Modify perplexity script #108
Closed
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
SunMarc
commented
Mar 5, 2024
| @@ -64,7 +275,7 @@ def perplexity( | |||
| stride: int = 512, | |||
| ): | |||
| dtype = torch.float32 if device.type == "cpu" else torch.float16 | |||
| tokenizer = AutoTokenizer.from_pretrained(model_id) | |||
| tokenizer = AutoTokenizer.from_pretrained(model_id, use_fast=False) | |||
There was a problem hiding this comment.
I set it to False since the slow tokenizer is better at processing long string. I only takes a few seconds to process the dataset this way compared to the fast tokenizer (default)
|
Merged as #110 that removes also the old code. |
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Outdated suggestions cannot be applied.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 ?
This PR changes the script to calculate the perplexity. This perplexity calculation is compatible with the one in llama.cpp, so we can compare the results with ggml model. See the following thread for more information. I used it a lot for calculating the perplexity of quantized models such as awq, gptq.
With this script, we get the correct perplexity for gemma or mistral. cc @younesbelkada