NousResearch/hermes-agent

MTP models spilling lose large prefill microbatch setting

Cluster 137685 · hermes cli. Data as of

3 open items. Fix attempted, not merged.

The short answer

1open issue
2open PRs
4items grouped
Known?
Yes. 4 related items grouped here, 0 issues already closed.
Fix event
No fix has been merged to main. 1 PR closed without merging; see the attempts below.
Tracker
The maintainers' tracker marks this unresolved. That is separate from the fix event.
Still open
1 issue and 2 PRs.

These items concern local-runtime handling of MTP models that are too large for the card and spill. One reports that the child process launches without -b/-ub flags (falling back to the engine default of 512), and the two fix-titled items appear to address retaining the large prefill microbatch in that case.

Open issues1

  • Issue#137685[Bug]: MTP models too large to fit the card never get the large prefill microbatch — child launches with no -b/-ub (engine default 512)

Open PRs2

  • PR#137689fix(local-runtime): retain MTP prefill when spilling
  • PR#141542fix(local-runtime): an MTP model that spills in both postures keeps the large prefill microbatch

1 PR closed without merging

From the close comments: superseded by another PR 1

  • #137694 fix(local-runtime): an MTP model that spills in both postures keeps the large prefill microbatch · superseded by another PR → #137689
How this cluster was grouped
  • Link evidence (structural links throughout): every member is connected by structural evidence: shared changed files, or a closing/superseded reference.
  • Pairwise link density 0.50: the share of item pairs that carry a link, counting a structural link in full and a relevance-only link at 0.36. 1.00 means every pair is linked.

Pick up a Lab task

Check whether a fix covers the open issues here, with your own agent. Bring back evidence for maintainers.

Open the Lab

At a glance

Subsystem
hermes cli
Closed issues
0
Attempts
1 PR closed without merging

For your agent

The same facts, typed.