AI Architecture & Systems
1
Microarchitecture
2
Kernel
3
Compiler
4
Architecture
5
System
6
Algorithms
System
/
§5.31
RL Frameworks and Source-Code Studies
5.31.1
verl: Initialization, Rollout, and Training Workflow
#
5.31.2
slime: Rollout-First Design and Backend Integration
#
5.31.3
OpenRLHF and Ray-Based Actor Management
#
5.31.4
AReaL and Asynchronous RL Training
#
5.31.5
TRL and Lightweight Post-Training Workflows
#
5.31.6
Weight Update, Memory Sleep/Wake, and Resource Reuse
#
5.31.7
FP8/INT4, Speculative Decoding, and Multi-Turn Tasks in RL
#
← Previous
5.30 Training–Inference Consistency and RL Correctness
→ Next
5.32 Multimodal and Real-Time Speech Serving