Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
44 commits
Select commit Hold shift + click to select a range
21b0540
Rebase changes onto correct git history
cmikeh2 Apr 12, 2023
bd74c32
Further generalize rotate half rotary position embeddings
cmikeh2 Apr 12, 2023
56e1de9
Upgrade local clang-format to match CI
cmikeh2 Apr 12, 2023
264d49b
Restore GeGLU behavior and template for SiLU. Add unit test.
cmikeh2 Apr 12, 2023
14c6a9b
Restore experimental qkv reset
cmikeh2 Apr 12, 2023
b01e7ea
Switch to named constant to improve readability
cmikeh2 Apr 12, 2023
0215867
Name refactor to align with functionality rather than implementation
cmikeh2 Apr 12, 2023
0f371d8
Finish hybrid engine integration
cmikeh2 Apr 13, 2023
48db17b
Fix for MLP dimensions
cmikeh2 Apr 13, 2023
ce19b9c
Refactor changes out of base container (#506)
cmikeh2 Apr 27, 2023
88a821d
Update explanations
cmikeh2 Apr 27, 2023
e42859b
Complete merge, fix BF16 integration
cmikeh2 Apr 28, 2023
6bad46d
Merge master
cmikeh2 May 1, 2023
eebfcdf
BF16_AVAILABLE should derive solely from the op_builder
cmikeh2 May 1, 2023
6ad5f0f
Merge remote-tracking branch 'public/master' into cholmes/llama-infer…
cmikeh2 May 1, 2023
63fe26f
Refactor on top of additional model support
cmikeh2 May 1, 2023
e9137d1
Guard is_bf16_supported check
cmikeh2 May 1, 2023
838b6f4
Even stronger guards
cmikeh2 May 1, 2023
a496dd1
Remove deprecated policy members
cmikeh2 May 1, 2023
123e195
Merge branch 'master' into cholmes/llama-inference-v2
cmikeh2 May 1, 2023
c670344
Another guard
cmikeh2 May 2, 2023
6830f5a
Merge branch 'cholmes/llama-inference-v2' of github.com:microsoft/Dee…
cmikeh2 May 2, 2023
708fb45
Bad check for BF16 support
cmikeh2 May 2, 2023
ace1967
Merge fix
cmikeh2 May 2, 2023
439f26f
Call correct parent func
cmikeh2 May 2, 2023
81c8e62
Revert API change for TP
cmikeh2 May 2, 2023
80a2c67
Reorder inheritance for Hybrid containers, provide justification
cmikeh2 May 2, 2023
4bd0c6e
Add missing policy kwargs
cmikeh2 May 2, 2023
a974832
Merge remote-tracking branch 'public/master' into cholmes/llama-infer…
cmikeh2 May 2, 2023
775282a
Remove debug code
cmikeh2 May 2, 2023
9725f09
BF16 model inference support
cmikeh2 May 2, 2023
e0e70fe
Remove debug code
cmikeh2 May 2, 2023
9d64515
Review feedback
cmikeh2 May 2, 2023
c94c67b
Merge branch 'master' into cholmes/llama-inference-v2
jeffra May 2, 2023
79ad7d0
Fix inheritance
cmikeh2 May 2, 2023
a044359
Merge branch 'cholmes/llama-inference-v2' of github.com:microsoft/Dee…
cmikeh2 May 2, 2023
5245e0a
don't use cache dir for torch installs
jeffra May 3, 2023
8c10a29
Merge branch 'master' into cholmes/llama-inference-v2
jeffra May 3, 2023
cd1b617
Align APIs
cmikeh2 May 3, 2023
149e193
Merge branch 'cholmes/llama-inference-v2' of github.com:microsoft/Dee…
cmikeh2 May 3, 2023
8adc056
add HE unit test for OPT
jeffra May 3, 2023
995c3ae
fix typo, missing policy ref to client module
jeffra May 3, 2023
1596640
Merge branch 'master' into cholmes/llama-inference-v2
jeffra May 3, 2023
d124817
Merge branch 'master' into cholmes/llama-inference-v2
jeffra May 4, 2023
File filter

Filter by extension

Filter by extension


Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion .github/workflows/nv-accelerate-v100.yml
Original file line number Diff line number Diff line change
Expand Up @@ -30,7 +30,7 @@ jobs:

- name: Install pytorch
run: |
pip install torch torchvision --extra-index-url https://download.pytorch.org/whl/cu111
pip install --no-cache-dir torch torchvision --extra-index-url https://download.pytorch.org/whl/cu111
python -c "import torch; print('torch:', torch.__version__, torch)"
python -c "import torch; print('CUDA available:', torch.cuda.is_available())"

Expand Down
2 changes: 1 addition & 1 deletion .github/workflows/nv-inference.yml
Original file line number Diff line number Diff line change
Expand Up @@ -30,7 +30,7 @@ jobs:

- name: Install pytorch
run: |
pip install torch==1.13.1 torchvision --extra-index-url https://download.pytorch.org/whl/cu116
pip install --no-cache-dir torch==1.13.1 torchvision --extra-index-url https://download.pytorch.org/whl/cu116
python -c "import torch; print('torch:', torch.__version__, torch)"
python -c "import torch; print('CUDA available:', torch.cuda.is_available())"

Expand Down
2 changes: 1 addition & 1 deletion .github/workflows/nv-lightning-v100.yml
Original file line number Diff line number Diff line change
Expand Up @@ -30,7 +30,7 @@ jobs:

- name: Install pytorch
run: |
pip install torch==1.9.1+cu111 torchvision==0.10.1+cu111 torchaudio==0.9.1 -f https://download.pytorch.org/whl/torch_stable.html
pip install --no-cache-dir torch==1.9.1+cu111 torchvision==0.10.1+cu111 torchaudio==0.9.1 -f https://download.pytorch.org/whl/torch_stable.html
python -c "import torch; print('torch:', torch.__version__, torch)"
python -c "import torch; print('CUDA available:', torch.cuda.is_available())"

Expand Down
2 changes: 1 addition & 1 deletion .github/workflows/nv-megatron.yml
Original file line number Diff line number Diff line change
Expand Up @@ -30,7 +30,7 @@ jobs:

- name: Install pytorch
run: |
pip install torch==1.13.1 torchvision --extra-index-url https://download.pytorch.org/whl/cu116
pip install --no-cache-dir torch==1.13.1 torchvision --extra-index-url https://download.pytorch.org/whl/cu116
python -c "import torch; print('torch:', torch.__version__, torch)"
python -c "import torch; print('CUDA available:', torch.cuda.is_available())"

Expand Down
2 changes: 1 addition & 1 deletion .github/workflows/nv-mii.yml
Original file line number Diff line number Diff line change
Expand Up @@ -30,7 +30,7 @@ jobs:

- name: Install pytorch
run: |
pip install torch==1.13.1 torchvision --extra-index-url https://download.pytorch.org/whl/cu116
pip install --no-cache-dir torch==1.13.1 torchvision --extra-index-url https://download.pytorch.org/whl/cu116
python -c "import torch; print('torch:', torch.__version__, torch)"
python -c "import torch; print('CUDA available:', torch.cuda.is_available())"

Expand Down
2 changes: 1 addition & 1 deletion .github/workflows/nv-torch19-p40.yml
Original file line number Diff line number Diff line change
Expand Up @@ -30,7 +30,7 @@ jobs:

- name: Install pytorch
run: |
pip install torch==1.9.0+cu111 torchvision==0.10.0+cu111 -f https://download.pytorch.org/whl/torch_stable.html
pip install --no-cache-dir torch==1.9.0+cu111 torchvision==0.10.0+cu111 -f https://download.pytorch.org/whl/torch_stable.html
python -c "import torch; print('torch:', torch.__version__, torch)"
python -c "import torch; print('CUDA available:', torch.cuda.is_available())"

Expand Down
2 changes: 1 addition & 1 deletion .github/workflows/nv-torch19-v100.yml
Original file line number Diff line number Diff line change
Expand Up @@ -30,7 +30,7 @@ jobs:

- name: Install pytorch
run: |
pip install torch==1.9.0+cu111 torchvision==0.10.0+cu111 -f https://download.pytorch.org/whl/torch_stable.html
pip install --no-cache-dir torch==1.9.0+cu111 torchvision==0.10.0+cu111 -f https://download.pytorch.org/whl/torch_stable.html
python -c "import torch; print('torch:', torch.__version__, torch)"
python -c "import torch; print('CUDA available:', torch.cuda.is_available())"

Expand Down
2 changes: 1 addition & 1 deletion .github/workflows/nv-transformers-v100.yml
Original file line number Diff line number Diff line change
Expand Up @@ -31,7 +31,7 @@ jobs:
- name: Install pytorch
run: |
# use the same pytorch version as transformers CI
pip install torch torchvision torchaudio -f https://download.pytorch.org/whl/torch_stable.html
pip install --no-cache-dir torch torchvision torchaudio -f https://download.pytorch.org/whl/torch_stable.html
python -c "import torch; print('torch:', torch.__version__, torch)"
python -c "import torch; print('CUDA available:', torch.cuda.is_available())"

Expand Down
1 change: 0 additions & 1 deletion csrc/includes/ds_kernel_utils.h
Original file line number Diff line number Diff line change
Expand Up @@ -34,7 +34,6 @@ constexpr int hw_warp_size = 32;

#if __CUDA_ARCH__ >= 800
#define ASYNC_COPY_AVAILABLE
#define BF16_AVAILABLE
#endif // __CUDA_ARCH__ >= 800

#include <cooperative_groups.h>
Expand Down
Loading