Support for Phi-3 MLP layer #84
Merged
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
…hByrneIntel/intel-npu-acceleration-library into sarah/feature/phi3MLP_layer
…hByrneIntel/intel-npu-acceleration-library into sarah/feature/phi3MLP_layer
SarahByrneIntel
changed the title
SarahByrneIntel
changed the title
Comment on lines
46
to
53
| if isinstance(model, Phi3MLP): | ||
| # Apply optimizations to a single MLP block model | ||
| model = model | ||
|
|
||
| if dtype in (int8, int4): | ||
| # Quantize model | ||
| model = quantize_model(model, dtype) | ||
| weights_quantization(model) |
There was a problem hiding this comment.
Why there is a specific branch about Phi3MLP?
There was a problem hiding this comment.
If only a single mlp block is passed in to be compiled, we don't want to pass it to the recursive function as it will break it down into the layers. When the block is contained within a larger model, then it is the model that is broken down and we can prevent the blocks being broken down through the NPUModuleWrapper check. However, this won't happen if it is only a single block
alessandropalla
requested changes
Jul 19, 2024
added 2 commits
July 19, 2024 11:07
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Outdated suggestions cannot be applied.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.
Adding support for Phi-3 MLP layer
Update compile functionality for model blocks
Add Phi-3 MLP optimization
Add testing for Phi-3 MLP
Add type operation for tensor dtype conversion
Implement new forward function for quantized models
Add toggling for model profiling
Add compiler configuration feature
Update tests and examples for compiler config
Update doc on usage