Skip to content

Add a simple sdpa#3037

Closed
cccclai wants to merge 7 commits intogh/cccclai/4/basefrom
gh/cccclai/4/head
Closed

Add a simple sdpa#3037
cccclai wants to merge 7 commits intogh/cccclai/4/basefrom
gh/cccclai/4/head

Conversation

@cccclai
Copy link
Copy Markdown
Contributor

@cccclai cccclai commented Apr 14, 2024 •
edited
Loading

Stack from ghstack (oldest at bottom):

Add a simple sdpa so it's decomposed to simpler ops instead of the decompose F.scaled_dot_product_attention, which includes 29 ops including torch.where

def forward(self, q, k, v):
    aten_mul_scalar = executorch_exir_dialects_edge__ops_aten_mul_Scalar(q, 0.5946035575013605);  q = None
    aten_full_default = executorch_exir_dialects_edge__ops_aten_full_default([8, 8], True, dtype = torch.bool, layout = torch.strided, device = device(type='cpu'), pin_memory = False)
    aten_arange_start_step = executorch_exir_dialects_edge__ops_aten_arange_start_step(0, 8, layout = torch.strided, device = device(type='cpu'), pin_memory = False)
    aten_unsqueeze_copy_default = executorch_exir_dialects_edge__ops_aten_unsqueeze_copy_default(aten_arange_start_step, -2);  aten_arange_start_step = None
    aten_arange_start_step_1 = executorch_exir_dialects_edge__ops_aten_arange_start_step(0, 8, layout = torch.strided, device = device(type='cpu'), pin_memory = False)
    aten_unsqueeze_copy_default_1 = executorch_exir_dialects_edge__ops_aten_unsqueeze_copy_default(aten_arange_start_step_1, -1);  aten_arange_start_step_1 = None
    aten_sub_tensor = executorch_exir_dialects_edge__ops_aten_sub_Tensor(aten_unsqueeze_copy_default, aten_unsqueeze_copy_default_1);  aten_unsqueeze_copy_default = aten_unsqueeze_copy_default_1 = None
    aten_le_scalar = executorch_exir_dialects_edge__ops_aten_le_Scalar(aten_sub_tensor, 0);  aten_sub_tensor = None
    aten_logical_and_default = executorch_exir_dialects_edge__ops_aten_logical_and_default(aten_le_scalar, aten_full_default);  aten_le_scalar = aten_full_default = None
    aten_full_like_default = executorch_exir_dialects_edge__ops_aten_full_like_default(aten_logical_and_default, 0, dtype = torch.float32, pin_memory = False, memory_format = torch.preserve_format)
    aten_logical_not_default = executorch_exir_dialects_edge__ops_aten_logical_not_default(aten_logical_and_default);  aten_logical_and_default = None
    aten_scalar_tensor_default = executorch_exir_dialects_edge__ops_aten_scalar_tensor_default(-inf, dtype = torch.float32, layout = torch.strided, device = device(type='cpu'))
    aten_where_self = executorch_exir_dialects_edge__ops_aten_where_self(aten_logical_not_default, aten_scalar_tensor_default, aten_full_like_default);  aten_logical_not_default = aten_scalar_tensor_default = aten_full_like_default = None
    aten_permute_copy_default = executorch_exir_dialects_edge__ops_aten_permute_copy_default(k, [0, 1, 3, 2]);  k = None
    aten_mul_scalar_1 = executorch_exir_dialects_edge__ops_aten_mul_Scalar(aten_permute_copy_default, 0.5946035575013605);  aten_permute_copy_default = None
    aten_expand_copy_default = executorch_exir_dialects_edge__ops_aten_expand_copy_default(aten_mul_scalar, [1, 1, 8, 8]);  aten_mul_scalar = None
    aten_view_copy_default = executorch_exir_dialects_edge__ops_aten_view_copy_default(aten_expand_copy_default, [1, 8, 8]);  aten_expand_copy_default = None
    aten_expand_copy_default_1 = executorch_exir_dialects_edge__ops_aten_expand_copy_default(aten_mul_scalar_1, [1, 1, 8, 8]);  aten_mul_scalar_1 = None
    aten_view_copy_default_1 = executorch_exir_dialects_edge__ops_aten_view_copy_default(aten_expand_copy_default_1, [1, 8, 8]);  aten_expand_copy_default_1 = None
    aten_bmm_default = executorch_exir_dialects_edge__ops_aten_bmm_default(aten_view_copy_default, aten_view_copy_default_1);  aten_view_copy_default = aten_view_copy_default_1 = None
    aten_view_copy_default_2 = executorch_exir_dialects_edge__ops_aten_view_copy_default(aten_bmm_default, [1, 1, 8, 8]);  aten_bmm_default = None
    aten_add_tensor = executorch_exir_dialects_edge__ops_aten_add_Tensor(aten_view_copy_default_2, aten_where_self);  aten_view_copy_default_2 = aten_where_self = None
    aten__softmax_default = executorch_exir_dialects_edge__ops_aten__softmax_default(aten_add_tensor, -1, False);  aten_add_tensor = None
    aten_expand_copy_default_2 = executorch_exir_dialects_edge__ops_aten_expand_copy_default(aten__softmax_default, [1, 1, 8, 8]);  aten__softmax_default = None
    aten_view_copy_default_3 = executorch_exir_dialects_edge__ops_aten_view_copy_default(aten_expand_copy_default_2, [1, 8, 8]);  aten_expand_copy_default_2 = None
    aten_expand_copy_default_3 = executorch_exir_dialects_edge__ops_aten_expand_copy_default(v, [1, 1, 8, 8]);  v = None
    aten_view_copy_default_4 = executorch_exir_dialects_edge__ops_aten_view_copy_default(aten_expand_copy_default_3, [1, 8, 8]);  aten_expand_copy_default_3 = None
    aten_bmm_default_1 = executorch_exir_dialects_edge__ops_aten_bmm_default(aten_view_copy_default_3, aten_view_copy_default_4);  aten_view_copy_default_3 = aten_view_copy_default_4 = None
    aten_view_copy_default_5 = executorch_exir_dialects_edge__ops_aten_view_copy_default(aten_bmm_default_1, [1, 1, 8, 8]);  aten_bmm_default_1 = None
    return (aten_view_copy_default_5,)

Differential Revision: D56119737

@cccclai
Add a simple sdpa
3bf691b 
Add a simple sdpa so it's decomposed to simpler ops instead of the decompose F.scaled_dot_product_attention, which includes 29 ops including `torch.where`
```
def forward(self, q, k, v):
    aten_mul_scalar = executorch_exir_dialects_edge__ops_aten_mul_Scalar(q, 0.5946035575013605);  q = None
    aten_full_default = executorch_exir_dialects_edge__ops_aten_full_default([8, 8], True, dtype = torch.bool, layout = torch.strided, device = device(type='cpu'), pin_memory = False)
    aten_arange_start_step = executorch_exir_dialects_edge__ops_aten_arange_start_step(0, 8, layout = torch.strided, device = device(type='cpu'), pin_memory = False)
    aten_unsqueeze_copy_default = executorch_exir_dialects_edge__ops_aten_unsqueeze_copy_default(aten_arange_start_step, -2);  aten_arange_start_step = None
    aten_arange_start_step_1 = executorch_exir_dialects_edge__ops_aten_arange_start_step(0, 8, layout = torch.strided, device = device(type='cpu'), pin_memory = False)
    aten_unsqueeze_copy_default_1 = executorch_exir_dialects_edge__ops_aten_unsqueeze_copy_default(aten_arange_start_step_1, -1);  aten_arange_start_step_1 = None
    aten_sub_tensor = executorch_exir_dialects_edge__ops_aten_sub_Tensor(aten_unsqueeze_copy_default, aten_unsqueeze_copy_default_1);  aten_unsqueeze_copy_default = aten_unsqueeze_copy_default_1 = None
    aten_le_scalar = executorch_exir_dialects_edge__ops_aten_le_Scalar(aten_sub_tensor, 0);  aten_sub_tensor = None
    aten_logical_and_default = executorch_exir_dialects_edge__ops_aten_logical_and_default(aten_le_scalar, aten_full_default);  aten_le_scalar = aten_full_default = None
    aten_full_like_default = executorch_exir_dialects_edge__ops_aten_full_like_default(aten_logical_and_default, 0, dtype = torch.float32, pin_memory = False, memory_format = torch.preserve_format)
    aten_logical_not_default = executorch_exir_dialects_edge__ops_aten_logical_not_default(aten_logical_and_default);  aten_logical_and_default = None
    aten_scalar_tensor_default = executorch_exir_dialects_edge__ops_aten_scalar_tensor_default(-inf, dtype = torch.float32, layout = torch.strided, device = device(type='cpu'))
    aten_where_self = executorch_exir_dialects_edge__ops_aten_where_self(aten_logical_not_default, aten_scalar_tensor_default, aten_full_like_default);  aten_logical_not_default = aten_scalar_tensor_default = aten_full_like_default = None
    aten_permute_copy_default = executorch_exir_dialects_edge__ops_aten_permute_copy_default(k, [0, 1, 3, 2]);  k = None
    aten_mul_scalar_1 = executorch_exir_dialects_edge__ops_aten_mul_Scalar(aten_permute_copy_default, 0.5946035575013605);  aten_permute_copy_default = None
    aten_expand_copy_default = executorch_exir_dialects_edge__ops_aten_expand_copy_default(aten_mul_scalar, [1, 1, 8, 8]);  aten_mul_scalar = None
    aten_view_copy_default = executorch_exir_dialects_edge__ops_aten_view_copy_default(aten_expand_copy_default, [1, 8, 8]);  aten_expand_copy_default = None
    aten_expand_copy_default_1 = executorch_exir_dialects_edge__ops_aten_expand_copy_default(aten_mul_scalar_1, [1, 1, 8, 8]);  aten_mul_scalar_1 = None
    aten_view_copy_default_1 = executorch_exir_dialects_edge__ops_aten_view_copy_default(aten_expand_copy_default_1, [1, 8, 8]);  aten_expand_copy_default_1 = None
    aten_bmm_default = executorch_exir_dialects_edge__ops_aten_bmm_default(aten_view_copy_default, aten_view_copy_default_1);  aten_view_copy_default = aten_view_copy_default_1 = None
    aten_view_copy_default_2 = executorch_exir_dialects_edge__ops_aten_view_copy_default(aten_bmm_default, [1, 1, 8, 8]);  aten_bmm_default = None
    aten_add_tensor = executorch_exir_dialects_edge__ops_aten_add_Tensor(aten_view_copy_default_2, aten_where_self);  aten_view_copy_default_2 = aten_where_self = None
    aten__softmax_default = executorch_exir_dialects_edge__ops_aten__softmax_default(aten_add_tensor, -1, False);  aten_add_tensor = None
    aten_expand_copy_default_2 = executorch_exir_dialects_edge__ops_aten_expand_copy_default(aten__softmax_default, [1, 1, 8, 8]);  aten__softmax_default = None
    aten_view_copy_default_3 = executorch_exir_dialects_edge__ops_aten_view_copy_default(aten_expand_copy_default_2, [1, 8, 8]);  aten_expand_copy_default_2 = None
    aten_expand_copy_default_3 = executorch_exir_dialects_edge__ops_aten_expand_copy_default(v, [1, 1, 8, 8]);  v = None
    aten_view_copy_default_4 = executorch_exir_dialects_edge__ops_aten_view_copy_default(aten_expand_copy_default_3, [1, 8, 8]);  aten_expand_copy_default_3 = None
    aten_bmm_default_1 = executorch_exir_dialects_edge__ops_aten_bmm_default(aten_view_copy_default_3, aten_view_copy_default_4);  aten_view_copy_default_3 = aten_view_copy_default_4 = None
    aten_view_copy_default_5 = executorch_exir_dialects_edge__ops_aten_view_copy_default(aten_bmm_default_1, [1, 1, 8, 8]);  aten_bmm_default_1 = None
    return (aten_view_copy_default_5,)
```

Differential Revision: [D56119737](https://our.internmc.facebook.com/intern/diff/D56119737/)

[ghstack-poisoned]
@pytorch-bot
Copy link
Copy Markdown

pytorch-bot Bot commented Apr 14, 2024 •
edited
Loading

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/3037

Note: Links to docs will display an error until the docs builds have been completed.

✅ No Failures

As of commit 5465fb7 with merge base 1eed125 (image):
💚 Looks good so far! There are no failures yet. 💚

This comment was automatically generated by Dr. CI and updates every 15 minutes.

@facebook-github-bot facebook-github-bot added the CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. label Apr 14, 2024
@facebook-github-bot
Copy link
Copy Markdown
Contributor

facebook-github-bot commented Apr 14, 2024

This pull request was exported from Phabricator. Differential Revision: D56119737

This was referenced Apr 14, 2024
Closed
Closed
@cccclai
Update on "Add a simple sdpa"
d6c63d1 
Add a simple sdpa so it's decomposed to simpler ops instead of the decompose F.scaled_dot_product_attention, which includes 29 ops including `torch.where`
```
def forward(self, q, k, v):
    aten_mul_scalar = executorch_exir_dialects_edge__ops_aten_mul_Scalar(q, 0.5946035575013605);  q = None
    aten_full_default = executorch_exir_dialects_edge__ops_aten_full_default([8, 8], True, dtype = torch.bool, layout = torch.strided, device = device(type='cpu'), pin_memory = False)
    aten_arange_start_step = executorch_exir_dialects_edge__ops_aten_arange_start_step(0, 8, layout = torch.strided, device = device(type='cpu'), pin_memory = False)
    aten_unsqueeze_copy_default = executorch_exir_dialects_edge__ops_aten_unsqueeze_copy_default(aten_arange_start_step, -2);  aten_arange_start_step = None
    aten_arange_start_step_1 = executorch_exir_dialects_edge__ops_aten_arange_start_step(0, 8, layout = torch.strided, device = device(type='cpu'), pin_memory = False)
    aten_unsqueeze_copy_default_1 = executorch_exir_dialects_edge__ops_aten_unsqueeze_copy_default(aten_arange_start_step_1, -1);  aten_arange_start_step_1 = None
    aten_sub_tensor = executorch_exir_dialects_edge__ops_aten_sub_Tensor(aten_unsqueeze_copy_default, aten_unsqueeze_copy_default_1);  aten_unsqueeze_copy_default = aten_unsqueeze_copy_default_1 = None
    aten_le_scalar = executorch_exir_dialects_edge__ops_aten_le_Scalar(aten_sub_tensor, 0);  aten_sub_tensor = None
    aten_logical_and_default = executorch_exir_dialects_edge__ops_aten_logical_and_default(aten_le_scalar, aten_full_default);  aten_le_scalar = aten_full_default = None
    aten_full_like_default = executorch_exir_dialects_edge__ops_aten_full_like_default(aten_logical_and_default, 0, dtype = torch.float32, pin_memory = False, memory_format = torch.preserve_format)
    aten_logical_not_default = executorch_exir_dialects_edge__ops_aten_logical_not_default(aten_logical_and_default);  aten_logical_and_default = None
    aten_scalar_tensor_default = executorch_exir_dialects_edge__ops_aten_scalar_tensor_default(-inf, dtype = torch.float32, layout = torch.strided, device = device(type='cpu'))
    aten_where_self = executorch_exir_dialects_edge__ops_aten_where_self(aten_logical_not_default, aten_scalar_tensor_default, aten_full_like_default);  aten_logical_not_default = aten_scalar_tensor_default = aten_full_like_default = None
    aten_permute_copy_default = executorch_exir_dialects_edge__ops_aten_permute_copy_default(k, [0, 1, 3, 2]);  k = None
    aten_mul_scalar_1 = executorch_exir_dialects_edge__ops_aten_mul_Scalar(aten_permute_copy_default, 0.5946035575013605);  aten_permute_copy_default = None
    aten_expand_copy_default = executorch_exir_dialects_edge__ops_aten_expand_copy_default(aten_mul_scalar, [1, 1, 8, 8]);  aten_mul_scalar = None
    aten_view_copy_default = executorch_exir_dialects_edge__ops_aten_view_copy_default(aten_expand_copy_default, [1, 8, 8]);  aten_expand_copy_default = None
    aten_expand_copy_default_1 = executorch_exir_dialects_edge__ops_aten_expand_copy_default(aten_mul_scalar_1, [1, 1, 8, 8]);  aten_mul_scalar_1 = None
    aten_view_copy_default_1 = executorch_exir_dialects_edge__ops_aten_view_copy_default(aten_expand_copy_default_1, [1, 8, 8]);  aten_expand_copy_default_1 = None
    aten_bmm_default = executorch_exir_dialects_edge__ops_aten_bmm_default(aten_view_copy_default, aten_view_copy_default_1);  aten_view_copy_default = aten_view_copy_default_1 = None
    aten_view_copy_default_2 = executorch_exir_dialects_edge__ops_aten_view_copy_default(aten_bmm_default, [1, 1, 8, 8]);  aten_bmm_default = None
    aten_add_tensor = executorch_exir_dialects_edge__ops_aten_add_Tensor(aten_view_copy_default_2, aten_where_self);  aten_view_copy_default_2 = aten_where_self = None
    aten__softmax_default = executorch_exir_dialects_edge__ops_aten__softmax_default(aten_add_tensor, -1, False);  aten_add_tensor = None
    aten_expand_copy_default_2 = executorch_exir_dialects_edge__ops_aten_expand_copy_default(aten__softmax_default, [1, 1, 8, 8]);  aten__softmax_default = None
    aten_view_copy_default_3 = executorch_exir_dialects_edge__ops_aten_view_copy_default(aten_expand_copy_default_2, [1, 8, 8]);  aten_expand_copy_default_2 = None
    aten_expand_copy_default_3 = executorch_exir_dialects_edge__ops_aten_expand_copy_default(v, [1, 1, 8, 8]);  v = None
    aten_view_copy_default_4 = executorch_exir_dialects_edge__ops_aten_view_copy_default(aten_expand_copy_default_3, [1, 8, 8]);  aten_expand_copy_default_3 = None
    aten_bmm_default_1 = executorch_exir_dialects_edge__ops_aten_bmm_default(aten_view_copy_default_3, aten_view_copy_default_4);  aten_view_copy_default_3 = aten_view_copy_default_4 = None
    aten_view_copy_default_5 = executorch_exir_dialects_edge__ops_aten_view_copy_default(aten_bmm_default_1, [1, 1, 8, 8]);  aten_bmm_default_1 = None
    return (aten_view_copy_default_5,)
```

Differential Revision: [D56119737](https://our.internmc.facebook.com/intern/diff/D56119737/)

[ghstack-poisoned]
@facebook-github-bot
Copy link
Copy Markdown
Contributor

facebook-github-bot commented Apr 14, 2024

This pull request was exported from Phabricator. Differential Revision: D56119737

@cccclai
Update on "Add a simple sdpa"
4bcbdce 
Add a simple sdpa so it's decomposed to simpler ops instead of the decompose F.scaled_dot_product_attention, which includes 29 ops including `torch.where`
```
def forward(self, q, k, v):
    aten_mul_scalar = executorch_exir_dialects_edge__ops_aten_mul_Scalar(q, 0.5946035575013605);  q = None
    aten_full_default = executorch_exir_dialects_edge__ops_aten_full_default([8, 8], True, dtype = torch.bool, layout = torch.strided, device = device(type='cpu'), pin_memory = False)
    aten_arange_start_step = executorch_exir_dialects_edge__ops_aten_arange_start_step(0, 8, layout = torch.strided, device = device(type='cpu'), pin_memory = False)
    aten_unsqueeze_copy_default = executorch_exir_dialects_edge__ops_aten_unsqueeze_copy_default(aten_arange_start_step, -2);  aten_arange_start_step = None
    aten_arange_start_step_1 = executorch_exir_dialects_edge__ops_aten_arange_start_step(0, 8, layout = torch.strided, device = device(type='cpu'), pin_memory = False)
    aten_unsqueeze_copy_default_1 = executorch_exir_dialects_edge__ops_aten_unsqueeze_copy_default(aten_arange_start_step_1, -1);  aten_arange_start_step_1 = None
    aten_sub_tensor = executorch_exir_dialects_edge__ops_aten_sub_Tensor(aten_unsqueeze_copy_default, aten_unsqueeze_copy_default_1);  aten_unsqueeze_copy_default = aten_unsqueeze_copy_default_1 = None
    aten_le_scalar = executorch_exir_dialects_edge__ops_aten_le_Scalar(aten_sub_tensor, 0);  aten_sub_tensor = None
    aten_logical_and_default = executorch_exir_dialects_edge__ops_aten_logical_and_default(aten_le_scalar, aten_full_default);  aten_le_scalar = aten_full_default = None
    aten_full_like_default = executorch_exir_dialects_edge__ops_aten_full_like_default(aten_logical_and_default, 0, dtype = torch.float32, pin_memory = False, memory_format = torch.preserve_format)
    aten_logical_not_default = executorch_exir_dialects_edge__ops_aten_logical_not_default(aten_logical_and_default);  aten_logical_and_default = None
    aten_scalar_tensor_default = executorch_exir_dialects_edge__ops_aten_scalar_tensor_default(-inf, dtype = torch.float32, layout = torch.strided, device = device(type='cpu'))
    aten_where_self = executorch_exir_dialects_edge__ops_aten_where_self(aten_logical_not_default, aten_scalar_tensor_default, aten_full_like_default);  aten_logical_not_default = aten_scalar_tensor_default = aten_full_like_default = None
    aten_permute_copy_default = executorch_exir_dialects_edge__ops_aten_permute_copy_default(k, [0, 1, 3, 2]);  k = None
    aten_mul_scalar_1 = executorch_exir_dialects_edge__ops_aten_mul_Scalar(aten_permute_copy_default, 0.5946035575013605);  aten_permute_copy_default = None
    aten_expand_copy_default = executorch_exir_dialects_edge__ops_aten_expand_copy_default(aten_mul_scalar, [1, 1, 8, 8]);  aten_mul_scalar = None
    aten_view_copy_default = executorch_exir_dialects_edge__ops_aten_view_copy_default(aten_expand_copy_default, [1, 8, 8]);  aten_expand_copy_default = None
    aten_expand_copy_default_1 = executorch_exir_dialects_edge__ops_aten_expand_copy_default(aten_mul_scalar_1, [1, 1, 8, 8]);  aten_mul_scalar_1 = None
    aten_view_copy_default_1 = executorch_exir_dialects_edge__ops_aten_view_copy_default(aten_expand_copy_default_1, [1, 8, 8]);  aten_expand_copy_default_1 = None
    aten_bmm_default = executorch_exir_dialects_edge__ops_aten_bmm_default(aten_view_copy_default, aten_view_copy_default_1);  aten_view_copy_default = aten_view_copy_default_1 = None
    aten_view_copy_default_2 = executorch_exir_dialects_edge__ops_aten_view_copy_default(aten_bmm_default, [1, 1, 8, 8]);  aten_bmm_default = None
    aten_add_tensor = executorch_exir_dialects_edge__ops_aten_add_Tensor(aten_view_copy_default_2, aten_where_self);  aten_view_copy_default_2 = aten_where_self = None
    aten__softmax_default = executorch_exir_dialects_edge__ops_aten__softmax_default(aten_add_tensor, -1, False);  aten_add_tensor = None
    aten_expand_copy_default_2 = executorch_exir_dialects_edge__ops_aten_expand_copy_default(aten__softmax_default, [1, 1, 8, 8]);  aten__softmax_default = None
    aten_view_copy_default_3 = executorch_exir_dialects_edge__ops_aten_view_copy_default(aten_expand_copy_default_2, [1, 8, 8]);  aten_expand_copy_default_2 = None
    aten_expand_copy_default_3 = executorch_exir_dialects_edge__ops_aten_expand_copy_default(v, [1, 1, 8, 8]);  v = None
    aten_view_copy_default_4 = executorch_exir_dialects_edge__ops_aten_view_copy_default(aten_expand_copy_default_3, [1, 8, 8]);  aten_expand_copy_default_3 = None
    aten_bmm_default_1 = executorch_exir_dialects_edge__ops_aten_bmm_default(aten_view_copy_default_3, aten_view_copy_default_4);  aten_view_copy_default_3 = aten_view_copy_default_4 = None
    aten_view_copy_default_5 = executorch_exir_dialects_edge__ops_aten_view_copy_default(aten_bmm_default_1, [1, 1, 8, 8]);  aten_bmm_default_1 = None
    return (aten_view_copy_default_5,)
```

Differential Revision: [D56119737](https://our.internmc.facebook.com/intern/diff/D56119737/)

[ghstack-poisoned]
@facebook-github-bot
Copy link
Copy Markdown
Contributor

facebook-github-bot commented Apr 15, 2024

This pull request was exported from Phabricator. Differential Revision: D56119737

@cccclai
Update on "Add a simple sdpa"
00347ea 
Add a simple sdpa so it's decomposed to simpler ops instead of the decompose F.scaled_dot_product_attention, which includes 29 ops including `torch.where`
```
def forward(self, q, k, v):
    aten_mul_scalar = executorch_exir_dialects_edge__ops_aten_mul_Scalar(q, 0.5946035575013605);  q = None
    aten_full_default = executorch_exir_dialects_edge__ops_aten_full_default([8, 8], True, dtype = torch.bool, layout = torch.strided, device = device(type='cpu'), pin_memory = False)
    aten_arange_start_step = executorch_exir_dialects_edge__ops_aten_arange_start_step(0, 8, layout = torch.strided, device = device(type='cpu'), pin_memory = False)
    aten_unsqueeze_copy_default = executorch_exir_dialects_edge__ops_aten_unsqueeze_copy_default(aten_arange_start_step, -2);  aten_arange_start_step = None
    aten_arange_start_step_1 = executorch_exir_dialects_edge__ops_aten_arange_start_step(0, 8, layout = torch.strided, device = device(type='cpu'), pin_memory = False)
    aten_unsqueeze_copy_default_1 = executorch_exir_dialects_edge__ops_aten_unsqueeze_copy_default(aten_arange_start_step_1, -1);  aten_arange_start_step_1 = None
    aten_sub_tensor = executorch_exir_dialects_edge__ops_aten_sub_Tensor(aten_unsqueeze_copy_default, aten_unsqueeze_copy_default_1);  aten_unsqueeze_copy_default = aten_unsqueeze_copy_default_1 = None
    aten_le_scalar = executorch_exir_dialects_edge__ops_aten_le_Scalar(aten_sub_tensor, 0);  aten_sub_tensor = None
    aten_logical_and_default = executorch_exir_dialects_edge__ops_aten_logical_and_default(aten_le_scalar, aten_full_default);  aten_le_scalar = aten_full_default = None
    aten_full_like_default = executorch_exir_dialects_edge__ops_aten_full_like_default(aten_logical_and_default, 0, dtype = torch.float32, pin_memory = False, memory_format = torch.preserve_format)
    aten_logical_not_default = executorch_exir_dialects_edge__ops_aten_logical_not_default(aten_logical_and_default);  aten_logical_and_default = None
    aten_scalar_tensor_default = executorch_exir_dialects_edge__ops_aten_scalar_tensor_default(-inf, dtype = torch.float32, layout = torch.strided, device = device(type='cpu'))
    aten_where_self = executorch_exir_dialects_edge__ops_aten_where_self(aten_logical_not_default, aten_scalar_tensor_default, aten_full_like_default);  aten_logical_not_default = aten_scalar_tensor_default = aten_full_like_default = None
    aten_permute_copy_default = executorch_exir_dialects_edge__ops_aten_permute_copy_default(k, [0, 1, 3, 2]);  k = None
    aten_mul_scalar_1 = executorch_exir_dialects_edge__ops_aten_mul_Scalar(aten_permute_copy_default, 0.5946035575013605);  aten_permute_copy_default = None
    aten_expand_copy_default = executorch_exir_dialects_edge__ops_aten_expand_copy_default(aten_mul_scalar, [1, 1, 8, 8]);  aten_mul_scalar = None
    aten_view_copy_default = executorch_exir_dialects_edge__ops_aten_view_copy_default(aten_expand_copy_default, [1, 8, 8]);  aten_expand_copy_default = None
    aten_expand_copy_default_1 = executorch_exir_dialects_edge__ops_aten_expand_copy_default(aten_mul_scalar_1, [1, 1, 8, 8]);  aten_mul_scalar_1 = None
    aten_view_copy_default_1 = executorch_exir_dialects_edge__ops_aten_view_copy_default(aten_expand_copy_default_1, [1, 8, 8]);  aten_expand_copy_default_1 = None
    aten_bmm_default = executorch_exir_dialects_edge__ops_aten_bmm_default(aten_view_copy_default, aten_view_copy_default_1);  aten_view_copy_default = aten_view_copy_default_1 = None
    aten_view_copy_default_2 = executorch_exir_dialects_edge__ops_aten_view_copy_default(aten_bmm_default, [1, 1, 8, 8]);  aten_bmm_default = None
    aten_add_tensor = executorch_exir_dialects_edge__ops_aten_add_Tensor(aten_view_copy_default_2, aten_where_self);  aten_view_copy_default_2 = aten_where_self = None
    aten__softmax_default = executorch_exir_dialects_edge__ops_aten__softmax_default(aten_add_tensor, -1, False);  aten_add_tensor = None
    aten_expand_copy_default_2 = executorch_exir_dialects_edge__ops_aten_expand_copy_default(aten__softmax_default, [1, 1, 8, 8]);  aten__softmax_default = None
    aten_view_copy_default_3 = executorch_exir_dialects_edge__ops_aten_view_copy_default(aten_expand_copy_default_2, [1, 8, 8]);  aten_expand_copy_default_2 = None
    aten_expand_copy_default_3 = executorch_exir_dialects_edge__ops_aten_expand_copy_default(v, [1, 1, 8, 8]);  v = None
    aten_view_copy_default_4 = executorch_exir_dialects_edge__ops_aten_view_copy_default(aten_expand_copy_default_3, [1, 8, 8]);  aten_expand_copy_default_3 = None
    aten_bmm_default_1 = executorch_exir_dialects_edge__ops_aten_bmm_default(aten_view_copy_default_3, aten_view_copy_default_4);  aten_view_copy_default_3 = aten_view_copy_default_4 = None
    aten_view_copy_default_5 = executorch_exir_dialects_edge__ops_aten_view_copy_default(aten_bmm_default_1, [1, 1, 8, 8]);  aten_bmm_default_1 = None
    return (aten_view_copy_default_5,)
```

Differential Revision: [D56119737](https://our.internmc.facebook.com/intern/diff/D56119737/)

[ghstack-poisoned]
@facebook-github-bot
Copy link
Copy Markdown
Contributor

facebook-github-bot commented Apr 16, 2024

This pull request was exported from Phabricator. Differential Revision: D56119737

This was referenced Apr 16, 2024
Closed
Closed
kimishpatel
kimishpatel reviewed Apr 17, 2024
View reviewed changes
Comment thread examples/models/llama2/export_llama_lib.py
return module


class SDPASimple(torch.nn.Module):
Copy link
Copy Markdown
Contributor

@kimishpatel kimishpatel Apr 17, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Can we just keep sdpa as is by registering it as a custom qnn op that qnn delegate can directly consume?

Copy link
Copy Markdown
Contributor Author

@cccclai cccclai Apr 17, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

It's not just for qnn. Basically for any backend that doesn't support sdpa, it makes more sense to use this version instead.

kimishpatel
kimishpatel reviewed Apr 17, 2024
View reviewed changes
Comment thread examples/models/llama2/export_llama_lib.py
attn_weight = torch.softmax(attn_weight, dim=-1)
y = attn_weight @ v

return y.transpose(1, 2).contiguous().view(bsz, seqlen, self.dim)
Copy link
Copy Markdown
Contributor

@kimishpatel kimishpatel Apr 17, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Do you need contiguous?

@cccclai
Update on "Add a simple sdpa"
8e99c7a 
Add a simple sdpa so it's decomposed to simpler ops instead of the decompose F.scaled_dot_product_attention, which includes 29 ops including `torch.where`
```
def forward(self, q, k, v):
    aten_mul_scalar = executorch_exir_dialects_edge__ops_aten_mul_Scalar(q, 0.5946035575013605);  q = None
    aten_full_default = executorch_exir_dialects_edge__ops_aten_full_default([8, 8], True, dtype = torch.bool, layout = torch.strided, device = device(type='cpu'), pin_memory = False)
    aten_arange_start_step = executorch_exir_dialects_edge__ops_aten_arange_start_step(0, 8, layout = torch.strided, device = device(type='cpu'), pin_memory = False)
    aten_unsqueeze_copy_default = executorch_exir_dialects_edge__ops_aten_unsqueeze_copy_default(aten_arange_start_step, -2);  aten_arange_start_step = None
    aten_arange_start_step_1 = executorch_exir_dialects_edge__ops_aten_arange_start_step(0, 8, layout = torch.strided, device = device(type='cpu'), pin_memory = False)
    aten_unsqueeze_copy_default_1 = executorch_exir_dialects_edge__ops_aten_unsqueeze_copy_default(aten_arange_start_step_1, -1);  aten_arange_start_step_1 = None
    aten_sub_tensor = executorch_exir_dialects_edge__ops_aten_sub_Tensor(aten_unsqueeze_copy_default, aten_unsqueeze_copy_default_1);  aten_unsqueeze_copy_default = aten_unsqueeze_copy_default_1 = None
    aten_le_scalar = executorch_exir_dialects_edge__ops_aten_le_Scalar(aten_sub_tensor, 0);  aten_sub_tensor = None
    aten_logical_and_default = executorch_exir_dialects_edge__ops_aten_logical_and_default(aten_le_scalar, aten_full_default);  aten_le_scalar = aten_full_default = None
    aten_full_like_default = executorch_exir_dialects_edge__ops_aten_full_like_default(aten_logical_and_default, 0, dtype = torch.float32, pin_memory = False, memory_format = torch.preserve_format)
    aten_logical_not_default = executorch_exir_dialects_edge__ops_aten_logical_not_default(aten_logical_and_default);  aten_logical_and_default = None
    aten_scalar_tensor_default = executorch_exir_dialects_edge__ops_aten_scalar_tensor_default(-inf, dtype = torch.float32, layout = torch.strided, device = device(type='cpu'))
    aten_where_self = executorch_exir_dialects_edge__ops_aten_where_self(aten_logical_not_default, aten_scalar_tensor_default, aten_full_like_default);  aten_logical_not_default = aten_scalar_tensor_default = aten_full_like_default = None
    aten_permute_copy_default = executorch_exir_dialects_edge__ops_aten_permute_copy_default(k, [0, 1, 3, 2]);  k = None
    aten_mul_scalar_1 = executorch_exir_dialects_edge__ops_aten_mul_Scalar(aten_permute_copy_default, 0.5946035575013605);  aten_permute_copy_default = None
    aten_expand_copy_default = executorch_exir_dialects_edge__ops_aten_expand_copy_default(aten_mul_scalar, [1, 1, 8, 8]);  aten_mul_scalar = None
    aten_view_copy_default = executorch_exir_dialects_edge__ops_aten_view_copy_default(aten_expand_copy_default, [1, 8, 8]);  aten_expand_copy_default = None
    aten_expand_copy_default_1 = executorch_exir_dialects_edge__ops_aten_expand_copy_default(aten_mul_scalar_1, [1, 1, 8, 8]);  aten_mul_scalar_1 = None
    aten_view_copy_default_1 = executorch_exir_dialects_edge__ops_aten_view_copy_default(aten_expand_copy_default_1, [1, 8, 8]);  aten_expand_copy_default_1 = None
    aten_bmm_default = executorch_exir_dialects_edge__ops_aten_bmm_default(aten_view_copy_default, aten_view_copy_default_1);  aten_view_copy_default = aten_view_copy_default_1 = None
    aten_view_copy_default_2 = executorch_exir_dialects_edge__ops_aten_view_copy_default(aten_bmm_default, [1, 1, 8, 8]);  aten_bmm_default = None
    aten_add_tensor = executorch_exir_dialects_edge__ops_aten_add_Tensor(aten_view_copy_default_2, aten_where_self);  aten_view_copy_default_2 = aten_where_self = None
    aten__softmax_default = executorch_exir_dialects_edge__ops_aten__softmax_default(aten_add_tensor, -1, False);  aten_add_tensor = None
    aten_expand_copy_default_2 = executorch_exir_dialects_edge__ops_aten_expand_copy_default(aten__softmax_default, [1, 1, 8, 8]);  aten__softmax_default = None
    aten_view_copy_default_3 = executorch_exir_dialects_edge__ops_aten_view_copy_default(aten_expand_copy_default_2, [1, 8, 8]);  aten_expand_copy_default_2 = None
    aten_expand_copy_default_3 = executorch_exir_dialects_edge__ops_aten_expand_copy_default(v, [1, 1, 8, 8]);  v = None
    aten_view_copy_default_4 = executorch_exir_dialects_edge__ops_aten_view_copy_default(aten_expand_copy_default_3, [1, 8, 8]);  aten_expand_copy_default_3 = None
    aten_bmm_default_1 = executorch_exir_dialects_edge__ops_aten_bmm_default(aten_view_copy_default_3, aten_view_copy_default_4);  aten_view_copy_default_3 = aten_view_copy_default_4 = None
    aten_view_copy_default_5 = executorch_exir_dialects_edge__ops_aten_view_copy_default(aten_bmm_default_1, [1, 1, 8, 8]);  aten_bmm_default_1 = None
    return (aten_view_copy_default_5,)
```

Differential Revision: [D56119737](https://our.internmc.facebook.com/intern/diff/D56119737/)

[ghstack-poisoned]
@facebook-github-bot
Copy link
Copy Markdown
Contributor

facebook-github-bot commented Apr 18, 2024

This pull request was exported from Phabricator. Differential Revision: D56119737

@cccclai
Update on "Add a simple sdpa"
f5ec6cf 
Add a simple sdpa so it's decomposed to simpler ops instead of the decompose F.scaled_dot_product_attention, which includes 29 ops including `torch.where`
```
def forward(self, q, k, v):
    aten_mul_scalar = executorch_exir_dialects_edge__ops_aten_mul_Scalar(q, 0.5946035575013605);  q = None
    aten_full_default = executorch_exir_dialects_edge__ops_aten_full_default([8, 8], True, dtype = torch.bool, layout = torch.strided, device = device(type='cpu'), pin_memory = False)
    aten_arange_start_step = executorch_exir_dialects_edge__ops_aten_arange_start_step(0, 8, layout = torch.strided, device = device(type='cpu'), pin_memory = False)
    aten_unsqueeze_copy_default = executorch_exir_dialects_edge__ops_aten_unsqueeze_copy_default(aten_arange_start_step, -2);  aten_arange_start_step = None
    aten_arange_start_step_1 = executorch_exir_dialects_edge__ops_aten_arange_start_step(0, 8, layout = torch.strided, device = device(type='cpu'), pin_memory = False)
    aten_unsqueeze_copy_default_1 = executorch_exir_dialects_edge__ops_aten_unsqueeze_copy_default(aten_arange_start_step_1, -1);  aten_arange_start_step_1 = None
    aten_sub_tensor = executorch_exir_dialects_edge__ops_aten_sub_Tensor(aten_unsqueeze_copy_default, aten_unsqueeze_copy_default_1);  aten_unsqueeze_copy_default = aten_unsqueeze_copy_default_1 = None
    aten_le_scalar = executorch_exir_dialects_edge__ops_aten_le_Scalar(aten_sub_tensor, 0);  aten_sub_tensor = None
    aten_logical_and_default = executorch_exir_dialects_edge__ops_aten_logical_and_default(aten_le_scalar, aten_full_default);  aten_le_scalar = aten_full_default = None
    aten_full_like_default = executorch_exir_dialects_edge__ops_aten_full_like_default(aten_logical_and_default, 0, dtype = torch.float32, pin_memory = False, memory_format = torch.preserve_format)
    aten_logical_not_default = executorch_exir_dialects_edge__ops_aten_logical_not_default(aten_logical_and_default);  aten_logical_and_default = None
    aten_scalar_tensor_default = executorch_exir_dialects_edge__ops_aten_scalar_tensor_default(-inf, dtype = torch.float32, layout = torch.strided, device = device(type='cpu'))
    aten_where_self = executorch_exir_dialects_edge__ops_aten_where_self(aten_logical_not_default, aten_scalar_tensor_default, aten_full_like_default);  aten_logical_not_default = aten_scalar_tensor_default = aten_full_like_default = None
    aten_permute_copy_default = executorch_exir_dialects_edge__ops_aten_permute_copy_default(k, [0, 1, 3, 2]);  k = None
    aten_mul_scalar_1 = executorch_exir_dialects_edge__ops_aten_mul_Scalar(aten_permute_copy_default, 0.5946035575013605);  aten_permute_copy_default = None
    aten_expand_copy_default = executorch_exir_dialects_edge__ops_aten_expand_copy_default(aten_mul_scalar, [1, 1, 8, 8]);  aten_mul_scalar = None
    aten_view_copy_default = executorch_exir_dialects_edge__ops_aten_view_copy_default(aten_expand_copy_default, [1, 8, 8]);  aten_expand_copy_default = None
    aten_expand_copy_default_1 = executorch_exir_dialects_edge__ops_aten_expand_copy_default(aten_mul_scalar_1, [1, 1, 8, 8]);  aten_mul_scalar_1 = None
    aten_view_copy_default_1 = executorch_exir_dialects_edge__ops_aten_view_copy_default(aten_expand_copy_default_1, [1, 8, 8]);  aten_expand_copy_default_1 = None
    aten_bmm_default = executorch_exir_dialects_edge__ops_aten_bmm_default(aten_view_copy_default, aten_view_copy_default_1);  aten_view_copy_default = aten_view_copy_default_1 = None
    aten_view_copy_default_2 = executorch_exir_dialects_edge__ops_aten_view_copy_default(aten_bmm_default, [1, 1, 8, 8]);  aten_bmm_default = None
    aten_add_tensor = executorch_exir_dialects_edge__ops_aten_add_Tensor(aten_view_copy_default_2, aten_where_self);  aten_view_copy_default_2 = aten_where_self = None
    aten__softmax_default = executorch_exir_dialects_edge__ops_aten__softmax_default(aten_add_tensor, -1, False);  aten_add_tensor = None
    aten_expand_copy_default_2 = executorch_exir_dialects_edge__ops_aten_expand_copy_default(aten__softmax_default, [1, 1, 8, 8]);  aten__softmax_default = None
    aten_view_copy_default_3 = executorch_exir_dialects_edge__ops_aten_view_copy_default(aten_expand_copy_default_2, [1, 8, 8]);  aten_expand_copy_default_2 = None
    aten_expand_copy_default_3 = executorch_exir_dialects_edge__ops_aten_expand_copy_default(v, [1, 1, 8, 8]);  v = None
    aten_view_copy_default_4 = executorch_exir_dialects_edge__ops_aten_view_copy_default(aten_expand_copy_default_3, [1, 8, 8]);  aten_expand_copy_default_3 = None
    aten_bmm_default_1 = executorch_exir_dialects_edge__ops_aten_bmm_default(aten_view_copy_default_3, aten_view_copy_default_4);  aten_view_copy_default_3 = aten_view_copy_default_4 = None
    aten_view_copy_default_5 = executorch_exir_dialects_edge__ops_aten_view_copy_default(aten_bmm_default_1, [1, 1, 8, 8]);  aten_bmm_default_1 = None
    return (aten_view_copy_default_5,)
```

Differential Revision: [D56119737](https://our.internmc.facebook.com/intern/diff/D56119737/)

[ghstack-poisoned]
@facebook-github-bot
Copy link
Copy Markdown
Contributor

facebook-github-bot commented Apr 19, 2024

This pull request was exported from Phabricator. Differential Revision: D56119737

@cccclai
Update on "Add a simple sdpa"
5465fb7 
Add a simple sdpa so it's decomposed to simpler ops instead of the decompose F.scaled_dot_product_attention, which includes 29 ops including `torch.where`
```
def forward(self, q, k, v):
    aten_mul_scalar = executorch_exir_dialects_edge__ops_aten_mul_Scalar(q, 0.5946035575013605);  q = None
    aten_full_default = executorch_exir_dialects_edge__ops_aten_full_default([8, 8], True, dtype = torch.bool, layout = torch.strided, device = device(type='cpu'), pin_memory = False)
    aten_arange_start_step = executorch_exir_dialects_edge__ops_aten_arange_start_step(0, 8, layout = torch.strided, device = device(type='cpu'), pin_memory = False)
    aten_unsqueeze_copy_default = executorch_exir_dialects_edge__ops_aten_unsqueeze_copy_default(aten_arange_start_step, -2);  aten_arange_start_step = None
    aten_arange_start_step_1 = executorch_exir_dialects_edge__ops_aten_arange_start_step(0, 8, layout = torch.strided, device = device(type='cpu'), pin_memory = False)
    aten_unsqueeze_copy_default_1 = executorch_exir_dialects_edge__ops_aten_unsqueeze_copy_default(aten_arange_start_step_1, -1);  aten_arange_start_step_1 = None
    aten_sub_tensor = executorch_exir_dialects_edge__ops_aten_sub_Tensor(aten_unsqueeze_copy_default, aten_unsqueeze_copy_default_1);  aten_unsqueeze_copy_default = aten_unsqueeze_copy_default_1 = None
    aten_le_scalar = executorch_exir_dialects_edge__ops_aten_le_Scalar(aten_sub_tensor, 0);  aten_sub_tensor = None
    aten_logical_and_default = executorch_exir_dialects_edge__ops_aten_logical_and_default(aten_le_scalar, aten_full_default);  aten_le_scalar = aten_full_default = None
    aten_full_like_default = executorch_exir_dialects_edge__ops_aten_full_like_default(aten_logical_and_default, 0, dtype = torch.float32, pin_memory = False, memory_format = torch.preserve_format)
    aten_logical_not_default = executorch_exir_dialects_edge__ops_aten_logical_not_default(aten_logical_and_default);  aten_logical_and_default = None
    aten_scalar_tensor_default = executorch_exir_dialects_edge__ops_aten_scalar_tensor_default(-inf, dtype = torch.float32, layout = torch.strided, device = device(type='cpu'))
    aten_where_self = executorch_exir_dialects_edge__ops_aten_where_self(aten_logical_not_default, aten_scalar_tensor_default, aten_full_like_default);  aten_logical_not_default = aten_scalar_tensor_default = aten_full_like_default = None
    aten_permute_copy_default = executorch_exir_dialects_edge__ops_aten_permute_copy_default(k, [0, 1, 3, 2]);  k = None
    aten_mul_scalar_1 = executorch_exir_dialects_edge__ops_aten_mul_Scalar(aten_permute_copy_default, 0.5946035575013605);  aten_permute_copy_default = None
    aten_expand_copy_default = executorch_exir_dialects_edge__ops_aten_expand_copy_default(aten_mul_scalar, [1, 1, 8, 8]);  aten_mul_scalar = None
    aten_view_copy_default = executorch_exir_dialects_edge__ops_aten_view_copy_default(aten_expand_copy_default, [1, 8, 8]);  aten_expand_copy_default = None
    aten_expand_copy_default_1 = executorch_exir_dialects_edge__ops_aten_expand_copy_default(aten_mul_scalar_1, [1, 1, 8, 8]);  aten_mul_scalar_1 = None
    aten_view_copy_default_1 = executorch_exir_dialects_edge__ops_aten_view_copy_default(aten_expand_copy_default_1, [1, 8, 8]);  aten_expand_copy_default_1 = None
    aten_bmm_default = executorch_exir_dialects_edge__ops_aten_bmm_default(aten_view_copy_default, aten_view_copy_default_1);  aten_view_copy_default = aten_view_copy_default_1 = None
    aten_view_copy_default_2 = executorch_exir_dialects_edge__ops_aten_view_copy_default(aten_bmm_default, [1, 1, 8, 8]);  aten_bmm_default = None
    aten_add_tensor = executorch_exir_dialects_edge__ops_aten_add_Tensor(aten_view_copy_default_2, aten_where_self);  aten_view_copy_default_2 = aten_where_self = None
    aten__softmax_default = executorch_exir_dialects_edge__ops_aten__softmax_default(aten_add_tensor, -1, False);  aten_add_tensor = None
    aten_expand_copy_default_2 = executorch_exir_dialects_edge__ops_aten_expand_copy_default(aten__softmax_default, [1, 1, 8, 8]);  aten__softmax_default = None
    aten_view_copy_default_3 = executorch_exir_dialects_edge__ops_aten_view_copy_default(aten_expand_copy_default_2, [1, 8, 8]);  aten_expand_copy_default_2 = None
    aten_expand_copy_default_3 = executorch_exir_dialects_edge__ops_aten_expand_copy_default(v, [1, 1, 8, 8]);  v = None
    aten_view_copy_default_4 = executorch_exir_dialects_edge__ops_aten_view_copy_default(aten_expand_copy_default_3, [1, 8, 8]);  aten_expand_copy_default_3 = None
    aten_bmm_default_1 = executorch_exir_dialects_edge__ops_aten_bmm_default(aten_view_copy_default_3, aten_view_copy_default_4);  aten_view_copy_default_3 = aten_view_copy_default_4 = None
    aten_view_copy_default_5 = executorch_exir_dialects_edge__ops_aten_view_copy_default(aten_bmm_default_1, [1, 1, 8, 8]);  aten_bmm_default_1 = None
    return (aten_view_copy_default_5,)
```

Differential Revision: [D56119737](https://our.internmc.facebook.com/intern/diff/D56119737/)

[ghstack-poisoned]
@facebook-github-bot
Copy link
Copy Markdown
Contributor

facebook-github-bot commented Apr 19, 2024

This pull request was exported from Phabricator. Differential Revision: D56119737

@facebook-github-bot
Copy link
Copy Markdown
Contributor

facebook-github-bot commented Apr 19, 2024

This pull request has been merged in cf78107.

cccclai added a commit to cccclai/executorch-1 that referenced this pull request Apr 19, 2024
@cccclai
Add a simple sdpa (pytorch#3037)
dc0b5bd 
Summary:
Pull Request resolved: pytorch#3037

Add a simple sdpa so it's decomposed to simpler ops instead of the decompose F.scaled_dot_product_attention, which includes 29 ops including `torch.where`
```
def forward(self, q, k, v):
    aten_mul_scalar = executorch_exir_dialects_edge__ops_aten_mul_Scalar(q, 0.5946035575013605);  q = None
    aten_full_default = executorch_exir_dialects_edge__ops_aten_full_default([8, 8], True, dtype = torch.bool, layout = torch.strided, device = device(type='cpu'), pin_memory = False)
    aten_arange_start_step = executorch_exir_dialects_edge__ops_aten_arange_start_step(0, 8, layout = torch.strided, device = device(type='cpu'), pin_memory = False)
    aten_unsqueeze_copy_default = executorch_exir_dialects_edge__ops_aten_unsqueeze_copy_default(aten_arange_start_step, -2);  aten_arange_start_step = None
    aten_arange_start_step_1 = executorch_exir_dialects_edge__ops_aten_arange_start_step(0, 8, layout = torch.strided, device = device(type='cpu'), pin_memory = False)
    aten_unsqueeze_copy_default_1 = executorch_exir_dialects_edge__ops_aten_unsqueeze_copy_default(aten_arange_start_step_1, -1);  aten_arange_start_step_1 = None
    aten_sub_tensor = executorch_exir_dialects_edge__ops_aten_sub_Tensor(aten_unsqueeze_copy_default, aten_unsqueeze_copy_default_1);  aten_unsqueeze_copy_default = aten_unsqueeze_copy_default_1 = None
    aten_le_scalar = executorch_exir_dialects_edge__ops_aten_le_Scalar(aten_sub_tensor, 0);  aten_sub_tensor = None
    aten_logical_and_default = executorch_exir_dialects_edge__ops_aten_logical_and_default(aten_le_scalar, aten_full_default);  aten_le_scalar = aten_full_default = None
    aten_full_like_default = executorch_exir_dialects_edge__ops_aten_full_like_default(aten_logical_and_default, 0, dtype = torch.float32, pin_memory = False, memory_format = torch.preserve_format)
    aten_logical_not_default = executorch_exir_dialects_edge__ops_aten_logical_not_default(aten_logical_and_default);  aten_logical_and_default = None
    aten_scalar_tensor_default = executorch_exir_dialects_edge__ops_aten_scalar_tensor_default(-inf, dtype = torch.float32, layout = torch.strided, device = device(type='cpu'))
    aten_where_self = executorch_exir_dialects_edge__ops_aten_where_self(aten_logical_not_default, aten_scalar_tensor_default, aten_full_like_default);  aten_logical_not_default = aten_scalar_tensor_default = aten_full_like_default = None
    aten_permute_copy_default = executorch_exir_dialects_edge__ops_aten_permute_copy_default(k, [0, 1, 3, 2]);  k = None
    aten_mul_scalar_1 = executorch_exir_dialects_edge__ops_aten_mul_Scalar(aten_permute_copy_default, 0.5946035575013605);  aten_permute_copy_default = None
    aten_expand_copy_default = executorch_exir_dialects_edge__ops_aten_expand_copy_default(aten_mul_scalar, [1, 1, 8, 8]);  aten_mul_scalar = None
    aten_view_copy_default = executorch_exir_dialects_edge__ops_aten_view_copy_default(aten_expand_copy_default, [1, 8, 8]);  aten_expand_copy_default = None
    aten_expand_copy_default_1 = executorch_exir_dialects_edge__ops_aten_expand_copy_default(aten_mul_scalar_1, [1, 1, 8, 8]);  aten_mul_scalar_1 = None
    aten_view_copy_default_1 = executorch_exir_dialects_edge__ops_aten_view_copy_default(aten_expand_copy_default_1, [1, 8, 8]);  aten_expand_copy_default_1 = None
    aten_bmm_default = executorch_exir_dialects_edge__ops_aten_bmm_default(aten_view_copy_default, aten_view_copy_default_1);  aten_view_copy_default = aten_view_copy_default_1 = None
    aten_view_copy_default_2 = executorch_exir_dialects_edge__ops_aten_view_copy_default(aten_bmm_default, [1, 1, 8, 8]);  aten_bmm_default = None
    aten_add_tensor = executorch_exir_dialects_edge__ops_aten_add_Tensor(aten_view_copy_default_2, aten_where_self);  aten_view_copy_default_2 = aten_where_self = None
    aten__softmax_default = executorch_exir_dialects_edge__ops_aten__softmax_default(aten_add_tensor, -1, False);  aten_add_tensor = None
    aten_expand_copy_default_2 = executorch_exir_dialects_edge__ops_aten_expand_copy_default(aten__softmax_default, [1, 1, 8, 8]);  aten__softmax_default = None
    aten_view_copy_default_3 = executorch_exir_dialects_edge__ops_aten_view_copy_default(aten_expand_copy_default_2, [1, 8, 8]);  aten_expand_copy_default_2 = None
    aten_expand_copy_default_3 = executorch_exir_dialects_edge__ops_aten_expand_copy_default(v, [1, 1, 8, 8]);  v = None
    aten_view_copy_default_4 = executorch_exir_dialects_edge__ops_aten_view_copy_default(aten_expand_copy_default_3, [1, 8, 8]);  aten_expand_copy_default_3 = None
    aten_bmm_default_1 = executorch_exir_dialects_edge__ops_aten_bmm_default(aten_view_copy_default_3, aten_view_copy_default_4);  aten_view_copy_default_3 = aten_view_copy_default_4 = None
    aten_view_copy_default_5 = executorch_exir_dialects_edge__ops_aten_view_copy_default(aten_bmm_default_1, [1, 1, 8, 8]);  aten_bmm_default_1 = None
    return (aten_view_copy_default_5,)
```
After applying the diff, we remove the following ops
```
    %aten_full_like_default : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.full_like.default](args = (%aten_index_tensor_2, 0), kwargs = {dtype: torch.float32, pin_memory: False, memory_format: torch.preserve_format})

    %aten_logical_not_default : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.logical_not.default](args = (%aten_index_tensor_2,), kwargs = {})

    %aten_scalar_tensor_default : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.scalar_tensor.default](args = (-inf,), kwargs = {dtype: torch.float32, layout: torch.strided, device: cpu})

    %aten_where_self : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.where.self](args = (%aten_logical_not_default, %aten_scalar_tensor_default, %aten_full_like_default), kwargs = {})

    %aten_mul_scalar : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.mul.Scalar](args = (%aten_permute_copy_default_3, 0.5946035575013605), kwargs = {})
    ...
    %aten_mul_scalar_1 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.mul.Scalar](args = (%aten_permute_copy_default_6, 0.5946035575013605), kwargs = {})
```
but introduce an add
    %aten_add_tensor_3 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.add.Tensor](args = (%aten_mul_tensor_11, %aten_index_tensor_2), kwargs = {})
```
ghstack-source-id: 223152096
exported-using-ghexport

Reviewed By: mergennachin, kimishpatel

Differential Revision: D56119737

fbshipit-source-id: ec8e875f0a4c4ec67b7493e4872c9a5b081e6de7
(cherry picked from commit cf78107)
@cccclai cccclai mentioned this pull request Apr 19, 2024
Merged
guangy10 pushed a commit that referenced this pull request Apr 19, 2024
@cccclai
Add a simple sdpa (#3037) (#3166)
efb7cf3 
Summary:
Pull Request resolved: #3037

Add a simple sdpa so it's decomposed to simpler ops instead of the decompose F.scaled_dot_product_attention, which includes 29 ops including `torch.where`
```
def forward(self, q, k, v):
    aten_mul_scalar = executorch_exir_dialects_edge__ops_aten_mul_Scalar(q, 0.5946035575013605);  q = None
    aten_full_default = executorch_exir_dialects_edge__ops_aten_full_default([8, 8], True, dtype = torch.bool, layout = torch.strided, device = device(type='cpu'), pin_memory = False)
    aten_arange_start_step = executorch_exir_dialects_edge__ops_aten_arange_start_step(0, 8, layout = torch.strided, device = device(type='cpu'), pin_memory = False)
    aten_unsqueeze_copy_default = executorch_exir_dialects_edge__ops_aten_unsqueeze_copy_default(aten_arange_start_step, -2);  aten_arange_start_step = None
    aten_arange_start_step_1 = executorch_exir_dialects_edge__ops_aten_arange_start_step(0, 8, layout = torch.strided, device = device(type='cpu'), pin_memory = False)
    aten_unsqueeze_copy_default_1 = executorch_exir_dialects_edge__ops_aten_unsqueeze_copy_default(aten_arange_start_step_1, -1);  aten_arange_start_step_1 = None
    aten_sub_tensor = executorch_exir_dialects_edge__ops_aten_sub_Tensor(aten_unsqueeze_copy_default, aten_unsqueeze_copy_default_1);  aten_unsqueeze_copy_default = aten_unsqueeze_copy_default_1 = None
    aten_le_scalar = executorch_exir_dialects_edge__ops_aten_le_Scalar(aten_sub_tensor, 0);  aten_sub_tensor = None
    aten_logical_and_default = executorch_exir_dialects_edge__ops_aten_logical_and_default(aten_le_scalar, aten_full_default);  aten_le_scalar = aten_full_default = None
    aten_full_like_default = executorch_exir_dialects_edge__ops_aten_full_like_default(aten_logical_and_default, 0, dtype = torch.float32, pin_memory = False, memory_format = torch.preserve_format)
    aten_logical_not_default = executorch_exir_dialects_edge__ops_aten_logical_not_default(aten_logical_and_default);  aten_logical_and_default = None
    aten_scalar_tensor_default = executorch_exir_dialects_edge__ops_aten_scalar_tensor_default(-inf, dtype = torch.float32, layout = torch.strided, device = device(type='cpu'))
    aten_where_self = executorch_exir_dialects_edge__ops_aten_where_self(aten_logical_not_default, aten_scalar_tensor_default, aten_full_like_default);  aten_logical_not_default = aten_scalar_tensor_default = aten_full_like_default = None
    aten_permute_copy_default = executorch_exir_dialects_edge__ops_aten_permute_copy_default(k, [0, 1, 3, 2]);  k = None
    aten_mul_scalar_1 = executorch_exir_dialects_edge__ops_aten_mul_Scalar(aten_permute_copy_default, 0.5946035575013605);  aten_permute_copy_default = None
    aten_expand_copy_default = executorch_exir_dialects_edge__ops_aten_expand_copy_default(aten_mul_scalar, [1, 1, 8, 8]);  aten_mul_scalar = None
    aten_view_copy_default = executorch_exir_dialects_edge__ops_aten_view_copy_default(aten_expand_copy_default, [1, 8, 8]);  aten_expand_copy_default = None
    aten_expand_copy_default_1 = executorch_exir_dialects_edge__ops_aten_expand_copy_default(aten_mul_scalar_1, [1, 1, 8, 8]);  aten_mul_scalar_1 = None
    aten_view_copy_default_1 = executorch_exir_dialects_edge__ops_aten_view_copy_default(aten_expand_copy_default_1, [1, 8, 8]);  aten_expand_copy_default_1 = None
    aten_bmm_default = executorch_exir_dialects_edge__ops_aten_bmm_default(aten_view_copy_default, aten_view_copy_default_1);  aten_view_copy_default = aten_view_copy_default_1 = None
    aten_view_copy_default_2 = executorch_exir_dialects_edge__ops_aten_view_copy_default(aten_bmm_default, [1, 1, 8, 8]);  aten_bmm_default = None
    aten_add_tensor = executorch_exir_dialects_edge__ops_aten_add_Tensor(aten_view_copy_default_2, aten_where_self);  aten_view_copy_default_2 = aten_where_self = None
    aten__softmax_default = executorch_exir_dialects_edge__ops_aten__softmax_default(aten_add_tensor, -1, False);  aten_add_tensor = None
    aten_expand_copy_default_2 = executorch_exir_dialects_edge__ops_aten_expand_copy_default(aten__softmax_default, [1, 1, 8, 8]);  aten__softmax_default = None
    aten_view_copy_default_3 = executorch_exir_dialects_edge__ops_aten_view_copy_default(aten_expand_copy_default_2, [1, 8, 8]);  aten_expand_copy_default_2 = None
    aten_expand_copy_default_3 = executorch_exir_dialects_edge__ops_aten_expand_copy_default(v, [1, 1, 8, 8]);  v = None
    aten_view_copy_default_4 = executorch_exir_dialects_edge__ops_aten_view_copy_default(aten_expand_copy_default_3, [1, 8, 8]);  aten_expand_copy_default_3 = None
    aten_bmm_default_1 = executorch_exir_dialects_edge__ops_aten_bmm_default(aten_view_copy_default_3, aten_view_copy_default_4);  aten_view_copy_default_3 = aten_view_copy_default_4 = None
    aten_view_copy_default_5 = executorch_exir_dialects_edge__ops_aten_view_copy_default(aten_bmm_default_1, [1, 1, 8, 8]);  aten_bmm_default_1 = None
    return (aten_view_copy_default_5,)
```
After applying the diff, we remove the following ops
```
    %aten_full_like_default : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.full_like.default](args = (%aten_index_tensor_2, 0), kwargs = {dtype: torch.float32, pin_memory: False, memory_format: torch.preserve_format})

    %aten_logical_not_default : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.logical_not.default](args = (%aten_index_tensor_2,), kwargs = {})

    %aten_scalar_tensor_default : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.scalar_tensor.default](args = (-inf,), kwargs = {dtype: torch.float32, layout: torch.strided, device: cpu})

    %aten_where_self : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.where.self](args = (%aten_logical_not_default, %aten_scalar_tensor_default, %aten_full_like_default), kwargs = {})

    %aten_mul_scalar : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.mul.Scalar](args = (%aten_permute_copy_default_3, 0.5946035575013605), kwargs = {})
    ...
    %aten_mul_scalar_1 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.mul.Scalar](args = (%aten_permute_copy_default_6, 0.5946035575013605), kwargs = {})
```
but introduce an add
    %aten_add_tensor_3 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.add.Tensor](args = (%aten_mul_tensor_11, %aten_index_tensor_2), kwargs = {})
```
ghstack-source-id: 223152096
exported-using-ghexport

Reviewed By: mergennachin, kimishpatel

Differential Revision: D56119737

fbshipit-source-id: ec8e875f0a4c4ec67b7493e4872c9a5b081e6de7
(cherry picked from commit cf78107)
@cccclai cccclai mentioned this pull request Apr 23, 2024
Closed
This was referenced Apr 25, 2024
Closed
Closed
kedarnath03 pushed a commit to kedarnath03/executorch that referenced this pull request Jun 25, 2025
@cccclai
Add a simple sdpa
0b6fb4c 
Pull Request resolved: pytorch/executorch#3037

Add a simple sdpa so it's decomposed to simpler ops instead of the decompose F.scaled_dot_product_attention, which includes 29 ops including `torch.where`
```
def forward(self, q, k, v):
    aten_mul_scalar = executorch_exir_dialects_edge__ops_aten_mul_Scalar(q, 0.5946035575013605);  q = None
    aten_full_default = executorch_exir_dialects_edge__ops_aten_full_default([8, 8], True, dtype = torch.bool, layout = torch.strided, device = device(type='cpu'), pin_memory = False)
    aten_arange_start_step = executorch_exir_dialects_edge__ops_aten_arange_start_step(0, 8, layout = torch.strided, device = device(type='cpu'), pin_memory = False)
    aten_unsqueeze_copy_default = executorch_exir_dialects_edge__ops_aten_unsqueeze_copy_default(aten_arange_start_step, -2);  aten_arange_start_step = None
    aten_arange_start_step_1 = executorch_exir_dialects_edge__ops_aten_arange_start_step(0, 8, layout = torch.strided, device = device(type='cpu'), pin_memory = False)
    aten_unsqueeze_copy_default_1 = executorch_exir_dialects_edge__ops_aten_unsqueeze_copy_default(aten_arange_start_step_1, -1);  aten_arange_start_step_1 = None
    aten_sub_tensor = executorch_exir_dialects_edge__ops_aten_sub_Tensor(aten_unsqueeze_copy_default, aten_unsqueeze_copy_default_1);  aten_unsqueeze_copy_default = aten_unsqueeze_copy_default_1 = None
    aten_le_scalar = executorch_exir_dialects_edge__ops_aten_le_Scalar(aten_sub_tensor, 0);  aten_sub_tensor = None
    aten_logical_and_default = executorch_exir_dialects_edge__ops_aten_logical_and_default(aten_le_scalar, aten_full_default);  aten_le_scalar = aten_full_default = None
    aten_full_like_default = executorch_exir_dialects_edge__ops_aten_full_like_default(aten_logical_and_default, 0, dtype = torch.float32, pin_memory = False, memory_format = torch.preserve_format)
    aten_logical_not_default = executorch_exir_dialects_edge__ops_aten_logical_not_default(aten_logical_and_default);  aten_logical_and_default = None
    aten_scalar_tensor_default = executorch_exir_dialects_edge__ops_aten_scalar_tensor_default(-inf, dtype = torch.float32, layout = torch.strided, device = device(type='cpu'))
    aten_where_self = executorch_exir_dialects_edge__ops_aten_where_self(aten_logical_not_default, aten_scalar_tensor_default, aten_full_like_default);  aten_logical_not_default = aten_scalar_tensor_default = aten_full_like_default = None
    aten_permute_copy_default = executorch_exir_dialects_edge__ops_aten_permute_copy_default(k, [0, 1, 3, 2]);  k = None
    aten_mul_scalar_1 = executorch_exir_dialects_edge__ops_aten_mul_Scalar(aten_permute_copy_default, 0.5946035575013605);  aten_permute_copy_default = None
    aten_expand_copy_default = executorch_exir_dialects_edge__ops_aten_expand_copy_default(aten_mul_scalar, [1, 1, 8, 8]);  aten_mul_scalar = None
    aten_view_copy_default = executorch_exir_dialects_edge__ops_aten_view_copy_default(aten_expand_copy_default, [1, 8, 8]);  aten_expand_copy_default = None
    aten_expand_copy_default_1 = executorch_exir_dialects_edge__ops_aten_expand_copy_default(aten_mul_scalar_1, [1, 1, 8, 8]);  aten_mul_scalar_1 = None
    aten_view_copy_default_1 = executorch_exir_dialects_edge__ops_aten_view_copy_default(aten_expand_copy_default_1, [1, 8, 8]);  aten_expand_copy_default_1 = None
    aten_bmm_default = executorch_exir_dialects_edge__ops_aten_bmm_default(aten_view_copy_default, aten_view_copy_default_1);  aten_view_copy_default = aten_view_copy_default_1 = None
    aten_view_copy_default_2 = executorch_exir_dialects_edge__ops_aten_view_copy_default(aten_bmm_default, [1, 1, 8, 8]);  aten_bmm_default = None
    aten_add_tensor = executorch_exir_dialects_edge__ops_aten_add_Tensor(aten_view_copy_default_2, aten_where_self);  aten_view_copy_default_2 = aten_where_self = None
    aten__softmax_default = executorch_exir_dialects_edge__ops_aten__softmax_default(aten_add_tensor, -1, False);  aten_add_tensor = None
    aten_expand_copy_default_2 = executorch_exir_dialects_edge__ops_aten_expand_copy_default(aten__softmax_default, [1, 1, 8, 8]);  aten__softmax_default = None
    aten_view_copy_default_3 = executorch_exir_dialects_edge__ops_aten_view_copy_default(aten_expand_copy_default_2, [1, 8, 8]);  aten_expand_copy_default_2 = None
    aten_expand_copy_default_3 = executorch_exir_dialects_edge__ops_aten_expand_copy_default(v, [1, 1, 8, 8]);  v = None
    aten_view_copy_default_4 = executorch_exir_dialects_edge__ops_aten_view_copy_default(aten_expand_copy_default_3, [1, 8, 8]);  aten_expand_copy_default_3 = None
    aten_bmm_default_1 = executorch_exir_dialects_edge__ops_aten_bmm_default(aten_view_copy_default_3, aten_view_copy_default_4);  aten_view_copy_default_3 = aten_view_copy_default_4 = None
    aten_view_copy_default_5 = executorch_exir_dialects_edge__ops_aten_view_copy_default(aten_bmm_default_1, [1, 1, 8, 8]);  aten_bmm_default_1 = None
    return (aten_view_copy_default_5,)
```
After applying the diff, we remove the following ops
```
    %aten_full_like_default : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.full_like.default](args = (%aten_index_tensor_2, 0), kwargs = {dtype: torch.float32, pin_memory: False, memory_format: torch.preserve_format})

    %aten_logical_not_default : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.logical_not.default](args = (%aten_index_tensor_2,), kwargs = {})

    %aten_scalar_tensor_default : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.scalar_tensor.default](args = (-inf,), kwargs = {dtype: torch.float32, layout: torch.strided, device: cpu})

    %aten_where_self : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.where.self](args = (%aten_logical_not_default, %aten_scalar_tensor_default, %aten_full_like_default), kwargs = {})

    %aten_mul_scalar : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.mul.Scalar](args = (%aten_permute_copy_default_3, 0.5946035575013605), kwargs = {})
    ...
    %aten_mul_scalar_1 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.mul.Scalar](args = (%aten_permute_copy_default_6, 0.5946035575013605), kwargs = {})
```
but introduce an add
    %aten_add_tensor_3 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.add.Tensor](args = (%aten_mul_tensor_11, %aten_index_tensor_2), kwargs = {})
```
ghstack-source-id: 223152096
@exported-using-ghexport

Differential Revision: [D56119737](https://our.internmc.facebook.com/intern/diff/D56119737/)
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@kimishpatel kimishpatel kimishpatel approved these changes

@kirklandsign kirklandsign kirklandsign approved these changes

@chunit-quic chunit-quic Awaiting requested review from chunit-quic

@haowhsu-quic haowhsu-quic Awaiting requested review from haowhsu-quic

@shewu-quic shewu-quic Awaiting requested review from shewu-quic

@chiwwang chiwwang Awaiting requested review from chiwwang

Assignees

No one assigned

Labels

CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. fb-exported Merged

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

4 participants

@cccclai @facebook-github-bot @kimishpatel @kirklandsign