fix falcon dummy input generator #1825
Open
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
Co-authored-by: Michael Benayoun <mickbenayoun@gmail.com>
|
The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update. |
echarlaix
reviewed
Apr 25, 2024
Comment on lines
+1093
to
+1098
| if normalized_config.new_decoder_architecture and normalized_config.multi_query: | ||
| self.num_kv_heads = normalized_config.num_attention_heads | ||
| elif normalized_config.new_decoder_architecture: | ||
| self.num_kv_heads = normalized_config.num_kv_heads | ||
| else: | ||
| self.num_kv_heads = 1 |
There was a problem hiding this comment.
shouldn't it be :
Suggested change
| if normalized_config.new_decoder_architecture and normalized_config.multi_query: | |
| self.num_kv_heads = normalized_config.num_attention_heads | |
| elif normalized_config.new_decoder_architecture: | |
| self.num_kv_heads = normalized_config.num_kv_heads | |
| else: | |
| self.num_kv_heads = 1 | |
| if normalized_config.new_decoder_architecture: | |
| self.num_kv_heads = normalized_config.num_attention_heads | |
| else: | |
| self.num_kv_heads = normalized_config.num_kv_heads if not normalized_config.multi_query else 1 |
?
| elif normalized_config.new_decoder_architecture: | ||
| self.num_kv_heads = normalized_config.num_kv_heads | ||
| else: | ||
| self.num_kv_heads = 1 | ||
| self.head_dim = self.hidden_size // self.num_attention_heads | ||
|
|
There was a problem hiding this comment.
also why not have this value set in the normalized_config directly and use it in both FalconDummyPastKeyValuesGenerator and ORTFalconForCausalLM @fxmarty ? (removing the need to have this check in both places)
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.
What does this PR do?
during enabling falcon40b export for openvino, I found that this model is not covered by dummy input generator configuration and failed with shape mismatch. This PR fixes that.
I propose these changed into optimum-intel, but the proper place for it here.
Who can review?