issues
search
awslabs
/
sockeye
Sequence-to-sequence framework with a focus on Neural Machine Translation based on PyTorch
https://awslabs.github.io/sockeye/
Apache License 2.0
1.21k
stars
324
forks
source link
issues
Newest
Newest
Most commented
Recently updated
Oldest
Least commented
Least recently updated
Add maintenance mode notice to README.
#1114
mjdenkowski
closed
3 months ago
3
Wrong parameter when creating vocabularies in sockeye.train.create_data_iters_and_vocabs
#1113
iPRET
opened
3 months ago
0
Use `weights_only=True` when loading SockeyeModel parameters.
#1112
mjdenkowski
closed
3 months ago
1
`accuracy` metric is no longer supported
#1111
AmitMY
opened
3 months ago
0
Factors inference is slow (3 seconds/token) on A100 GPU
#1110
AmitMY
opened
3 months ago
6
Missing `--softmax-temperature T`
#1109
AmitMY
opened
3 months ago
1
Add alignment matrix learning
#1108
iPRET
closed
3 months ago
4
Allow easier programmatic use by extracting code from `translate.py`
#1107
AmitMY
closed
3 months ago
2
Out-of-memory when using target factors
#1106
AmitMY
opened
6 months ago
4
Guided alignments in Sockeye.
#1105
tomsbergmanis
opened
6 months ago
6
feat(evaluate): add signwriting evaluation metrics
#1104
AmitMY
closed
3 months ago
2
Adding custom metrics
#1103
AmitMY
opened
7 months ago
3
IndexError when I try to continue training
#1102
chrkell
opened
10 months ago
1
TracerWarning while translating but no error showing
#1101
chrkell
opened
10 months ago
1
Update url for nightly builds.
#1100
fhieber
closed
10 months ago
0
Update README.md w/ recent publications using Sockeye
#1099
fhieber
closed
10 months ago
0
sockeye.utils.SockeyeError: Target sequences are not token-parallel:
#1098
jingshu-liu
opened
11 months ago
1
Fine Tuning on Different Data size mismatch for embedding_target
#1097
AmitMY
closed
1 year ago
1
beam_search fix for running with torch.use_deterministic_algorithms(True)
#1096
Jehovan
closed
1 year ago
0
`learning_rate_scheduler_type: inv-sqrt-decay` not reaching `initial_learning_rate`
#1095
SamuelLarkin
closed
1 year ago
3
Refactor inference optimizations
#1094
xingniu
closed
1 year ago
0
Use fault tolerant symlink during training to handle temporary file system inconsistencies.
#1093
mjdenkowski
closed
1 year ago
0
Automatic termination of training
#1092
SKT-T1-Thecai
closed
1 year ago
2
fix typos in testset names, allow symlinked datadir
#1091
thammegowda
closed
1 year ago
0
pca + testset overlap scripts
#1090
thompsonb
closed
1 year ago
0
How to finetune no-factor translation model with factors?
#1089
SKT-T1-Thecai
closed
1 year ago
1
ppl decreasing very slow
#1088
muramaso
closed
1 year ago
3
Fixes to blocking cross-attention
#1087
xingniu
closed
1 year ago
0
Two fixes to SampleK
#1086
tdomhan
closed
1 year ago
0
Add blocking cross-attention between decoder and encoded prepended tokens
#1085
xingniu
closed
1 year ago
0
RuntimeError: UNSUPPORTED DTYPE (sockeye-translate --use-cpu --dtype bfloat16)
#1084
SamuelLarkin
opened
1 year ago
0
RuntimeError: "LayerNormKernelImpl" not implemented for 'Half' (sockeye==3.1.27)
#1083
SamuelLarkin
closed
1 year ago
2
Modified factor decoding for automatic dubbing
#1082
Proyag
closed
3 months ago
0
UnicodeEncodeError: 'charmap' codec can't encode character
#1081
Verkolle
closed
1 year ago
2
ValueError: setting an array element with a sequence.
#1080
Verkolle
closed
1 year ago
5
BFloat16 Integration/System Test Update
#1079
mjdenkowski
closed
1 year ago
0
Specif or omit --shared-vocab consistently when training and preparing the data error
#1078
NLP-Peter
closed
1 year ago
2
Enable bfloat16 inference and fix unrelated NumPy issue
#1077
mjdenkowski
closed
1 year ago
0
Code cleanup: refactoring, type checking, and formatting
#1076
mjdenkowski
closed
1 year ago
0
Update to sacrebleu 2.3.1
#1075
fhieber
closed
1 year ago
0
Update to PyTorch 1.13 (+ 1.14 compatibility)
#1074
fhieber
closed
1 year ago
1
tok/sec throughput
#1073
vince62s
closed
1 year ago
3
OSError: libcudnn.so.7: cannot open shared object file: No such file or directory
#1072
nitin-bansal46
closed
1 year ago
1
Update DeepSpeed checkpoint conversion to support newer DeepSpeed versions
#1071
mjdenkowski
closed
1 year ago
0
How to avoid out of memory errors on large models
#1070
ejkitchen
closed
1 year ago
3
translation speed with quantization into int8
#1069
ZhenYangIACAS
closed
1 year ago
1
Change decoder softmax size logging level from info to debug.
#1068
mjdenkowski
closed
1 year ago
0
AssertionError: If capturable=False, state_steps should not be CUDA tensors.
#1067
SamuelLarkin
closed
1 year ago
3
Add --tf32 device flag for transparent float32 acceleration
#1066
graehl
closed
1 year ago
5
BrokenPipeError: [Errno 32]
#1065
whher
closed
1 year ago
6
Next