Update release v0.1.2#5
Closed
nuzant wants to merge 34 commits into
Closed
Conversation
Merge branch fix-loss-scale of git@code.alipay.com:inclusionAI/AReaL.git into main https://code.alipay.com/inclusionAI/AReaL/pull_requests/1 Signed-off-by: 博惟 <bowei.fw@antgroup.com>
… window and initial loss scale. Merge branch fw/loss-scale of git@code.alipay.com:inclusionAI/AReaL.git into main https://code.alipay.com/inclusionAI/AReaL/pull_requests/2 Signed-off-by: 晓雷 <meizhiyu.mzy@antgroup.com> * .
different seeds for every epoch & every rank
Merge branch whj/fix-dataloader of git@code.alipay.com:inclusionAI/AReaL.git into main https://code.alipay.com/inclusionAI/AReaL/pull_requests/3 Signed-off-by: 晓雷 <meizhiyu.mzy@antgroup.com>
Merge branch fw/fix-dataloading-not-shuffle of git@code.alipay.com:inclusionAI/AReaL.git into main https://code.alipay.com/inclusionAI/AReaL/pull_requests/4 Signed-off-by: 晓雷 <meizhiyu.mzy@antgroup.com> * fw/fix-dataloading-not-shuffle * . * . * .
…vLLM generation. Merge branch fw/vllm-key-value-alloc of git@code.alipay.com:inclusionAI/AReaL.git into main https://code.alipay.com/inclusionAI/AReaL/pull_requests/7?tab=commit Signed-off-by: 晓雷 <meizhiyu.mzy@antgroup.com> * . * remove pring
Merge branch fw/balanced-datapck of git@code.alipay.com:inclusionAI/AReaL.git into main https://code.alipay.com/inclusionAI/AReaL/pull_requests/5 Signed-off-by: 温差 <xushusheng.xss@antgroup.com> * . * fw/fix-dataloading-not-shuffle * . * . * . * . * .
Merge branch math_verifier_timeout of git@code.alipay.com:inclusionAI/AReaL.git into main https://code.alipay.com/inclusionAI/AReaL/pull_requests/8 Signed-off-by: 博惟 <bowei.fw@antgroup.com> * debug: math verifier timeout * debug: math verifier timeout * timeout in math_verify_utils * format
…ecover. Merge branch fw/fix-recover-filter of git@code.alipay.com:inclusionAI/AReaL.git into main https://code.alipay.com/inclusionAI/AReaL/pull_requests/10 Signed-off-by: 晓雷 <meizhiyu.mzy@antgroup.com> * . * cleanup * cleanup
Merge branch update-tutorial of git@code.alipay.com:inclusionAI/AReaL.git into main https://code.alipay.com/inclusionAI/AReaL/pull_requests/12?tab=comment Signed-off-by: 晓雷 <meizhiyu.mzy@antgroup.com> * support setup env and start train * update tutorial * update training script
Merge branch fw/update-requirements of git@code.alipay.com:inclusionAI/AReaL.git into main https://code.alipay.com/inclusionAI/AReaL/pull_requests/14 Signed-off-by: 晓雷 <meizhiyu.mzy@antgroup.com> * format and update requirements.txt * cleanup
Merge branch 0304-7b-zero-readme of git@code.alipay.com:inclusionAI/AReaL.git into main https://code.alipay.com/inclusionAI/AReaL/pull_requests/13 Signed-off-by: 晓雷 <meizhiyu.mzy@antgroup.com> * update 7B zero figures * update readme * update amc * update readme * Delete Data Comparison * typo * update figures * fix typo
…and memory allocation Merge branch ranghou-math of git@code.alipay.com:inclusionAI/AReaL.git into main https://code.alipay.com/inclusionAI/AReaL/pull_requests/15?tab=diff Signed-off-by: 晓雷 <meizhiyu.mzy@antgroup.com> * support specifying number of gpus and mems for actors * PR fix
Merge branch mzy/auto-eval of git@code.alipay.com:inclusionAI/AReaL.git into main https://code.alipay.com/inclusionAI/AReaL/pull_requests/11 Signed-off-by: 博惟 <bowei.fw@antgroup.com> * test * move evaluator to main process * . * clear codes * add docstring * . * separate wandb groups * .
Merge branch fw/fix-avgseqlen of git@code.alipay.com:inclusionAI/AReaL.git into main https://code.alipay.com/inclusionAI/AReaL/pull_requests/16 Signed-off-by: 温差 <xushusheng.xss@antgroup.com> * .
Merge branch ranghou-math of git@code.alipay.com:inclusionAI/AReaL.git into main https://code.alipay.com/inclusionAI/AReaL/pull_requests/18 Signed-off-by: 晓雷 <meizhiyu.mzy@antgroup.com> Signed-off-by: 博惟 <bowei.fw@antgroup.com> * support specifying number of gpus and mems for actors * PR fix * support tensorboard * bug fix * add doc
…support Merge branch fw/uvloop of git@code.alipay.com:inclusionAI/AReaL.git into main https://code.alipay.com/inclusionAI/AReaL/pull_requests/6 Signed-off-by: 晓雷 <meizhiyu.mzy@antgroup.com> * . * fw/fix-dataloading-not-shuffle * . * . * . * . * . * add v2 master worker * cpu test pass * ppo run * . * format * fix * merge and format * change default env vars to v1 worker
Merge branch functioncall-code of git@code.alipay.com:inclusionAI/AReaL.git into main https://code.alipay.com/inclusionAI/AReaL/pull_requests/17 Signed-off-by: 晓雷 <meizhiyu.mzy@antgroup.com> * test code evaluation with faas * support functioncall for code * fix code crash bug * format * .
Merge branch fw/datatransfer-v2 of git@code.alipay.com:inclusionAI/AReaL.git into main https://code.alipay.com/inclusionAI/AReaL/pull_requests/9 Signed-off-by: 晓雷 <meizhiyu.mzy@antgroup.com> * . * fw/fix-dataloading-not-shuffle * . * . * . * . * . * add v2 master worker * cpu test pass * ppo run * . * pass sft test * pass ppo dp test * format * fix * run * . * cleanup * . * format * run * merge and format * refactor * sft pass * . * format * format * format * . * .
Merge branch fix-tutorials of git@code.alipay.com:inclusionAI/AReaL.git into main https://code.alipay.com/inclusionAI/AReaL/pull_requests/19 Signed-off-by: 晓雷 <meizhiyu.mzy@antgroup.com>
Merge branch fw/fix-timeutil-recover of git@code.alipay.com:inclusionAI/AReaL.git into main https://code.alipay.com/inclusionAI/AReaL/pull_requests/21 Signed-off-by: 步偶 <sam.gjx@antgroup.com> * fix timeutil consistency during recover
Merge branch mzy/fix-slurm of git@code.alipay.com:inclusionAI/AReaL.git into main https://code.alipay.com/inclusionAI/AReaL/pull_requests/23 Signed-off-by: 博惟 <bowei.fw@antgroup.com> * fix slurm + singularity environment
Merge branch rw_save_version of git@code.alipay.com:inclusionAI/AReaL.git into main https://code.alipay.com/inclusionAI/AReaL/pull_requests/25 Signed-off-by: 博惟 <bowei.fw@antgroup.com> * fix the save version in rw interface * format
Merge branch fix/math of git@code.alipay.com:inclusionAI/AReaL.git into main https://code.alipay.com/inclusionAI/AReaL/pull_requests/26 Signed-off-by: 晓雷 <meizhiyu.mzy@antgroup.com>
zhshgmail
pushed a commit
to HsiaoTsan/AReaL
that referenced
this pull request
Oct 29, 2025
This commit addresses design issues areal-project#1-areal-project#4 by reorganizing interfaces: ## Changes **1. Moved to api module (proper location for interfaces):** - `QueueAPI` → `areal/api/queue_api.py` - `CacheAPI` → `areal/api/cache_api.py` - `EventType`, `EventContext`, `EventHandler` → `areal/api/event_api.py` - `QueueFilter` → `Filter` in `areal/api/filter_api.py` **2. Created separate queue/cache event handler protocols:** - `QueueEventHandler` + `QueueEventContext` in `areal/api/queue_event_handler.py` - `CacheEventHandler` + `CacheEventContext` in `areal/api/cache_event_handler.py` **3. Simplified event_system.py:** - Now only contains `EventRegistry` class - Imports interfaces from api module - No longer defines protocols **4. Renamed QueueFilter → Filter:** - Generic name since it applies to both queue and cache - Located in api module as proper interface - StalenessFilter now implements Filter protocol ## Why Protocol as superclass? (Answer to areal-project#4) Protocol enables **structural subtyping** (duck typing with types): ```python from typing import Protocol class QueueAPI(Protocol): def put(self, item): ... # No need to inherit! class MyQueue: def put(self, item): pass # Type checker accepts this ✓ def foo(q: QueueAPI): q.put(1) foo(MyQueue()) # Works! Structural typing ``` vs ABC (requires inheritance): ```python from abc import ABC class QueueAPI(ABC): ... class MyQueue: # Must inherit from QueueAPI ... ``` **Protocol = Interface without inheritance requirement** ## Remaining Work Still TODO (per user feedback): - Redesign ProximalRecomputer with propagator pattern (areal-project#3) - Fix AsyncTaskRunner to require QueueAPI/CacheAPI (areal-project#5) - Remove executor._filter_context (areal-project#6) - Add CacheFilter or apply Filter to all operations (areal-project#7, areal-project#8) - Fix RemoteInfEngine to not expose queue/cache (areal-project#9) - Merge event_factory into workflow_factory (areal-project#10) - Add vLLM recompute support (areal-project#11) 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude <noreply@anthropic.com>
zhshgmail
pushed a commit
to HsiaoTsan/AReaL
that referenced
this pull request
Oct 29, 2025
Addresses Issue areal-project#5 from user feedback: AsyncTaskRunner should not support both QueueAPI and plain queue.Queue. It should REQUIRE QueueAPI/CacheAPI. Changes: 1. Made output_queue and result_cache REQUIRED parameters (not optional) 2. Typed as QueueAPI and CacheAPI using TYPE_CHECKING 3. Removed default creation logic (no fallback to plain queue.Queue or list) 4. Updated event_factory to pass parameters in correct order 5. Added __future__ annotations for forward references This enforces the Spring @configuration pattern where the factory is responsible for creating and configuring queue/cache instances. AsyncTaskRunner no longer creates defaults - it MUST receive properly configured QueueAPI and CacheAPI implementations. Benefits: - Enforces proper factory pattern - No mixing of plain queue.Queue and QueueAPI - Clearer separation of concerns - Enables future distributed implementations 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude <noreply@anthropic.com>
guozhihao-224
added a commit
to guozhihao-224/AReaL
that referenced
this pull request
Apr 18, 2026
…, and metrics
Fix high-priority documentation issues that can directly cause user errors:
1. **Proxy endpoints** (agent_workflow.md): Fix paths from /v1/chat/completions
and /v1/responses to /chat/completions and /responses. The OpenAI SDK puts
/v1/ in the base_url, not in the request path.
2. **Rollout config** (rollout_workflow.md): Remove non-existent
rollout.group_size YAML config (group_size is a runtime API parameter).
Fix trajectory dump path to include /logs/{user}/ segment.
3. **LoRA support** (cli_reference.md, lora.md, cli_args.py): Update "Only
support FSDP" to include Megatron (requires megatron-bridge). Add note
that SGLang LoRA requires weight_update_mode: disk.
4. **Metrics tracking** (metrics_tracking.md): Fix SwanLab mode from "online"
(invalid) to "cloud". Add Trackio as a supported logging backend.
Ref: areal-project#1165 (Findings areal-project#3, areal-project#4, areal-project#5, areal-project#8, areal-project#9, areal-project#14, areal-project#15)
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
3 tasks
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
5 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.
No description provided.