issues
search
NVIDIA
/
NeMo-Aligner
Scalable toolkit for efficient model alignment
Apache License 2.0
522
stars
58
forks
source link
issues
Newest
Newest
Most commented
Recently updated
Oldest
Least commented
Least recently updated
Error during saving checkpoint with TensorRT-enabled PPO actor training
#281
haizadinia
opened
3 weeks ago
2
fix micro batch size logic in DPO
#280
gshennvm
closed
2 weeks ago
1
Implement KTO
#279
soumye
closed
2 weeks ago
4
[Question] TransfomerEngine and Apex dependencies
#278
peri044
opened
4 weeks ago
0
Add 405B numbers to readme
#277
gshennvm
closed
4 weeks ago
0
add comments about rlhf on readme
#276
gshennvm
closed
1 month ago
0
Add sequence packing support for SFTPackedDataset
#275
ashors1
closed
3 weeks ago
0
[TEST] sequence packing
#274
ashors1
closed
1 month ago
0
make build_dataloader not take in cfg
#273
gshennvm
opened
1 month ago
0
common class for aligner models
#272
gshennvm
opened
1 month ago
0
Request for Context Parallel Support in MegatronGPTDPOModel
#271
Wolfwjs
opened
1 month ago
0
Added "Annealed importance guidance" and DRaFT+ docs
#270
rohitrango
closed
3 weeks ago
2
call
#268
gshennvm
opened
1 month ago
0
Fix num_micro_batches when using forward_mbs
#267
shengyangs
closed
1 month ago
1
vm model guided inference
#266
arendu
opened
1 month ago
0
Does NeMo Aligner support tensor parallel and pipeline parallel?
#265
cizhenshi
opened
1 month ago
0
GPTGenerateTRTLLM.trt_llm_exporter.refit failed due to empty weights in the refit engine during PPO actor training
#264
renweizhukov
opened
1 month ago
1
support for mamba hybrid models
#263
arendu
opened
1 month ago
0
DPO Training error: NameError: name 'RetroConfig' is not defined
#262
sunilitggu
closed
1 month ago
1
DPO Training error: NameError: name 'RetroConfig' is not defined
#261
sunilitggu
closed
1 month ago
0
0.4 doc tech edit
#260
terrykong
closed
1 month ago
1
Support for Packed Sequence Dataset
#259
RadhaGulhane13
closed
1 month ago
1
Point the TRTLLM documentation to Nemo docs
#258
terrykong
closed
1 month ago
0
Point the TRTLLM documentation to Nemo docs
#257
terrykong
closed
2 months ago
0
Updates requirements to reflect necessary mcore version (>=0.8)
#256
terrykong
closed
2 months ago
0
0.4 mcore constraint
#255
terrykong
closed
2 months ago
0
Enforce that mcore is >=0.8 for Aligner 0.4.0
#254
terrykong
closed
2 months ago
1
Updates main Dockerfile
#253
terrykong
closed
1 month ago
1
0.4 dockerfile
#252
terrykong
closed
1 month ago
0
job hangs or IndexError when train reward model with PP> 1
#251
zirui
opened
2 months ago
5
How to shuffle data before the start of each epoch?
#250
Cppowboy
opened
2 months ago
0
Add Pickscore/HPSv2 style reward training and evaluation
#249
rohitrango
closed
1 month ago
2
fix HF links in readme
#248
gshennvm
closed
2 months ago
0
Fixes typo in apex installation
#247
terrykong
closed
2 months ago
0
Fixes nemofw doc links
#246
terrykong
closed
2 months ago
1
Addresses documentation bugs
#245
terrykong
closed
2 months ago
0
Addresses documentation bugs
#244
terrykong
closed
2 months ago
0
Different performance from TRL DPO
#243
Cppowboy
closed
2 months ago
1
Updates microbatch APIs to use megatron's instead of apex's
#242
terrykong
closed
2 months ago
0
Updates microbatch APIs to use megatron's instead of apex's
#241
terrykong
closed
2 months ago
0
Update the hash of the conversion script to include TE fix for mcore conversion
#240
terrykong
closed
2 months ago
0
Update the hash of the conversion script to include TE fix for mcore conversion
#239
terrykong
closed
2 months ago
1
Update the hash of the conversion script to include TE fix for mcore conversion
#238
terrykong
closed
2 months ago
0
sync: main to dev
#237
github-actions[bot]
closed
2 weeks ago
0
SFT not working on nemo:24.05.01 container
#236
vecorro
opened
2 months ago
0
Fixed minor float check bug in compute_num_steps_per_epoch
#235
trias702
closed
2 months ago
0
Avoids crash in PPOTrainer if using adafactor w/o learning rate
#234
terrykong
closed
2 months ago
0
Adds support for trainer.ppo.export_rollouts_jsonl
#233
terrykong
closed
2 months ago
1
Raise exceptions if using trtllm and use_Greedy in sampling params is set but temperature/topk are not correct for greedy
#232
terrykong
closed
2 months ago
0
better add_BOS and add_EOS support in reward models
#231
gshennvm
opened
2 months ago
0
Previous
Next