[BYOC][ACL] Add support for dense (fully connected) layer #6254
Conversation
This file contains 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 patch adds the ability to offload dense (or fully connected) operators to ACL. For fp32 a single dense layer can be offloaded, or the composite variant: nn.dense, nn.bias_add? (ACL does not currently offer fused activation). For uint8: qnn.dense, nn.bias_add?, qnn.requantize Change-Id: I83ea00b2aa6bdc5d9ef5cd6d54bbf981e523bd14
Change-Id: I856eb2298499fdf22c172ba7f85d21033d3cc920
comaniac
approved these changes
Aug 12, 2020
| @@ -114,8 +141,26 @@ def check_qnn_conv(extract): | |||
| call = call.args[0] | |||
| return qnn_conv2d(call.attrs, call.args) | |||
|
|
|||
| def check_dense(extract): | |||
| """Check conv pattern is supported by ACL.""" | |||
| call = extract | |||
There was a problem hiding this comment.
After reading check_qnn_dense, do you need to check the out_dtype here?
There was a problem hiding this comment.
The out_dtype attribute check is actually on the requantize node. The first node in the extract for fp32 would be nn.bias_add which doesn't have the out_dtype attribute. The out_dtype is checked though in the dense function below. Hope that makes sense :)
zhiics
approved these changes
Aug 13, 2020
trevor-m
pushed a commit
to trevor-m/tvm
that referenced
this pull request
Aug 26, 2020
* [BYOC][ACL] Add support for dense (fully connected) layer This patch adds the ability to offload dense (or fully connected) operators to ACL. For fp32 a single dense layer can be offloaded, or the composite variant: nn.dense, nn.bias_add? (ACL does not currently offer fused activation). For uint8: qnn.dense, nn.bias_add?, qnn.requantize Change-Id: I83ea00b2aa6bdc5d9ef5cd6d54bbf981e523bd14 * Don't offload dense layer with unsupported datatype Change-Id: I856eb2298499fdf22c172ba7f85d21033d3cc920
trevor-m
pushed a commit
to trevor-m/tvm
that referenced
this pull request
Aug 26, 2020
* [BYOC][ACL] Add support for dense (fully connected) layer This patch adds the ability to offload dense (or fully connected) operators to ACL. For fp32 a single dense layer can be offloaded, or the composite variant: nn.dense, nn.bias_add? (ACL does not currently offer fused activation). For uint8: qnn.dense, nn.bias_add?, qnn.requantize Change-Id: I83ea00b2aa6bdc5d9ef5cd6d54bbf981e523bd14 * Don't offload dense layer with unsupported datatype Change-Id: I856eb2298499fdf22c172ba7f85d21033d3cc920
trevor-m
pushed a commit
to trevor-m/tvm
that referenced
this pull request
Aug 26, 2020
* [BYOC][ACL] Add support for dense (fully connected) layer This patch adds the ability to offload dense (or fully connected) operators to ACL. For fp32 a single dense layer can be offloaded, or the composite variant: nn.dense, nn.bias_add? (ACL does not currently offer fused activation). For uint8: qnn.dense, nn.bias_add?, qnn.requantize Change-Id: I83ea00b2aa6bdc5d9ef5cd6d54bbf981e523bd14 * Don't offload dense layer with unsupported datatype Change-Id: I856eb2298499fdf22c172ba7f85d21033d3cc920
trevor-m
pushed a commit
to trevor-m/tvm
that referenced
this pull request
Sep 2, 2020
* [BYOC][ACL] Add support for dense (fully connected) layer This patch adds the ability to offload dense (or fully connected) operators to ACL. For fp32 a single dense layer can be offloaded, or the composite variant: nn.dense, nn.bias_add? (ACL does not currently offer fused activation). For uint8: qnn.dense, nn.bias_add?, qnn.requantize Change-Id: I83ea00b2aa6bdc5d9ef5cd6d54bbf981e523bd14 * Don't offload dense layer with unsupported datatype Change-Id: I856eb2298499fdf22c172ba7f85d21033d3cc920
trevor-m
pushed a commit
to neo-ai/tvm
that referenced
this pull request
Sep 3, 2020
* [BYOC][ACL] Add support for dense (fully connected) layer This patch adds the ability to offload dense (or fully connected) operators to ACL. For fp32 a single dense layer can be offloaded, or the composite variant: nn.dense, nn.bias_add? (ACL does not currently offer fused activation). For uint8: qnn.dense, nn.bias_add?, qnn.requantize Change-Id: I83ea00b2aa6bdc5d9ef5cd6d54bbf981e523bd14 * Don't offload dense layer with unsupported datatype Change-Id: I856eb2298499fdf22c172ba7f85d21033d3cc920
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
This patch adds the ability to offload dense (or fully connected) operator to ACL.
For fp32 a single dense layer can be offloaded, or the composite variant: nn.dense, nn.bias_add? (ACL does not currently offer fused activation).
For uint8: qnn.dense, nn.bias_add?, qnn.requantize
cc @mbaret @Leo-arm @giuseros @comaniac @zhiics