[GH-ISSUE #112] can vllm + mooncake 1P1D support DeepSeek R1 with TP16? #59

New issue

Closed

opened 2026-10-29 21:08:34 +00:00 by bsakaguchi · 1 comment

bsakaguchi commented

2026-10-29 21:08:34 +00:00

Owner

Originally created by @neiltian-tencent on GitHub (Feb 18, 2025).
Original GitHub issue: https://github.com/kvcache-ai/Mooncake/issues/112

P node needs 2 machines with 16 cards, D node needs 2 machines with 16 cards

Originally created by @neiltian-tencent on GitHub (Feb 18, 2025). Original GitHub issue: https://github.com/kvcache-ai/Mooncake/issues/112 P node needs 2 machines with 16 cards, D node needs 2 machines with 16 cards

bsakaguchi closed this issue

2026-10-29 21:08:34 +00:00

bsakaguchi commented

2026-10-29 21:08:35 +00:00

Author

Owner

@neiltian-tencent commented on GitHub (Feb 19, 2025):

Configuring different prefill_url and decoding_url on different nodes solves this problem

@neiltian-tencent commented on GitHub (Feb 19, 2025): Configuring different prefill_url and decoding_url on different nodes solves this problem

bsakaguchi referenced this issue

2026-10-29 21:26:11 +00:00

[PR #59] [CLOSED] Bump golang.org/x/crypto from 0.23.0 to 0.31.0 in /mooncake-transfer-engine/example/http-metadata-server #1303