Data stream 2 - #5
Merged
Merged
Conversation
Signed-off-by: Sami Jaghouar <sami.jaghouar@gmail.com>
Signed-off-by: Sami Jaghouar <sami.jaghouar@gmail.com>
Signed-off-by: Sami Jaghouar <sami.jaghouar@gmail.com>
Signed-off-by: Sami Jaghouar <sami.jaghouar@gmail.com>
Signed-off-by: Sami Jaghouar <sami.jaghouar@gmail.com>
Signed-off-by: Sami Jaghouar <sami.jaghouar@gmail.com>
Signed-off-by: Sami Jaghouar <sami.jaghouar@gmail.com>
Signed-off-by: Sami Jaghouar <sami.jaghouar@gmail.com>
Signed-off-by: Sami Jaghouar <sami.jaghouar@gmail.com>
Signed-off-by: Sami Jaghouar <sami.jaghouar@gmail.com>
Signed-off-by: Sami Jaghouar <sami.jaghouar@gmail.com>
Signed-off-by: Sami Jaghouar <sami.jaghouar@gmail.com>
Signed-off-by: Sami Jaghouar <sami.jaghouar@gmail.com>
Signed-off-by: Sami Jaghouar <sami.jaghouar@gmail.com>
Signed-off-by: Sami Jaghouar <sami.jaghouar@gmail.com>
Signed-off-by: Sami Jaghouar <sami.jaghouar@gmail.com>
Signed-off-by: Sami Jaghouar <sami.jaghouar@gmail.com>
Signed-off-by: Sami Jaghouar <sami.jaghouar@gmail.com>
Signed-off-by: Sami Jaghouar <sami.jaghouar@gmail.com>
samsja
added a commit
that referenced
this pull request
Mar 30, 2026
* add swebench lite * fix dataset name --------- Co-authored-by: Mika Senghaas <mail@mikasenghaas.de>
seanbell
added a commit
to clouddatalabs/scalerl-prime-rl
that referenced
this pull request
Apr 26, 2026
…polish - scripts/prebuild_tb_images.sh: drop `set -e`; report per-task succeeded/failed at end and exit non-zero if any failed (an upstream task with a broken Dockerfile, e.g. log-summary's missing logs/ dir, no longer aborts the remaining 27 builds). - docs/SCALERL.md "4xB200" → "single 8-GPU B200 node (4 train + 4 infer GPUs split per the [deployment] block)" — disambiguates the steady-state perf number for Snowflake reading capacity sizing (Snowflake POC critique PrimeIntellect-ai#5). - configs/scalerl_math/rl.toml: add doc note that the implicit `use_token_client = true` default is correct ONLY for single-turn envs. Cross-references the TB config's full rationale so a future user copying scalerl_math to bootstrap a multi-turn env doesn't silently get the linear-history corruption mode (Snowflake POC critique PrimeIntellect-ai#6).
snimu
added a commit
that referenced
this pull request
Jun 4, 2026
- enabled_losses=None now validated as the full term list, so >1 echo term per env is caught at config time instead of at rollout time. [review #6] - loss_overrides keys validated against `losses`; non-echo overrides rejected. [#7] - warn (don't fail) when prompt-role echo is configured with renderer=None (MITO), where prompt_attribution is unavailable so it would silently no-op. [#8] - token_export: add echo_mask/echo_weight columns + export sequences trained only via echo (gate on loss_mask OR echo_mask). [#9] - doc notes: echo CE uses the rollout temperature (scale alpha to compensate, kept as-is); negative alpha is intentional (suppresses tokens). [#1, #10] - tests for the new config validators. Deferred to a follow-up pass (per the review): full per-sample primary routing / rl-disable [#2b] + the <=1-primary validation it enables [#5], and the multi-run losses fingerprint [#3]. Not run locally; ruff + py_compile clean. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
snimu
added a commit
that referenced
this pull request
Jun 4, 2026
…#2/#5) - enabled_losses can now disable the primary loss for an env: when the term matching training_mode isn't enabled, train_sink zeroes that env's samples' completion_mask (after the decode/prefill metric), so the primary core trains nothing while echo (its own mask) still applies. [#2b] - setup_loss_fns no longer fabricates a default rl core when no rl/custom term exists; the rl core raises if an rl-mode batch is applied. [#2a, trainer-side] - OrchestratorConfig.validate_primary_loss: training_mode must have a matching term (rl/custom | sft | opd) in `losses`. [#2a, config-time] - LossList rejects more than one primary (rl/custom) term. [#5] - tests for the validators + the no-rl-term rl-core error. Not run locally; ruff + py_compile clean. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
snimu
added a commit
that referenced
this pull request
Jun 5, 2026
…ht resolver - #1 reserve loss-term names {sft, opd} (any term) and `rl` (non-primary), so an overlay can't silently overwrite a training_mode dispatch core or the rl primary in the trainer registry. - #5 resolve the primary's advantage weight orchestrator-side: scale the per-token advantage by the advantage-weight's tau in process_group and drop adv_tau from the dppo_kl core / RLLossConfig. Now *any* primary core (dppo_kl or custom) gets the resolved advantage × tau — no per-core special-case. Bit-identical for the default tau=1.0. - #4 overlay trainability = non-None AND non-zero, so a zero-weight overlay (e.g. advantage-weighted with zero advantage) no longer keeps an otherwise-empty batch alive past the empty-batch guard. - #6 custom overlay weight resolver is group-aware: it now receives `WeightInputs{sample, rollouts}` (the full GRPO group) instead of a lone sample, so it can compute group-relative weights. - #10 document the overlay_mask/overlay_weight token-export columns (schema v2) in the configs skill. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
snimu
added a commit
that referenced
this pull request
Jun 10, 2026
…ted) rollout Bug review #5: the token-backfill path (renderer build_trajectory_step) doesn't emit prompt_attribution, so prompt-side role tokens (system/user/tool) stay unattributed and prompt-role echo silently trains nothing on backfilled rollouts (e.g. SFT against external APIs). Completion/assistant echo is unaffected. Loud guard rather than a silent no-op: TrainSink precomputes whether any echo term targets a role beyond "assistant", and process_rollout warns once when such a config meets a backfilled rollout. The full fix (emit prompt_attribution from build_trajectory_step, or attribute in the backfill path) is noted in the warning and left for the renderer submodule. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
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.
Okay so this new version of the data loader
As discussed offline we are following this pattern:
seems that we don't have performance issue, ( done on 2xa40 with a 150m)
Maybe need to check performance a bit more once we have e2e run