sampling: add Top-nσ sampler#11223
Merged
ggerganov merged 23 commits intoggml-org:masterfrom Feb 13, 2025
Merged
Conversation
Contributor
|
Thank you for this implementation! Top-nσ is definitely special and needs a lot of testing. I like the results so far, especially since high temperature is not a problem, as shown in the paper, and I'm going to test it more and see what its limitations are. |
slaren
reviewed
Jan 16, 2025
slaren
approved these changes
Jan 24, 2025
|
Sorry for the ping, but what is needed so that this approved PR gets merged? I like the new feature very much as it allows the benefits of a higher temperature without the usual drawbacks. |
Contributor
Author
|
@hdu-hh I agree! This would be a cool feature to have but I don't think it's a priority for the maintainers at the moment. |
ggerganov
approved these changes
Feb 12, 2025
Co-authored-by: Georgi Gerganov <[email protected]>
Co-authored-by: Georgi Gerganov <[email protected]>
Co-authored-by: Georgi Gerganov <[email protected]>
Co-authored-by: Georgi Gerganov <[email protected]>
Co-authored-by: Georgi Gerganov <[email protected]>
Co-authored-by: Georgi Gerganov <[email protected]>
orca-zhang
pushed a commit
to orca-zhang/llama.cpp
that referenced
this pull request
Feb 26, 2025
* initial sampling changes: * completed top nsigma sampler implementation * apply parameter to only llama-cli * updated readme * added tests and fixed nsigma impl * cleaned up pr * format * format * format * removed commented tests * cleanup pr and remove explicit floats * added top-k sampler to improve performance * changed sigma to float * fixed string format to float * Update src/llama-sampling.cpp Co-authored-by: Georgi Gerganov <[email protected]> * Update common/sampling.cpp Co-authored-by: Georgi Gerganov <[email protected]> * Update src/llama-sampling.cpp Co-authored-by: Georgi Gerganov <[email protected]> * Update src/llama-sampling.cpp Co-authored-by: Georgi Gerganov <[email protected]> * Update src/llama-sampling.cpp Co-authored-by: Georgi Gerganov <[email protected]> * Update src/llama-sampling.cpp Co-authored-by: Georgi Gerganov <[email protected]> * added llama_sampler_init --------- Co-authored-by: Georgi Gerganov <[email protected]>
arthw
pushed a commit
to arthw/llama.cpp
that referenced
this pull request
Feb 26, 2025
* initial sampling changes: * completed top nsigma sampler implementation * apply parameter to only llama-cli * updated readme * added tests and fixed nsigma impl * cleaned up pr * format * format * format * removed commented tests * cleanup pr and remove explicit floats * added top-k sampler to improve performance * changed sigma to float * fixed string format to float * Update src/llama-sampling.cpp Co-authored-by: Georgi Gerganov <[email protected]> * Update common/sampling.cpp Co-authored-by: Georgi Gerganov <[email protected]> * Update src/llama-sampling.cpp Co-authored-by: Georgi Gerganov <[email protected]> * Update src/llama-sampling.cpp Co-authored-by: Georgi Gerganov <[email protected]> * Update src/llama-sampling.cpp Co-authored-by: Georgi Gerganov <[email protected]> * Update src/llama-sampling.cpp Co-authored-by: Georgi Gerganov <[email protected]> * added llama_sampler_init --------- Co-authored-by: Georgi Gerganov <[email protected]>
ddh0
added a commit
to ddh0/easy-llama
that referenced
this pull request
Apr 23, 2025
- update libllama to match `llama.cpp/include/llama.h@71e90e8813f90097701e62f7fce137d96ddf41e2` - add top-n-sigma sampler (ref: [llama.cpp#11223](ggml-org/llama.cpp#11223)) - add missing newlines to Gemma3 prompt format - add new `Llama.n_head_kv()`, `Llama.bpw()`, `Llama.warmup()` - add sampler parameter info the the sampler chain string - adopt new sampler parameter defaults (temp=1.0, min_p=0.1, all others neutral) - improved logic for applying the penalties sampler - warming-up the model now does a single token decode as well as a full batch decode - fix `Llama.chat_template()` - fix return type annotations for some token methods of Llama (`int` -> `Optional[int]`) - remove old sampler presets - continue working on new server + webui
|
We are just a small step away from bringing it to users: #14637 |
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
7 participants
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.
Top-nσ: Not All Logits Are You Need
https://arxiv.org/pdf/2411.07641
The authors of this paper propose a new sampling method known as Top-nσ. The main feature of this sampler is that "unlike existing methods (e.g., top-p, min-p) that inadvertently include more noise tokens at higher temperatures, top-nσ maintains a stable sampling space regardless of temperature scaling". They discovered that logits natually separate into a gaussian-distributed noisy region and an informative region.
This PR implements the sampling method proposed in the paper. Here the algorithm implemented from the paper:
Since the manipulation is done directly on the logits pre-softmax, I added it as a stand-alone sampler instead of chaining it with the common samplers. The changes only add support for
llama-cli.sampler chain: logits -> logit-bias -> temp -> top-n-sigma -> distI'm aware that this algorithm is still in it's early phases so we could tag this as demo for now but I'll leave that choice up to the maintainers.
resolves #11057
Relavent Links:
https://huggingface.co/papers/2411.07641
https://arxiv.org/pdf/2411.07641
https://github.com/Tomorrowdawn/top_nsigma
#11057