Commits · f5eac3d19b89991d2e2c88f50efeb0fe5cc69a45 · OpenDAS / Megatron-LM

05 Jan, 2021 1 commit
- Only create task heads on last pipeline stage. · f772fbc9
  Jared Casper authored Jan 05, 2021
  
  f772fbc9
27 Dec, 2020 1 commit
- moved module to model and removed fp16 · b84d7a90
  mohammad authored Dec 26, 2020
  
  b84d7a90
25 Dec, 2020 1 commit
- moved entire optimizer build and tested · 28062e14
  mohammad authored Dec 24, 2020
  
  28062e14
22 Dec, 2020 3 commits
- Address Deepak's comments · 83671bbf
  mshoeybi authored Dec 21, 2020
  
  83671bbf
- Add residual connection in fp32 machinery · 8bed1d63
  mohammad authored Dec 21, 2020
  
  8bed1d63
- Add the option for fp32 residual connection (fp32 residual connection... · 62632d39
  mohammad authored Dec 21, 2020
```
Add the option for fp32 residual connection (fp32 residual connection machinery still needs to be added)
```
  62632d39
19 Dec, 2020 1 commit
- Add pipelining to GLUE and RACE tasks · caa9dca5
  Jared Casper authored Nov 30, 2020
  
  caa9dca5
17 Nov, 2020 1 commit
- Update code used for finetuning to latest API. · b219ff00
  Jared Casper authored Nov 16, 2020
  
  b219ff00
12 Nov, 2020 9 commits
- Small bugfix in bert_model.py: make sure word_embeddings is initialized before... · 69a546be
  Deepak Narayanan authored Nov 12, 2020
```
Small bugfix in bert_model.py: make sure word_embeddings is initialized before instantiating lm_head
```
  69a546be
- Refactor code according to Jared's comments: move pipelining and... · 1979c242
  Deepak Narayanan authored Nov 12, 2020
```
Refactor code according to Jared's comments: move pipelining and non-pipelining training loops into separate methods

Also, use mpu.get_*_model_parallel_size() instead of args.*_model_parallel_size
```
  1979c242
- Only transpose hidden_states when necessary · 6abf39be
  Deepak Narayanan authored Nov 03, 2020
  
  6abf39be
- Refactor word_embeddings_weight() logic into separate method, and other Mohammad comments · 57c3b364
  Deepak Narayanan authored Nov 03, 2020
  
  57c3b364
- Clarifications in comments and minor refactoring to make main training loop more readable · 8fb2bc8c
  Deepak Narayanan authored Oct 28, 2020
  
  8fb2bc8c
- Remove unused parameter sharing logic · 1271fd73
  Deepak Narayanan authored Oct 28, 2020
  
  1271fd73
- Removal of unneeded changes so that diff is smaller · 275d4e64
  Deepak Narayanan authored Oct 20, 2020
  
  275d4e64
- Intra-layer MP -> Tensor MP, Inter-layer MP -> Pipeline MP · 52a5f2f2
  Deepak Narayanan authored Oct 20, 2020
  
  52a5f2f2
- Pipeline parallelism implementation with periodic full-pipeline syncs · 7abd3e90
  Deepak Narayanan authored Aug 29, 2020
```
Also includes following changes for inter-layer model-parallel implementation:
- Refactoring of model implementations
- Training loop changes to support inter-layer communication using `ring_exchange`
- New groups for inter-layer communication
- Checkpoint changes
- Command line arguments
```
  7abd3e90
13 Oct, 2020 2 commits
- typo fix · 01dffdb5
  Vijay Korthikanti authored Oct 12, 2020
  
  01dffdb5
- address review comments · 0bc75448
  Vijay Korthikanti authored Oct 12, 2020
  
  0bc75448
12 Oct, 2020 2 commits
- Incrementing checkpoint version to 2.0 · 9b0083ea
  Vijay Korthikanti authored Oct 12, 2020
  
  9b0083ea
- reordering perf fix · ee327acd
  Vijay Korthikanti authored Oct 12, 2020
  
  ee327acd
09 Oct, 2020 1 commit
- support for different query key sequence lengths · d763f54a
  Vijay Korthikanti authored Oct 08, 2020
  
  d763f54a
30 Sep, 2020 1 commit
- checkpoint versioning · a8d47812
  Vijay Korthikanti authored Sep 29, 2020
  
  a8d47812
29 Sep, 2020 2 commits
- bug fix · 7d4ad51e
  Vijay Korthikanti authored Sep 29, 2020
  
  7d4ad51e
- support for old checkpoint format · 37ae6646
  Vijay Korthikanti authored Sep 28, 2020
  
  37ae6646
18 Sep, 2020 1 commit
- scaling-generalized masking-softmax fusion kernel · d73def35
  root authored Sep 18, 2020
  
  d73def35
11 Sep, 2020 1 commit
- Various speed optimizations. · 51a2e6b0
  Vijay Korthikanti authored Sep 11, 2020
  
  51a2e6b0
02 Sep, 2020 1 commit
- added splitting checkpointed activations · d6c4248b
  mohammad authored Sep 02, 2020
  
  d6c4248b
06 Aug, 2020 1 commit
- Implementing lazy parallel initialization · 5c04ceb3
  Boris Fomitchev authored Aug 05, 2020
```
Signed-off-by: Boris Fomitchev <bfomitchev@nvidia.com>
```
  5c04ceb3
29 Jul, 2020 1 commit
- Restored @script for openai_gelu · bb7c638f
  Boris Fomitchev authored Jul 29, 2020
```
Signed-off-by: Boris Fomitchev <bfomitchev@nvidia.com>
```
  bb7c638f
28 Jul, 2020 1 commit
- Replacing --erf-gelu option with explicit --onnx-safe option · fe794c5a
  Boris Fomitchev authored Jul 27, 2020
```
Signed-off-by: Boris Fomitchev <bfomitchev@nvidia.com>
```
  fe794c5a
26 Jul, 2020 2 commits
- fixing copy-paste · 7917774a
  Boris Fomitchev authored Jul 25, 2020
  
  7917774a
- Complete fix for APEX absence in NeMo · aa0ee72e
  Boris Fomitchev authored Jul 25, 2020
  
  aa0ee72e
24 Jul, 2020 3 commits
- Fixed APEX test · 8f0af7cd
  Boris Fomitchev authored Jul 24, 2020
```
Signed-off-by: Boris Fomitchev <bfomitchev@nvidia.com>
```
  8f0af7cd
- Warning added · b04eb0a2
  Boris Fomitchev authored Jul 23, 2020
  
  b04eb0a2
- ONNX export fix, including one for gelu() · 158a99c3
  Boris Fomitchev authored Jul 23, 2020
```
Signed-off-by: Boris Fomitchev <bfomitchev@nvidia.com>
```
  158a99c3
23 Jul, 2020 1 commit
- Address comments from last week · 11f76cd3
  Neel Kant authored Jul 22, 2020
  
  11f76cd3
22 Jul, 2020 1 commit
- Initialization fixes: allowing simple case like pytest pass, also making apex optional · 9026b86d
  Boris Fomitchev authored Jul 22, 2020
```
Signed-off-by: Boris Fomitchev <bfomitchev@nvidia.com>
```
  9026b86d
07 Jul, 2020 1 commit
- Addressed Jared's comments · 8d7f508a
  Neel Kant authored Jul 06, 2020
  
  8d7f508a
29 Jun, 2020 1 commit
- Create tensors on cuda rather than copying · 76928caa
  Neel Kant authored Jun 29, 2020
  
  76928caa