fix: memory auto-cap strategy for SSD MoE streaming + speculative decoding (Issue #72) by solderzzc · Pull Request #77 · SharpAI/SwiftLM

fix(ssd-stream): prevent inference-time swap explosion with --draft-model (#72 follow-up)

github-actions[bot]

committed

5390216

fix(ssd-stream): auto-cap draft tokens to 1 when --stream-experts + --draft-model (#72)

github-actions[bot]

committed

dfd0935

test(benchmark): add Test 10 — Issue #72 SSD + draft model RAM regression

github-actions[bot]

committed

7a14a67

ci: add ssd-draft-memory-guard job + vm_stat readings for Issue #72

github-actions[bot]

committed

3f6bad5

docs: document --stream-experts + --draft-model auto-cap strategy (Issue #72)

github-actions[bot]

committed

bb29e36

fix(ci): repair YAML corruption in ci.yml (retention-days merged with comment)

github-actions[bot]

committed

be8353f

ci: trigger run after YAML fix

github-actions[bot]

committed

c8b236d

fix(ci): use bash variable for PID in ssd-draft-memory-guard

github-actions[bot]

committed

58249c2

fix: address Copilot review feedback on PR #77

github-actions[bot]

committed

7b0bfd4

fix: allow custom model selection in benchmark test 10

github-actions[bot]

committed

8385350

Navigation Menu

Repository navigation

fix: memory auto-cap strategy for SSD MoE streaming + speculative decoding (Issue #72) - #77

fix: memory auto-cap strategy for SSD MoE streaming + speculative decoding (Issue #72)#77
solderzzc merged 10 commits into
mainfrom
fix/issue-72-draft-model-ssd-ram

Commits on Apr 23, 2026

fix(ssd-stream): prevent inference-time swap explosion with --draft-model (#72 follow-up)

fix(ssd-stream): auto-cap draft tokens to 1 when --stream-experts + --draft-model (#72)

test(benchmark): add Test 10 — Issue #72 SSD + draft model RAM regression

ci: add ssd-draft-memory-guard job + vm_stat readings for Issue #72

docs: document --stream-experts + --draft-model auto-cap strategy (Issue #72)

fix(ci): repair YAML corruption in ci.yml (retention-days merged with comment)

ci: trigger run after YAML fix

fix(ci): use bash variable for PID in ssd-draft-memory-guard

fix: address Copilot review feedback on PR #77

fix: allow custom model selection in benchmark test 10