Improve tool call success rate: Allow arbitrary tool call parameter ordering for up to 8 params#13
Closed
florianbrede-ayet wants to merge 21 commits intopwilkin:autoparserfrom
Closed
Conversation
… fix Apertus template to require proper OpenAI compatible paths to tools
Co-authored-by: Xuan-Son Nguyen <thichthat@gmail.com>
Co-authored-by: Xuan-Son Nguyen <thichthat@gmail.com>
…tent in template [no ci]
…d Kimi 2.5 Thinking PEG parser
…rdering for up to 8 params.
pwilkin
force-pushed
the
autoparser
branch
2 times, most recently
from
c8f7024 to
2595444
Compare
Owner
|
See ggml-org#20171 |
Author
|
@pwilkin I built your branch and ran my toolcall-tester.py which I used before to verify my solution. I still have the feeling that your narrower implementation could generate tool call failures for XML trained models. I'll close this PR. |
Owner
|
@florianbrede-ayet originally I had the same approach as you did with permutations, but the co-maintainers convinced me that it's needless complication in practice. |
moonshadow-25
pushed a commit
to moonshadow-25/llama.cpp
that referenced
this pull request
Mar 14, 2026
* FlashAttention (pwilkin#13) * Add inplace softmax * Move rms_norm to split row approach * Update debug for supports_op * clean up debug statements * neg f16xf32xip builds and runs, havent actually ran a model that uses neg kernel yet though * neg passes backend test * unary operators pass ggml tests * rms_norm double declaration bug atoned * abides by editor-config * removed vestigial files * fixed autoconfig * All operators (inlcluding xielu) working * removed unnecesarry checking if node->src[1] exists for unary operators * responded and dealt with PR comments * implemented REPL_Template support and removed bug in unary operators kernel * formatted embed wgsl and ggml-webgpu.cpp * Faster tensors (pwilkin#8) Add fast matrix and matrix/vector multiplication. * Use map for shader replacements instead of pair of strings * Wasm (pwilkin#9) * webgpu : fix build on emscripten * more debugging stuff * test-backend-ops: force single thread on wasm * fix single-thread case for init_tensor_uniform * use jspi * add pthread * test: remember to set n_thread for cpu backend * Add buffer label and enable dawn-specific toggles to turn off some checks * Intermediate state * Fast working f16/f32 vec4 * Working float fast mul mat * Clean up naming of mul_mat to match logical model, start work on q mul_mat * Setup for subgroup matrix mat mul * Basic working subgroup matrix * Working subgroup matrix tiling * Handle weirder sg matrix sizes (but still % sg matrix size) * Working start to gemv * working f16 accumulation with shared memory staging * Print out available subgroup matrix configurations * Vectorize dst stores for sg matrix shader * Gemv working scalar * Minor set_rows optimization (pwilkin#4) * updated optimization, fixed errors * non vectorized version now dispatches one thread per element * Simplify * Change logic for set_rows pipelines --------- Co-authored-by: Neha Abbas <nehaabbas@macbookpro.lan> Co-authored-by: Neha Abbas <nehaabbas@ReeseLevines-MacBook-Pro.local> Co-authored-by: Reese Levine <reeselevine1@gmail.com> * Comment on dawn toggles * Working subgroup matrix code for (semi)generic sizes * Remove some comments * Cleanup code * Update dawn version and move to portable subgroup size * Try to fix new dawn release * Update subgroup size comment * Only check for subgroup matrix configs if they are supported * Add toggles for subgroup matrix/f16 support on nvidia+vulkan * Make row/col naming consistent * Refactor shared memory loading * Move sg matrix stores to correct file * Working q4_0 * Formatting * Work with emscripten builds * Fix test-backend-ops emscripten for f16/quantized types * Use emscripten memory64 to support get_memory * Add build flags and try ci --------- Co-authored-by: Xuan Son Nguyen <son@huggingface.co> * Remove extra whitespace * Move wasm single-thread logic out of test-backend-ops for cpu backend * Disable multiple threads for emscripten single-thread builds in ggml_graph_plan * Refactored pipelines and workgroup calculations (pwilkin#10) * refactored pipelines * refactored workgroup calculation * removed commented out block of prior maps * Clean up ceiling division pattern --------- Co-authored-by: Neha Abbas <nehaabbas@eduroam-169-233-141-223.ucsc.edu> Co-authored-by: Reese Levine <reeselevine1@gmail.com> * Start work on flash attention * Shader structure set up (many bugs still) * debugging * Working first test * Working with head grouping, head sizes to 128, logit softcap, mask/sinks enabled, f32 * Generalize softmax to work with multiple subgroups, f16 accumulation, mask shared memory tiling * Start work on integrating pre-wgsl * Separate structs/initial shader compilation library into separate files * Work on compilation choices for flashattention * Work on subgroup matrix/tile size portability * subgroup size agnostic online softmax * Cleanups, quantization types * more cleanup * fix wasm build * Refactor flashattention to increase parallelism, use direct loads for KV in somce cases * Checkpoint * formatting * Update to account for default kv cache padding * formatting shader * Add workflow for ggml-ci webgpu * Try passing absolute path to dawn in ggml-ci * Avoid error on device destruction, add todos for proper cleanup * Fix unused warning * Forgot one parameter unused * Move some flashattn computation to f32 for correctness
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
2 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.
Thanks @pwilkin for working on this branch, it's a nice improvement over the fixed (and often partially broken) chat templates.
However, I noticed tool call failures especially under mistral-vibe which I could reliably reproduce (
read_filewithoffsetandlimit).This seems to happen for any qwen35 model, including 27b dense (albeit with a higher chance of correct order).
I debugged your autoparser with claude and found that tool calls enforce a strict parameter order.
Qwen with his native XML tool calls has a very high chance to generate the parameter entities in a very particular order (also tested at different temperatures and penalties - with different seeds you can have a random chance of tool calls succeeding) which does not neccessarily match the expected order in the autoparser.
To limit the number of permutations, I set the "allow any order" to a hard cap of 8 parameters (fallback to sequential order otherwise).
Disclosure: Code was mostly written by Opus, I ran the tests and built it against ROCm and tested it with several hundred tool calls. I don't have a CI/CD setup locally.
Make sure to read the contributing guidelines before submitting a PR