Skip to content

Rename the trainer group and cell to TrainerController and TrainerCell - #2176

Open
fzyzcjy wants to merge 2 commits into
tom/refactor-miles/op13-171from
tom/refactor-miles/op13-142
Open

Rename the trainer group and cell to TrainerController and TrainerCell#2176
fzyzcjy wants to merge 2 commits into
tom/refactor-miles/op13-171from
tom/refactor-miles/op13-142

Conversation

@fzyzcjy

@fzyzcjy fzyzcjy commented Aug 4, 2026

Copy link
Copy Markdown
Collaborator

Part of #1837

ci-sglang-pr: tom/refactor-miles-repo-sglang/deliver-1

@gemini-code-assist

Copy link
Copy Markdown
Contributor

Caution

The consumer version of Gemini Code Assist on GitHub has been sunset. All code review activity has officially ceased.

@fzyzcjy
fzyzcjy force-pushed the tom/refactor-miles/op13-142 branch from c31b653 to fac7194 Compare August 6, 2026 03:47
@fzyzcjy
fzyzcjy force-pushed the tom/refactor-miles/op13-160 branch from 897c0ec to 03a6a74 Compare August 6, 2026 03:47
@fzyzcjy
fzyzcjy force-pushed the tom/refactor-miles/op13-142 branch 2 times, most recently from dfc4680 to d11bd3a Compare August 9, 2026 23:32
@fzyzcjy
fzyzcjy force-pushed the tom/refactor-miles/op13-160 branch from 98c9464 to 3f4bf39 Compare August 10, 2026 00:44
@fzyzcjy
fzyzcjy force-pushed the tom/refactor-miles/op13-142 branch 2 times, most recently from 190c45a to 7bbce62 Compare August 10, 2026 06:01
@fzyzcjy
fzyzcjy force-pushed the tom/refactor-miles/op13-160 branch from 3f4bf39 to c240135 Compare August 10, 2026 06:02
@fzyzcjy
fzyzcjy force-pushed the tom/refactor-miles/op13-142 branch from 7bbce62 to ca43f31 Compare August 10, 2026 16:06
@fzyzcjy
fzyzcjy requested a review from Ying1123 as a code owner August 10, 2026 16:06
@fzyzcjy
fzyzcjy force-pushed the tom/refactor-miles/op13-160 branch from c240135 to 6e08c81 Compare August 10, 2026 16:06
@fzyzcjy
fzyzcjy changed the base branch from tom/refactor-miles/op13-160 to tom/refactor-miles/op13-171 August 10, 2026 16:12
@fzyzcjy
fzyzcjy force-pushed the tom/refactor-miles/op13-142 branch from ca43f31 to c706796 Compare August 10, 2026 16:33
@fzyzcjy
fzyzcjy force-pushed the tom/refactor-miles/op13-171 branch from b95dc89 to 3a78d51 Compare August 10, 2026 16:34
@fzyzcjy
fzyzcjy force-pushed the tom/refactor-miles/op13-142 branch from c706796 to 098853c Compare August 11, 2026 00:45
@fzyzcjy
fzyzcjy force-pushed the tom/refactor-miles/op13-171 branch from 3a78d51 to 95de5ea Compare August 12, 2026 02:48
@fzyzcjy
fzyzcjy force-pushed the tom/refactor-miles/op13-142 branch from 098853c to 8021c90 Compare August 12, 2026 02:48
@fzyzcjy fzyzcjy mentioned this pull request Aug 12, 2026
The train actor used to tell the rollout side which weights the engines
now serve via `self.rollout_manager`, but that attribute stopped being
assigned once the actor was decoupled from the rollout side, leaving two
dead call sites: the FSDP one also referenced an unimported `ray`. The
executor's `weight_version` therefore stayed None forever and every
sample it collected went to training unstamped.

The actor's `update_weights` now answers the version it broadcast, the
trainer controller passes it up, and the driver writes it into the
executor through one `update_weights` helper the three training loops
share. An actor that skipped the broadcast answers None and nothing is
published.

The driver owns the write because it is the only place that holds both
handles for the whole chain: the controller's own executor handle goes
away when the trainer controller becomes an independent worker.
@fzyzcjy
fzyzcjy force-pushed the tom/refactor-miles/op13-171 branch from 95de5ea to 3da72e3 Compare August 12, 2026 09:32
@fzyzcjy
fzyzcjy force-pushed the tom/refactor-miles/op13-142 branch from a58b3f6 to 0963e0e Compare August 12, 2026 09:32
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant