{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/comet-fine-grained-computation-communication","title":"Comet: Fine-grained Computation-communication Overlapping for Mixture-of-Experts","arxiv_id":"2502.19811","date":"2025-02-27","proceeding":null,"authors":["Shulai Zhang","Ningxin Zheng","Haibin Lin","Ziheng Jiang","Wenlei Bao","Chengquan Jiang","Qi Hou","Weihao Cui","Size Zheng","Li-Wen Chang","Quan Chen","Xin Liu"],"abstract":"Mixture-of-experts (MoE) has been extensively employed to scale large language models to trillion-plus parameters while maintaining a fixed computational cost. The development of large MoE models in the distributed scenario encounters the problem of large communication overhead. The inter-device communication of a MoE layer can occupy 47% time of the entire model execution with popular models and frameworks. Therefore, existing methods suggest the communication in a MoE layer to be pipelined with the computation for overlapping. However, these coarse grained overlapping schemes introduce a notable impairment of computational efficiency and the latency concealing is sub-optimal. To this end, we present COMET, an optimized MoE system with fine-grained communication-computation overlapping. Leveraging data dependency analysis and task rescheduling, COMET achieves precise fine-grained overlapping of communication and computation. Through adaptive workload assignment, COMET effectively eliminates fine-grained communication bottlenecks and enhances its adaptability across various scenarios. Our evaluation shows that COMET accelerates the execution of a single MoE layer by $1.96\\times$ and for end-to-end execution, COMET delivers a $1.71\\times$ speedup on average. COMET has been adopted in the production environment of clusters with ten-thousand-scale of GPUs, achieving savings of millions of GPU hours.","url_abs":"https://arxiv.org/abs/2502.19811v3","url_pdf":"https://arxiv.org/pdf/2502.19811v3.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"comet-fine-grained-computation-communication","repo_url":"https://github.com/bytedance/flux","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"Apache-2.0"}},{"paper_slug":"comet-fine-grained-computation-communication","repo_url":"https://github.com/multi-swe-bench/magentless","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":{"status":"ok","spdx":"MIT"}},{"paper_slug":"comet-fine-grained-computation-communication","repo_url":"https://github.com/multi-swe-bench/mopenhands","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":{"status":"ok","spdx":"MIT"}},{"paper_slug":"comet-fine-grained-computation-communication","repo_url":"https://github.com/multi-swe-bench/mswe-agent","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":{"status":"ok"}}],"tasks":[{"task_slug":"computational-efficiency","task_name":"Computational Efficiency"},{"task_slug":null,"task_name":"GPU"},{"task_slug":"mixture-of-experts","task_name":"Mixture-of-Experts"}],"methods":[{"method_slug":"moe","method_name":"MoE"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":"https://syntology.ai/paper/2502.19811","atlas_url":"https://app.syntology.ai/?focus=2502.19811","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2502.19811"}},"developers":"https://syntology.ai/developers","read_at":"2026-09-25T09:33:49+00:00","read_at_is":"when the build read Syntology's graph, not when any sample ran","claim":"Per-sample execution status on synthesized fixtures; not a correctness claim about the paper. Samples come from repositories linked to the paper, official or community; repo_kind says which.","repos":[{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/multi-swe-bench/mopenhands","reach":{"status":"ok","spdx":"MIT"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/multi-swe-bench/mswe-agent","reach":{"status":"ok"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/multi-swe-bench/magentless","reach":{"status":"ok","spdx":"MIT"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/bytedance/flux","reach":{"status":"ok","spdx":"Apache-2.0"}}],"summary":{"ran":7,"unverified":13},"by_repo_kind":{"official":{"samples":17,"ran":6,"repositories":1},"listed":{"samples":3,"ran":1,"repositories":1}},"repo_kind_vocabulary":{"official":"The archive marks this repository official for the paper","named_in_paper":"The archive records that the paper mentions this repository; it is not marked official","listed":"In the archive's code links for this paper, not marked official and not recorded as mentioned in the paper","found_in_text":"Syntology found this repository in the paper's own text; whether it is the authors' implementation is not asserted","community":"Not in the archive's code links for this paper; a community repository Syntology harvested"},"n_pointer_only_for_licence":0,"samples":[{"code_sha256_prefix":"c252afed940d6e30","entry":"calc_scatter_index_stable","repo":"bytedance/flux","repo_kind":"official","path":"python/flux/testing/moe_utils.py","file_url":"https://github.com/bytedance/flux/blob/HEAD/python/flux/testing/moe_utils.py","link_basis":"harvester_set","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":true,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"c252afed940d6e30"}},{"code_sha256_prefix":"4dc6308a692bee4f","entry":"construct_file_meta_data","repo":"multi-swe-bench/magentless","repo_kind":"listed","path":"agentless/fl/Index.py","file_url":"https://github.com/multi-swe-bench/magentless/blob/HEAD/agentless/fl/Index.py","link_basis":"harvester_set","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"4dc6308a692bee4f"}},{"code_sha256_prefix":"3e48284521d53051","entry":"get_inter_node_ag_group","repo":"bytedance/flux","repo_kind":"official","path":"python/flux/ag_kernel_internode.py","file_url":"https://github.com/bytedance/flux/blob/HEAD/python/flux/ag_kernel_internode.py","link_basis":"harvester_set","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"3e48284521d53051"}},{"code_sha256_prefix":"318a824a192ac02f","entry":"get_torch_prof_ctx","repo":"bytedance/flux","repo_kind":"official","path":"python/flux/util.py","file_url":"https://github.com/bytedance/flux/blob/HEAD/python/flux/util.py","link_basis":"harvester_set","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"318a824a192ac02f"}},{"code_sha256_prefix":"f47e5cdd7171b461","entry":"is_fp8_dtype","repo":"bytedance/flux","repo_kind":"official","path":"python/flux/testing/gpu_perf_model.py","file_url":"https://github.com/bytedance/flux/blob/HEAD/python/flux/testing/gpu_perf_model.py","link_basis":"harvester_set","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"f47e5cdd7171b461"}},{"code_sha256_prefix":"6469bfc20e491271","entry":"is_fp8_dtype","repo":"bytedance/flux","repo_kind":"official","path":"python/flux/util.py","file_url":"https://github.com/bytedance/flux/blob/HEAD/python/flux/util.py","link_basis":"harvester_set","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"6469bfc20e491271"}},{"code_sha256_prefix":"36283518ba9f282d","entry":"torch_pre_attn_all_to_all_transpose","repo":"bytedance/flux","repo_kind":"official","path":"python/flux/testing/ulysses_sp_utils.py","file_url":"https://github.com/bytedance/flux/blob/HEAD/python/flux/testing/ulysses_sp_utils.py","link_basis":"harvester_set","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"36283518ba9f282d"}},{"code_sha256_prefix":"b2e2521687ac95ce","entry":"bench_func","repo":"bytedance/flux","repo_kind":"official","path":"python/flux/util.py","file_url":"https://github.com/bytedance/flux/blob/HEAD/python/flux/util.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"b2e2521687ac95ce"}},{"code_sha256_prefix":"b0bd0e05915bf19c","entry":"calc_gather_index","repo":"bytedance/flux","repo_kind":"official","path":"python/flux/testing/moe_utils.py","file_url":"https://github.com/bytedance/flux/blob/HEAD/python/flux/testing/moe_utils.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"b0bd0e05915bf19c"}},{"code_sha256_prefix":"c7158b2a6c069b16","entry":"calc_gather_index_stable","repo":"bytedance/flux","repo_kind":"official","path":"python/flux/testing/moe_utils.py","file_url":"https://github.com/bytedance/flux/blob/HEAD/python/flux/testing/moe_utils.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"c7158b2a6c069b16"}},{"code_sha256_prefix":"bdad45577321a1d2","entry":"get_device_property","repo":"bytedance/flux","repo_kind":"official","path":"python/flux/testing/gpu_perf_model.py","file_url":"https://github.com/bytedance/flux/blob/HEAD/python/flux/testing/gpu_perf_model.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"bdad45577321a1d2"}},{"code_sha256_prefix":"c23cadd88b362499","entry":"get_dist_env","repo":"bytedance/flux","repo_kind":"official","path":"python/flux/dist_utils.py","file_url":"https://github.com/bytedance/flux/blob/HEAD/python/flux/dist_utils.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"c23cadd88b362499"}},{"code_sha256_prefix":"80683aed2fa79e07","entry":"get_intra_node_ag_group","repo":"bytedance/flux","repo_kind":"official","path":"python/flux/ag_kernel_internode.py","file_url":"https://github.com/bytedance/flux/blob/HEAD/python/flux/ag_kernel_internode.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"80683aed2fa79e07"}},{"code_sha256_prefix":"673d58d820da47c9","entry":"get_intra_node_pg_group","repo":"bytedance/flux","repo_kind":"official","path":"python/flux/gemm_rs_sm80.py","file_url":"https://github.com/bytedance/flux/blob/HEAD/python/flux/gemm_rs_sm80.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"673d58d820da47c9"}},{"code_sha256_prefix":"0c0fc4a132ede3cb","entry":"nvsmi","repo":"bytedance/flux","repo_kind":"official","path":"python/flux/testing/gpu_perf_model.py","file_url":"https://github.com/bytedance/flux/blob/HEAD/python/flux/testing/gpu_perf_model.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"0c0fc4a132ede3cb"}},{"code_sha256_prefix":"f2c4e9391d254c7e","entry":"parse_patch","repo":"multi-swe-bench/magentless","repo_kind":"listed","path":"get_repo_structure/get_patch_info.py","file_url":"https://github.com/multi-swe-bench/magentless/blob/HEAD/get_repo_structure/get_patch_info.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"f2c4e9391d254c7e"}},{"code_sha256_prefix":"94dc3cbe589f0afa","entry":"parse_python_file","repo":"multi-swe-bench/magentless","repo_kind":"listed","path":"get_repo_structure/get_repo_structure.py","file_url":"https://github.com/multi-swe-bench/magentless/blob/HEAD/get_repo_structure/get_repo_structure.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"94dc3cbe589f0afa"}},{"code_sha256_prefix":"146ee989a1ad16cb","entry":"run_perf","repo":"bytedance/flux","repo_kind":"official","path":"python/flux/testing/utils.py","file_url":"https://github.com/bytedance/flux/blob/HEAD/python/flux/testing/utils.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"146ee989a1ad16cb"}},{"code_sha256_prefix":"24ef3f53f604fc80","entry":"torch_post_attn_all_to_all_transpose","repo":"bytedance/flux","repo_kind":"official","path":"python/flux/testing/ulysses_sp_utils.py","file_url":"https://github.com/bytedance/flux/blob/HEAD/python/flux/testing/ulysses_sp_utils.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"24ef3f53f604fc80"}},{"code_sha256_prefix":"e693c5491138e1e3","entry":"torch_pre_attn_qkv_pack_a2a","repo":"bytedance/flux","repo_kind":"official","path":"python/flux/testing/ulysses_sp_utils.py","file_url":"https://github.com/bytedance/flux/blob/HEAD/python/flux/testing/ulysses_sp_utils.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"e693c5491138e1e3"}}]},"arxiv_metadata":null,"syntology_extracted_results":null}