imatrix: calculate activation-based statistics for new format (GGUF) imatrices #14891
Conversation
This file contains hidden or 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
compilade
reviewed
Jul 26, 2025
| if (!stat.activations.empty()) { | ||
| const int32_t nact = (int32_t) stat.activations.size(); | ||
| struct ggml_tensor * in_sum = ggml_new_tensor_2d(ctx, GGML_TYPE_F32, nact / nmat, nmat); | ||
| ggml_format_name(in_sum, "%s.in_sum", name.c_str()); // ToDo: consider a better name. 'in_act' maybe? |
There was a problem hiding this comment.
I think in_sum is fine, this fits with the intention of in_sum2.
Comment on lines
+41
to
42
| std::vector<float> activations; | ||
| std::vector<float> values; |
There was a problem hiding this comment.
It might make sense to rename Stats.values to Stats.in_sum2, and Stats.activations to Stats.in_sum.
It should make it more obvious what maps to what in the resulting GGUF.
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Outdated suggestions cannot be applied.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.
Following up from #9400 and #12718, I've started tinkering with activation-based statistics, in addition to what's currently available via
--show-statistics.At the moment, I'm exploring three options going from from easy to implement and OK approximation, to some assembly required but fairly accurate:
llama-perplexity --save-all-logitsrun)llama-imatrixalready generates the actual logits to compute PPL, use Thông T. Nguyễn's logit prism approach to calculate the exact contribution of each layer to the final logit scoresSharing with the readers, and in particular @compilade and @jukofyork, in case anyone's willing to double check assumptions and/or suggest alternative approaches I haven't considered.