[uc] add activation, norm, pos emb, quant, sdpa operators for uc backend. #342
There are no files selected for viewing
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change | ||||||||||||||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| @@ -0,0 +1,5 @@ | ||||||||||||||||||||||||||||||||
| from mojo_opset.utils.platform import get_impl_by_platform | ||||||||||||||||||||||||||||||||
|
|
||||||||||||||||||||||||||||||||
| _op_map = get_impl_by_platform() | ||||||||||||||||||||||||||||||||
| globals().update(_op_map) | ||||||||||||||||||||||||||||||||
| __all__ = list(_op_map.keys()) | ||||||||||||||||||||||||||||||||
|
Comment on lines
+1
to
+5
Contributor
There was a problem hiding this comment. The
Suggested change
|
||||||||||||||||||||||||||||||||
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1 @@ | ||
|
|
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change | ||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| @@ -0,0 +1,70 @@ | ||||||||||||||||||
| from functools import lru_cache | ||||||||||||||||||
|
|
||||||||||||||||||
| import torch | ||||||||||||||||||
|
|
||||||||||||||||||
|
|
||||||||||||||||||
| _DTYPE_API_SUFFIX = { | ||||||||||||||||||
| torch.float16: "fp16", | ||||||||||||||||||
| torch.bfloat16: "bf16", | ||||||||||||||||||
| torch.float32: "fp32", | ||||||||||||||||||
| } | ||||||||||||||||||
|
|
||||||||||||||||||
|
|
||||||||||||||||||
| @lru_cache(maxsize=1) | ||||||||||||||||||
| def _uc_kernels(): | ||||||||||||||||||
| import uc_kernel | ||||||||||||||||||
|
|
||||||||||||||||||
| return uc_kernel.load() | ||||||||||||||||||
|
|
||||||||||||||||||
|
|
||||||||||||||||||
| def _matrix_shape(tensor: torch.Tensor) -> tuple[int, int]: | ||||||||||||||||||
| if tensor.dim() == 0: | ||||||||||||||||||
| return 1, 1 | ||||||||||||||||||
| if tensor.dim() == 1: | ||||||||||||||||||
| return 1, tensor.numel() | ||||||||||||||||||
| return tensor.numel() // tensor.shape[-1], tensor.shape[-1] | ||||||||||||||||||
|
|
||||||||||||||||||
|
|
||||||||||||||||||
| def _typed_api(api: str, dtype: torch.dtype) -> str: | ||||||||||||||||||
| suffix = _DTYPE_API_SUFFIX.get(dtype) | ||||||||||||||||||
| if suffix is None: | ||||||||||||||||||
| raise NotImplementedError(f"UC backend {api} does not support dtype {dtype}.") | ||||||||||||||||||
|
|
||||||||||||||||||
| kernels = _uc_kernels() | ||||||||||||||||||
| typed_api = f"{api}_{suffix}" | ||||||||||||||||||
| if typed_api in kernels.keys(): | ||||||||||||||||||
| return typed_api | ||||||||||||||||||
| if dtype == torch.float16 and api in kernels.keys(): | ||||||||||||||||||
| return api | ||||||||||||||||||
|
Contributor
There was a problem hiding this comment. In Python, checking membership directly in a dictionary (e.g.,
Suggested change
|
||||||||||||||||||
| raise NotImplementedError(f"UC backend {api} does not provide a {suffix} kernel artifact.") | ||||||||||||||||||
|
|
||||||||||||||||||
|
|
||||||||||||||||||
| def run_unary_kernel(api: str, x: torch.Tensor) -> torch.Tensor: | ||||||||||||||||||
| if x.numel() == 0: | ||||||||||||||||||
| return torch.empty_like(x) | ||||||||||||||||||
|
|
||||||||||||||||||
| kernel_input = x.contiguous() | ||||||||||||||||||
| kernel_output = torch.empty_like(kernel_input) | ||||||||||||||||||
| rows, cols = _matrix_shape(kernel_input) | ||||||||||||||||||
| _uc_kernels()[_typed_api(api, kernel_input.dtype)](kernel_input, kernel_output, rows, cols) | ||||||||||||||||||
| return kernel_output.reshape(x.shape) | ||||||||||||||||||
|
|
||||||||||||||||||
|
|
||||||||||||||||||
| def run_binary_kernel(api: str, lhs: torch.Tensor, rhs: torch.Tensor) -> torch.Tensor: | ||||||||||||||||||
| if lhs.shape != rhs.shape: | ||||||||||||||||||
| raise ValueError(f"UC backend {api} expects matching input shapes, got {lhs.shape} and {rhs.shape}.") | ||||||||||||||||||
| if lhs.dtype != rhs.dtype: | ||||||||||||||||||
| raise ValueError(f"UC backend {api} expects matching input dtypes, got {lhs.dtype} and {rhs.dtype}.") | ||||||||||||||||||
| if lhs.numel() == 0: | ||||||||||||||||||
| return torch.empty_like(lhs) | ||||||||||||||||||
|
|
||||||||||||||||||
| kernel_lhs = lhs.contiguous() | ||||||||||||||||||
| kernel_rhs = rhs.contiguous() | ||||||||||||||||||
| kernel_output = torch.empty_like(kernel_lhs) | ||||||||||||||||||
| rows, cols = _matrix_shape(kernel_lhs) | ||||||||||||||||||
| _uc_kernels()[_typed_api(api, kernel_lhs.dtype)](kernel_lhs, kernel_rhs, kernel_output, rows, cols) | ||||||||||||||||||
| return kernel_output.reshape(lhs.shape) | ||||||||||||||||||
|
|
||||||||||||||||||
|
|
||||||||||||||||||
| def run_kernel(api: str, dtype: torch.dtype, *args) -> None: | ||||||||||||||||||
| _uc_kernels()[_typed_api(api, dtype)](*args) | ||||||||||||||||||
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,31 @@ | ||
| import torch | ||
|
|
||
| from mojo_opset.core import MojoGelu | ||
| from mojo_opset.core import MojoSilu | ||
| from mojo_opset.core import MojoSwiGLU | ||
|
|
||
| from ._utils import run_binary_kernel | ||
| from ._utils import run_unary_kernel | ||
|
|
||
|
|
||
| class UCGelu(MojoGelu): | ||
| supported_platforms_list = ["npu"] | ||
|
|
||
| def forward(self, x: torch.Tensor) -> torch.Tensor: | ||
| return run_unary_kernel("mojo_gelu", x) | ||
|
|
||
|
|
||
| class UCSilu(MojoSilu): | ||
| supported_platforms_list = ["npu"] | ||
|
|
||
| def forward(self, x: torch.Tensor) -> torch.Tensor: | ||
| return run_unary_kernel("mojo_silu", x) | ||
|
|
||
|
|
||
| class UCSwiGLU(MojoSwiGLU): | ||
| supported_platforms_list = ["npu"] | ||
|
|
||
| def forward(self, gate_out: torch.Tensor, up_out: torch.Tensor) -> torch.Tensor: | ||
| if self.swiglu_limit > 0: | ||
| return super().forward(gate_out, up_out) | ||
| return run_binary_kernel("mojo_swiglu", gate_out, up_out) |
Oops, something went wrong.
Oops, something went wrong.
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
This block of code is completely redundant. Setting
CLOSE_MATMUL_K_SHIFTwhenMOJO_DETERMINISTICis enabled is already handled at lines 11-17. Removing this redundant block keeps the code clean and maintainable.