Open Models Almanac
Methodssoftware implementationinfrastructure service

methods · software implementation

infrastructure service

33 methods filed at this node or below it, from the sources of 18 models.

In this branch 33

Everything filed at this node or below it, with one collapsible heading per child node.

filed here 33

Automatic provider failover and retry core · 7 sources · 7 quotes
32 single mentions
Core scheduling by priority class core · 1 source · 1 quote
End-to-end lineage tracking core · 1 source · 1 quote
Parallel preprocessing with Ray core · 1 source · 1 quote
Sandbox pause and resume core · 1 source · 1 quote
Allowlisted sandbox egress proxy used · 1 source · 1 quote
Apptainer image conversion and caching used · 1 source · 1 quote
Central scheduler used · 1 source · 1 quote
DeepEP v2 with ElasticBuffer used · 1 source · 1 quote
Dependency version unification used · 1 source · 1 quote
Fresh-clone repository sanitization used · 1 source · 1 quote
HTTP/HTTPS sidecar proxy used · 1 source · 1 quote
In-house batch scheduler used · 1 source · 1 quote
Local container-image caching used · 1 source · 1 quote
Model Express used · 1 source · 1 quote
Model Factory used · 1 source · 1 quote
Mooncake used · 1 source · 1 quote
NeMo Guardrails used · 1 source · 1 quote
NeMo Switchyard used · 1 source · 1 quote
NIXL toy proxy used · 1 source · 1 quote
NVIDIA Dynamo used · 1 source · 1 quote
Per-node pooling of initialization actors used · 1 source · 1 quote
Ray data plane used · 1 source · 1 quote
Shared read-only SquashFS root used · 1 source · 1 quote
Vendoring external dependencies used · 1 source · 1 quote
Write-ahead log (WAL) used · 1 source · 1 quote
Z3 SMT solver integration used · 1 source · 1 quote
Ray optional · 1 source · 1 quote

By model

Which of this branch's methods each model's own documents describe, and how strongly. Under each model: its strongest treatment anywhere in the branch.

Modelmethods
GLM-5.3-Flash usedAutomatic provider failover and retry used—
DeepSeek-V4.1-Flash coreCore scheduling by priority class coreRelaxed-consistency sandbox placement engine coreSharding and relaxed-consistency scheduling core—
Step-5-Preview usedAutomatic provider failover and retry used—
NVIDIA-Nemotron-3-Ultra-550B-A55B usedDependency version unification usedFresh-clone repository sanitization usedLocal container-image caching usedPer-node pooling of initialization actors usedReplacing short-lived Ray actors with tasks usedRay optional—
DeepSeek-V4-Flash usedWrite-ahead log (WAL) usedZ3 SMT solver integration used—
DeepSeek-V4-Pro usedWrite-ahead log (WAL) usedZ3 SMT solver integration used—
DeepSeek-V4-Pro-0813 optionalAutomatic provider failover and retry optional—
Gemma 4 31B coreAutomatic provider failover and retry core—
Inkling usedAutomatic provider failover and retry used—
Kimi K3 coreIncremental sandbox checkpointing and resumption coreSandbox pause and resume coreMooncake used—
Kolibri-1 usedAllowlisted sandbox egress proxy usedApptainer image conversion and caching usedDeepEP v2 with ElasticBuffer usedPassive GPU health monitoring with NVIDIA DCGM usedShared read-only SquashFS root used—
Laguna-S-2.1 coreEnd-to-end lineage tracking coreHTTP/HTTPS sidecar proxy usedIn-house batch scheduler usedModel Factory usedVendoring external dependencies used—
Mistral-Large-4 usedCentral scheduler usedModel Express usedNIXL toy proxy usedRay data plane used—
NVIDIA-Nemotron-3.5-Lightning-30B-A3B usedNeMo Guardrails usedNeMo Switchyard used—
NVIDIA-Nemotron-3.5-Super-VL-120B-A12B usedNVIDIA Dynamo used—
Qwen3.5-397B-A17B usedAutomatic provider failover and retry used—
Qwen3.6-35B-A3B coreAutomatic provider failover and retry core—
VLA Foundry (Foundry-VLA-1.7B) coreParallel preprocessing with Ray core—
↑