Repository navigation
test: add real-expert packed-scale GPU correctness gate - #86
Merged
Merged
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Adds an executable one-GPU gate for #84 using eight real Vision drafter experts at TP2 w13/w2 shapes, indexed top6 routing, FP8 activation quantization and both packed/unpacked scale formats. No full model loading.
Checks finite outputs, repeat stability, bitwise format equality and individual CUDA graph capture/replay with changed/restored inputs. Fail-closed: exit0 plus PACKED_SCALE_KERNEL_GATE_PASS only on success. Timings include harness allocation/initialization, not isolated GEMM latency. Compilation workers capped1 and CPU threads2.
Validated here: syntax, --help and15 repository CPU tests. GPU execution is pending: this script is a test handoff, not evidence the kernels pass. It does not replace full MoE/serving correctness or memory profiling.
Example inside the #84 image, script mounted /probe.py and model mounted read-only:
Use an isolated compile cache, one SM90 GPU,4 CPUs,16GiB RAM. Repeat layers1/2 and TP rank1 if the window permits. No production changes.