ST
stabilityai/stable-diffusion-3.5-large
text-to-imagediffusersother8.2B params · updated Oct 2024
Mainframe serving
Can a COBOL transaction program call this checkpoint, and under which transport? The model returns evidence; the host keeps the decision.
Export status
Exportable with care
Pipeline of several graphs; export per component
Recommended pattern
None
Not cleared
Latency class
—
ONNX artifact
Not measured
measured once the harness posts results
Screening note
Not cleared for the transaction core.
text-to-image has no transaction-core use case; sampling loops are Python control flow
No integration pattern recommended
- Multi-graph diffusion pipeline — each component exports separately
osai-screening/1.0.0 · verdict 282b62325ddd98da…
Serving controlskill switch offshadow onconfidence floor 0.85GET /serving-status
Promotion gates
| Gate | What it checks | State |
|---|---|---|
| Structural check | Graph loads; node inputs, outputs and shapes resolve | Not run |
| Numerical parity | Same inputs agree with the source framework within 1e-4 | Not run |
| Eight-bit quantisation | INT8 weights with parity re-run afterwards | Not run |
| Shadow cycle | Full cycle scored against live traffic with no influence on outcomes | Not run |
Interface contract
No contract for this checkpoint
Models not cleared for a pattern get no copybook — there is no transport to generate one for.
Standards this deployment inherits
Download ML-BOMS1Models ship as ONNX artifacts. No ML framework runtime is installed on z/OS.
S2The copybook is the interface contract — generated, versioned, never hand-edited on either side.
S3Latency class selects the pattern. Team preference and vendor enthusiasm do not.
S4Every inference is logged with the record state and model version that produced it.
S5PHI leaves the platform only tokenized, over mTLS, to an endpoint under an executed BAA.
S6A model influences payment only after a full shadow cycle and with a tested fallback path.