stoa
RL post-training with your Apple-silicon Macs as the rollout fleet: Mac workers generate experience, GPUs learn, synced over Cloudflare R2
파일 탐색기
- SKILL.md
- SKILL.md
- ci.yml
- how-it-works.png
- how-it-works.svg
- how_it_works_fig.py
- stoa-banner.svg
- ab_kvopt.py
- analyze_ab.py
- README.md
- run_ab.sh
- async_uploader.py
- logging_setup.py
- pulse_chain.py
- pulse_delta.py
- pulse_quant.py
- pulse_sync.py
- r2_commit.py
- r2_io.py
- r2_rollout_log.py
- r2_stats_log.py
- filtered_training_set.jsonl
- README.md
- architecture.md
- async-learner-design.md
- r2-setup.md
- testing.md
- README.md
- run_qwen25_gsm8k_decoupled.sh
- MODEL_ARGS.sh
- provision.sh
- trainer_setup_decoupled.sh
- ckpt.py
- mbridge__lfm2_moe.py
- megatron_to_hf__lfm2_moe.py
- models__lfm2_moe.py
- NOTICE
- test_plugin_load.py
- eval_serve_sglang.sh
- generate_with_search_lfm2.py
- launch_fleet.env.example
- preflight_check.py
- prep_dataset.py
- prep_eval.py
- prep_worker_dataset.py
- psqa_env.py
- qa_em.py
- README.md
- resume_from_full.sh
- run_lfm25_psqa_fleet.sh
- patch_mlx_lfm2_moe_padslab.py
- patch_mlx_lfm2_moe_router.py
- retrieval_server.py
- clean_node.sh
- detect_mac_tier.sh
- install_supervisor.sh
- install_worker_key.sh
- join_and_run.sh
- mac_fleet.hosts.example
- provision_mac_fleet.sh
- run.env.example
- run.env.fleet-16gb.example
- run.env.fleet-24gb.example
- run.env.fleet.example
- setup_mac_worker.sh
- setup_retriever.sh
- setup_retriever_linux.sh
- apply_slime_patches.py
- heldout_full5000_n8.csv
- REPRODUCE.md
- build_pubmed_bm25.py
- fetch_data.sh
- hf_pull.py
- README.md
- lfm25_8bit_tensors.json
- lfm25_bf16_tensors.json
- pulse_bf16_plane_golden.json
- pulse_quant_golden.json
- _fake_s3.py
- _pulse_test_ids.py
- test_producer_consumer.py
- test_pulse_chain.py
- test_pulse_edge.py
- test_pulse_failclosed.py
- test_pulse_gaps.py
- test_pulse_plane.py
- test_pulse_quant.py
- test_pulse_realr2.py
- test_pulse_realr2_correctpath.py
- test_pulse_roundtrip.py
- test_pulse_scenarios.py
- test_pulse_sparsity.py
- test_pulse_taxonomy.py
- test_pulse_v2_protocol.py
- test_pulse_wedge.py
- test_r2_commit.py
- test_r2_io.py
- test_r2_io_hardening.py
- test_r2_rollout_log.py
- test_preflight_profiles.py
- test_psqa_env.py
- test_psqa_realtok.py
- test_resume_from_full.py
- test_reward_cover_em.py
- test_padslab_plan.py
- test_apply_slime_patches.py
- test_adaptive_fetch.py
- test_ckpt_export.py
- test_cursor_handoff.py
- test_drgrpo_reducer.py
- test_full_ckpt.py
- test_math_rm_grading.py
- test_math_rm_threading.py
- test_publish_async.py
- test_pulse_stats.py
- test_replay_buffer.py
- test_rollout_metrics.py
- test_cb_native.py
- test_cb_scheduler.py
- test_class_ds.py
- test_dynamic_sampling.py
- test_multigroup.py
- test_native_kernels.py
- test_paged_cache.py
- test_paged_fullscale.py
- test_paged_ledger_safety.py
- test_paged_race.py
- test_paged_refcount_lock.py
- test_paged_table_guards.py
- test_plane_native_load.py
- test_prefix_store.py
- test_psqa_streaming_phase2.py
- test_quant_gap.py
- test_search_floor.py
- test_stop_sequences.py
- test_stop_signal.py
- test_streaming_e2e_mlx.py
- test_streaming_phase2.py
- test_streaming_upload.py
- test_weight_prefetch.py
- test_worker_config.py
- test_worker_curriculum.py
- test_worker_identity.py
- test_worker_stats.py
- conftest.py
- mlx-lm-LICENSE
- PaperSearchQA-LICENSE
- Search-R1-LICENSE
- slime-LICENSE
- verl-LICENSE
- ckpt_export.py
- dppo_gate.py
- drgrpo_reducer.py
- eval_curve.py
- eval_gsm8k.py
- eval_pass_k.py
- full_ckpt.py
- gsm8k_rm.py
- inspect_rollouts.py
- learning_curve.py
- live_anchor_eval.py
- math_rm.py
- prep_gsm8k.py
- prep_math.py
- psqa_anchor_watcher.py
- psqa_eval_shakedown.py
- r2_rollout.py
- replay_buffer.py
- wordle_advantage.py
- canary_oracle.py
- cb_scheduler.py
- dRL_producer.py
- ds_gate.py
- native_kernels.py
- paged_cache.py
- paged_pool.py
- prefix_store.py
- producer_decode.py
- producer_weights.py
- psqa_mlx_smoke.py
- psqa_mlx_smoke_streaming.py
- psqa_retriever.py
- psqa_rollout.py
- requirements.txt
- run_b0_battery.sh
- run_dRL_worker_gsm8k.sh
- run_dRL_worker_psqa.sh
- run_local_e2e.py
- stop_sequences.py
- sweep_telemetry.py
- turn_mask.py
- worker_config.py
- worker_curriculum.py
- worker_stats.py
- .env.example
- .gitignore
- .pre-commit-config.yaml
- .python-version
- .r2env.example
- CITATION.cff
- CLAUDE.md
- CONTRIBUTING.md
- LICENSE
- NOTICE
- pyproject.toml
- README.md
- SECURITY.md
- THIRD_PARTY_NOTICES.md
- uv.lock
# CDN으로 사용하기
jsDelivrjsDelivr는 공개 GitHub 리포지토리를 별도 설정 없이 CDN으로 즉시 서빙합니다. 버전과 파일을 고르면 웹페이지에 바로 붙일 수 있는 링크와 예시 코드가 만들어집니다.
링크
예시
// repository documentation
Was this content helpful?
(0 ratings)