Moonshot AI Researchers Introduce Seer: An On-line Context Studying System for Quick Synchronous Reinforcement Studying RL Rollouts
How do you retain reinforcement studying for big reasoning fashions from stalling on a number of very lengthy, very sluggish rollouts whereas GPUs sit beneath used? a group of researchers from Moonshot AI and Tsinghua College introduce ‘Seer’, a brand new on-line context studying system that targets a selected programs bottleneck in reinforcement studying for big language fashions. In synchronous on coverage setups, the rollout part dominates the price of every iteration. Seer restructures this part and experiences rollout throughput beneficial properties of 74 p.c to 97 p.c and tail latency reductions of 75 p.c to 93 p.c in contrast with a robust synchronous baseline known as veRL.
Why synchronous rollout is sluggish for reasoning fashions?
Fashionable reasoning RL workloads use lengthy chain of thought model outputs. Within the Seer experiments, the researchers apply GRPO to a few completely different fashions, Moonlight, Qwen2 VL 72B and Kimi K2. These workloads run on 32 compute nodes with 8 H800 GPUs per node. The three duties use 32, 128 and 256 GPUs respectively, with 400, 600 and 800 prompts per iteration and eight or 16 responses per immediate.
Most technology size is massive. Moonlight is configured for 65,536 tokens, Qwen2 VL 72B for 40,960 tokens and Kimi K2 for 98,304 tokens. A single lengthy chain of thought request can develop from a number of hundred megabytes of KVCache to tens of gigabytes as decoding progresses. This reminiscence development forces cases to cut back concurrency or to preempt requests, which triggers costly re decoding.
The analysis group defines tail requests because the final 10 p.c of requests to complete in a rollout. For Moonlight and Qwen2 VL 72B, this tail alone can eat as much as 50 p.c of the overall rollout time within the baseline system. Rollout already dominates iteration time, so this tail impact instantly slows RL.
Seer structure on high of Mooncake and vLLM
Seer retains the RL algorithm equivalent to synchronous veRL. Every coaching iteration makes use of solely knowledge from the present rollout iteration, so the system preserves on coverage habits. The coaching part makes use of Megatron for distributed optimization. The rollout part makes use of an in home implementation of vLLM because the inference engine.
To assist aggressive request scheduling, Seer depends on a World KVCache Pool constructed on the Mooncake disaggregated KVCache structure utilized in manufacturing for Kimi. Mooncake gives a two tier DRAM and SSD KV cache retailer shared throughout inference nodes, which permits Seer emigrate requests with out recomputing prefills.
On high of this substrate, Seer introduces three key mechanisms:
- Divided Rollout
- Context Conscious Scheduling
- Adaptive Grouped Speculative Decoding
These are orchestrated by a Request Buffer, a Context Supervisor and an Inference Engine Pool related to the World KVCache Pool.
Divided Rollout, high-quality grained scheduling and migration
Standard synchronous rollout assigns entire GRPO teams to inference cases. A bunch is a set of requests that share one immediate. As soon as assigned, a bunch stays on the identical occasion till all responses end. Because of massive variance in output lengths, this results in load imbalance and lengthy working stragglers.
Seer breaks teams down in two steps. It first decomposes every group into particular person requests. It then divides every request into a number of chunks based mostly on technology size. When the scheduler dispatches a request from the Request Buffer, it units a small max tokens worth comparable to 8,000 tokens for that chunk. After every chunk, the request is re enqueued till it reaches an finish of sequence token or its authentic max tokens restrict.
As a result of KVCache is saved within the World KVCache Pool, divided requests can transfer between cases at chunk boundaries with out re working the prefill. The scheduler maintains a concurrency stage that retains reminiscence utilization excessive whereas avoiding preemption. This reduces waste and smooths KVCache utilization throughout the iteration.
Context Conscious Scheduling utilizing group size statistics
The analysis group observe that completely different requests in the identical group are inclined to have correlated output lengths. Seer makes use of this construction as on-line context. For every immediate group, it designates one request because the speculative request. The scheduler retains speculative requests in a excessive precedence queue and serves them with a smallest first coverage based mostly on generated tokens thus far. Quick requests full rapidly and exit. Lengthy requests stay and establish teams which might be potential tail candidates.
The Context Supervisor maintains a size estimate for every group. It updates this estimate to the utmost generated size amongst accomplished requests within the group. If no request has completed, it makes use of the unique max tokens as a conservative sure. As soon as speculative requests are in flight or executed, Seer schedules remaining requests with an approximate longest first coverage at group stage. This design achieves throughput and tail habits near an oracle scheduler that is aware of all output lengths prematurely.
Adaptive Grouped Speculative Decoding
Seer provides Adaptive Grouped Speculative Decoding on high of the earlier two elements to speed up decoding, particularly for lengthy requests within the tail. It introduces a Distributed Grouped Draft Server, or DGDS. DGDS maintains a Compressed Suffix Tree for every group and aggregates token sequences from all requests in that group. Cases asynchronously append generated tokens to DGDS, periodically fetch up to date suffix bushes and carry out native speculative decoding based mostly on the shared sample statistics.
The system adjusts draft size and the variety of paths in line with mannequin structure, batch dimension and measured acceptance size. For dense and Combination of Specialists fashions, it pre-computes completely different hypothesis thresholds and makes use of them to sure draft depth for every batch. In late tail levels, concurrency is low, so Seer will increase draft depth and permits multi path drafting to lift accepted tokens per step.
Ablation outcomes present that divided rollout yields as much as 35 p.c throughput enchancment over the baseline. Including Context Conscious Scheduling will increase this to as much as 47 p.c over baseline. Enabling grouped speculative decoding raises the overall speedup to 77 p.c to 87 p.c over the baseline within the evaluated iteration.
Finish to finish affect on RL coaching
The analysis group consider Seer on three RL duties constructed on Moonlight, Qwen2 VL 72B and Kimi K2. They run 10 rollout iterations per job and measure output tokens per second and completion time for every rollout. Seer improves rollout throughput by 74 p.c to 97 p.c throughout these workloads relative to veRL with the identical RL algorithm and vLLM based mostly inference engine.
Tail latency is decreased by 75 p.c to 93 p.c. For reminiscence constrained duties, the baseline system spends as much as half of its time on the final 10 p.c of requests. Seer removes most of this tail by combining divided rollout, Context Conscious Scheduling and Adaptive Grouped Speculative Decoding on high of the Mooncake based mostly World KVCache Pool.
Key Takeaways
- Rollout bottleneck: Seer targets the rollout part of synchronous RL, which accounts for about 63% to 87% of iteration time and is dominated by lengthy tail requests and KV cache fragmentation.
- Three core mechanisms: Seer combines divided rollout, context conscious scheduling and adaptive grouped speculative decoding to take advantage of output size and sample similarity amongst GRPO responses that share a immediate.
- Nice grained scheduling on a worldwide KV cache: Requests are cut up into chunks and migrated throughout a Mooncake model World KVCache Pool, which preserves synchronous on coverage RL whereas retaining GPU reminiscence utilization excessive and lowering preemptions.
- On-line context for tail latency discount: Group stage size statistics from speculative requests drive context conscious scheduling that approximates an oracle longest first scheduler and sharply reduces the time spent on the final 10 p.c of requests.
- Measured finish to finish beneficial properties: On manufacturing grade RL workloads with Moonlight, Qwen2 VL 72B and Kimi K2, Seer improves rollout throughput by 74% to 97% and reduces lengthy tail latency by 75% to 93% relative to a state-of-the-art synchronous vLLM based mostly baseline.
Editorial Feedback
Seer is a vital programs contribution as a result of it optimizes the rollout part in synchronous RL with out altering the underlying GRPO algorithm, so it preserves on coverage ensures and reproducibility whereas fixing an actual infrastructure bottleneck. The mixture of divided rollout, context conscious scheduling and adaptive grouped speculative decoding affords a sensible template for different RL stacks that depend on lengthy chain of thought reasoning fashions and enormous KVCache footprints. General, Seer exhibits that on-line context studying on the programs stage is now as essential as mannequin structure for scaling reasoning RL effectively.
Try the Paper here. Be happy to take a look at our GitHub Page for Tutorials, Codes and Notebooks. Additionally, be happy to observe us on Twitter and don’t overlook to affix our 100k+ ML SubReddit and Subscribe to our Newsletter. Wait! are you on telegram? now you can join us on telegram as well.
The publish Moonshot AI Researchers Introduce Seer: An Online Context Learning System for Fast Synchronous Reinforcement Learning RL Rollouts appeared first on MarkTechPost.
