Commits · 69a546be21cef5e72000295275a23cb7491d26cc · OpenDAS / Megatron-LM

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
24 Jun, 2020 1 commit
- More ict_merge changes and interactive testing · 3354081f
  Neel Kant authored Jun 23, 2020
  
  3354081f
22 Jun, 2020 1 commit
- Address most of comments from code review w/ Mohammad · 6495927e
  Neel Kant authored Jun 22, 2020
  
  6495927e
10 Jun, 2020 1 commit
- Ran and cleaned up · fcc500d6
  Neel Kant authored Jun 10, 2020
  
  fcc500d6
09 Jun, 2020 1 commit
- Correct retrieval utility and add salient span preprocessing · bf599e86
  Neel Kant authored Jun 08, 2020
  
  bf599e86
05 Jun, 2020 4 commits
- addressed jareds comments · 197c132e
  mohammad authored Jun 05, 2020
  
  197c132e
- added fp16 lm cross entropy to bert · 78022005
  mohammad authored Jun 05, 2020
  
  78022005
- added fp16 cross entropy loss option for gpt2 · 22e3c7e6
  mohammad authored Jun 05, 2020
  
  22e3c7e6
- added fp16 cross entropy loss option for gpt2 · acfe848e
  mohammad authored Jun 05, 2020
  
  acfe848e