-
Notifications
You must be signed in to change notification settings - Fork 402
[Plugin TRT EP] Complete GetCapability() implementation #553
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
[Plugin TRT EP] Complete GetCapability() implementation #553
Conversation
plugin_execution_providers/tensorrt/src/tensorrt_execution_provider.cc
Outdated
Show resolved
Hide resolved
| size_t max_workspace_size_ = 1 << 30; // 1GB | ||
| bool fp16_enable_ = false; | ||
| bool int8_enable_ = false; | ||
| bool bf16_enable_ = false; |
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.
is bf16 support fully implemented? it looks like this is only used to determine whether to set the flag nvinfer1::NetworkDefinitionCreationFlag::kSTRONGLY_TYPED.
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.
Good catch, added the missing parts for bf16 support.
This PR makes following changes:
GetSupportedList()inGetCapability()so that it determines whether the entire graph or a subgraph (i.e., a subset of nodes from the original graph forming a new graph) should be assigned to the plugin TensorRT EP, based on the results from calling the TensorRT parser.src/folder