236.64883423 -197.38874817 -198.74374390 -183.10134888 +230.86734009 +176.34066772 +178.55038452 +161.01515198 diff --git a/tests/training/multi-gpu/sync_sgd_1gpu_expsmooth.expected b/tests/training/multi-gpu/sync_sgd_1gpu_expsmooth.expected index 0600ba2..b826610 100644 --- a/tests/training/multi-gpu/sync_sgd_1gpu_expsmooth.expected +++ b/tests/training/multi-gpu/sync_sgd_1gpu_expsmooth.expected @@ -1,4 +1,4 @@ -236.64883423 -197.38874817 -198.74374390 -183.10137939 +230.86734009 +176.34066772 +178.55038452 +161.01515198 diff --git a/tests/training/multi-gpu/test_sync_sgd_1gpu.sh b/tests/training/multi-gpu/test_sync_sgd_1gpu.sh index 2b9c72b..a1a4453 100644 --- a/tests/training/multi-gpu/test_sync_sgd_1gpu.sh +++ b/tests/training/multi-gpu/test_sync_sgd_1gpu.sh @@ -8,8 +8,8 @@ rm -rf sync_sgd_1gpu sync_sgd_1gpu.log mkdir -p sync_sgd_1gpu $MRT_MARIAN/marian \ - --no-shuffle --seed 888 --mini-batch 4 --maxi-batch 1 --maxi-batch-sort none \ - --dim-rnn 64 --dim-emb 32 --learn-rate 0.1 \ + --no-shuffle --clip-norm 0 --seed 888 --mini-batch 4 --maxi-batch 1 --maxi-batch-sort none \ + --dim-rnn 64 --dim-emb 32 --learn-rate 0.02 \ --devices 0 --sync-sgd --optimizer sgd --cost-type ce-mean \ -m sync_sgd_1gpu/model.npz -t $MRT_DATA/europarl.de-en/corpus.bpe.{en,de} -v vocab.en.yml vocab.de.yml \ --disp-freq 5 --save-freq 10 --after-batches 20 \ @@ -19,7 +19,7 @@ test -e sync_sgd_1gpu/model.npz test -e sync_sgd_1gpu.log cat sync_sgd_1gpu.log | $MRT_TOOLS/extract-costs.sh > sync_sgd_1gpu.out -$MRT_TOOLS/diff-nums.py sync_sgd_1gpu.out sync_sgd_1gpu.expected -o sync_sgd_1gpu.diff +$MRT_TOOLS/diff-nums.py -p 0.02 sync_sgd_1gpu.out sync_sgd_1gpu.expected -o sync_sgd_1gpu.diff # Exit with success code exit 0 diff --git a/tests/training/multi-gpu/test_sync_sgd_1gpu_expsmooth.sh b/tests/training/multi-gpu/test_sync_sgd_1gpu_expsmooth.sh index d1d77d3..586a57a 100644 --- a/tests/training/multi-gpu/test_sync_sgd_1gpu_expsmooth.sh +++ b/tests/training/multi-gpu/test_sync_sgd_1gpu_expsmooth.sh @@ -8,8 +8,8 @@ rm -rf sync_sgd_1gpu_expsmooth sync_sgd_1gpu_expsmooth.log mkdir -p sync_sgd_1gpu_expsmooth $MRT_MARIAN/marian \ - --no-shuffle --seed 888 --mini-batch 4 --maxi-batch 1 --maxi-batch-sort none \ - --dim-rnn 64 --dim-emb 32 --learn-rate 0.1 \ + --no-shuffle --clip-norm 0 --seed 888 --mini-batch 4 --maxi-batch 1 --maxi-batch-sort none \ + --dim-rnn 64 --dim-emb 32 --learn-rate 0.02 \ --devices 0 --sync-sgd --optimizer sgd --exponential-smoothing --cost-type ce-mean \ -m sync_sgd_1gpu_expsmooth/model.npz -t $MRT_DATA/europarl.de-en/corpus.bpe.{en,de} -v vocab.en.yml vocab.de.yml \ --disp-freq 5 --save-freq 10 --after-batches 20 \ @@ -19,7 +19,7 @@ test -e sync_sgd_1gpu_expsmooth/model.npz test -e sync_sgd_1gpu_expsmooth.log cat sync_sgd_1gpu_expsmooth.log | $MRT_TOOLS/extract-costs.sh > sync_sgd_1gpu_expsmooth.out -$MRT_TOOLS/diff-nums.py sync_sgd_1gpu_expsmooth.out sync_sgd_1gpu_expsmooth.expected -o sync_sgd_1gpu_expsmooth.diff +$MRT_TOOLS/diff-nums.py -p 0.02 sync_sgd_1gpu_expsmooth.out sync_sgd_1gpu_expsmooth.expected -o sync_sgd_1gpu_expsmooth.diff # Exit with success code exit 0 |