Repository navigation
Conversation
T0's group backend could not abort GLM's communicator before the RoCE group was installed, because it only borrowed it. GLM now hands the backend the group that owns the communicator from Engine.init, so an abort always reaches it once, and the start-up unwind uses the group's guarded destroy, which skips the destroy after an abort.
A failed spawn returned while the ranks already started still wrote to the test's frame, the same pattern #2700 fixed in host.runWith. Both tests now join the ranks they started on every return; in the gather test that also covers rank 0's own steps between the spawns and the join. Every wait in Bootstrap has a deadline, so the join is bounded.
Each app run uploads its ~230 MB AppImage, and with GitHub's default 90-day retention 2,069 copies (475 GB) had piled up in Actions storage by 12 Nov, enough to hit the account's Actions limit. The build is smoke-tested in the job itself; the upload is only for a quick download, which 3 days covers.
A reply that keeps a tool-call envelope as text now carries tensorfold.kept_tool_calls with unoffered and malformed counts and the unoffered names the model wrote, capped at eight names of 64 bytes each. /metrics counts both families. docs/api.md no longer claims an unoffered call comes back as a tool call.
The four bad_tool cases (whole, stream, one call allowed, and its stream) keep an envelope that names the unoffered tool "other", so their tensorfold block now ends with kept_tool_calls: one unoffered, none malformed, names ["other"]. No other expected reply changes; the /metrics cases stay on the known-differences list, which they reach before the new counters.
…3 on other cards and groups
With the profiler on, prompt_pair.pairedAt and prompt_solo.at return false (the e.prompt_profs[3] != null term in cuda_prompt_pair.zig and cuda_prompt_solo.zig), so every prompt step is a single chunk. The environment table did not say so, and a profile could be read as the served shape.
Owner
Author
|
Stopped: main is frozen for the 1.2.8 release. These PRs land first after it. |
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.
CI for a landing train: #2814 #2697 #2990 stacked on main in that order. Nothing merges from this PR: each PR above lands on its own once this tip is green, and this PR is then closed.