@atomic-chat/core 0.5.1 → 0.13.6
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/README.md +22 -0
- package/dist/app-daemon.js +20 -67
- package/dist/app-daemon.js.map +1 -1
- package/dist/backend/advisor/advisor.d.ts +132 -0
- package/dist/backend/advisor/advisor.js +537 -0
- package/dist/backend/advisor/advisor.js.map +1 -0
- package/dist/backend/advisor/index.d.ts +6 -0
- package/dist/backend/advisor/index.js +7 -0
- package/dist/backend/advisor/index.js.map +1 -0
- package/dist/backend/advisor/policy.d.ts +81 -0
- package/dist/backend/advisor/policy.js +154 -0
- package/dist/backend/advisor/policy.js.map +1 -0
- package/dist/backend/catalog/archive.d.ts +21 -7
- package/dist/backend/catalog/archive.js +57 -17
- package/dist/backend/catalog/archive.js.map +1 -1
- package/dist/backend/catalog/bundled-manifest-baseline.js +42 -26
- package/dist/backend/catalog/bundled-manifest-baseline.js.map +1 -1
- package/dist/backend/catalog/cuda-family.d.ts +12 -2
- package/dist/backend/catalog/cuda-family.js +30 -7
- package/dist/backend/catalog/cuda-family.js.map +1 -1
- package/dist/backend/catalog/index.d.ts +5 -0
- package/dist/backend/catalog/index.js +5 -0
- package/dist/backend/catalog/index.js.map +1 -1
- package/dist/backend/catalog/manifest.d.ts +5 -3
- package/dist/backend/catalog/manifest.js +12 -9
- package/dist/backend/catalog/manifest.js.map +1 -1
- package/dist/backend/catalog/migrate.d.ts +8 -6
- package/dist/backend/catalog/migrate.js +26 -15
- package/dist/backend/catalog/migrate.js.map +1 -1
- package/dist/backend/catalog/prism-catalog.d.ts +42 -0
- package/dist/backend/catalog/prism-catalog.js +101 -0
- package/dist/backend/catalog/prism-catalog.js.map +1 -0
- package/dist/backend/catalog/prism-manifest-baseline.d.ts +2 -0
- package/dist/backend/catalog/prism-manifest-baseline.js +11 -0
- package/dist/backend/catalog/prism-manifest-baseline.js.map +1 -0
- package/dist/backend/catalog/prism-manifest-baseline.json +134 -0
- package/dist/backend/catalog/prism-manifest.d.ts +90 -0
- package/dist/backend/catalog/prism-manifest.js +171 -0
- package/dist/backend/catalog/prism-manifest.js.map +1 -0
- package/dist/backend/catalog/turboquant-catalog.d.ts +60 -0
- package/dist/backend/catalog/turboquant-catalog.js +162 -0
- package/dist/backend/catalog/turboquant-catalog.js.map +1 -0
- package/dist/backend/catalog/turboquant-index.d.ts +104 -0
- package/dist/backend/catalog/turboquant-index.js +260 -0
- package/dist/backend/catalog/turboquant-index.js.map +1 -0
- package/dist/backend/index.d.ts +1 -0
- package/dist/backend/index.js +1 -0
- package/dist/backend/index.js.map +1 -1
- package/dist/backend/install/runtime-backend.d.ts +6 -6
- package/dist/backend/install/runtime-backend.js +23 -20
- package/dist/backend/install/runtime-backend.js.map +1 -1
- package/dist/backend/install/service.d.ts +106 -5
- package/dist/backend/install/service.js +247 -21
- package/dist/backend/install/service.js.map +1 -1
- package/dist/backend/installed/bundled.d.ts +19 -0
- package/dist/backend/installed/bundled.js +45 -0
- package/dist/backend/installed/bundled.js.map +1 -0
- package/dist/backend/installed/index.d.ts +1 -0
- package/dist/backend/installed/index.js +1 -0
- package/dist/backend/installed/index.js.map +1 -1
- package/dist/backend/installed/installed.js +3 -0
- package/dist/backend/installed/installed.js.map +1 -1
- package/dist/backend/optimal/optimal-cache.d.ts +27 -8
- package/dist/backend/optimal/optimal-cache.js +38 -19
- package/dist/backend/optimal/optimal-cache.js.map +1 -1
- package/dist/backend/optimal/optimal-store.js +1 -1
- package/dist/backend/optimal/optimal-store.js.map +1 -1
- package/dist/backend/select/categories.d.ts +2 -1
- package/dist/backend/select/categories.js +8 -1
- package/dist/backend/select/categories.js.map +1 -1
- package/dist/backend/select/features.d.ts +13 -5
- package/dist/backend/select/features.js +36 -6
- package/dist/backend/select/features.js.map +1 -1
- package/dist/backend/select/index.d.ts +2 -0
- package/dist/backend/select/index.js +2 -0
- package/dist/backend/select/index.js.map +1 -1
- package/dist/backend/select/prism.d.ts +54 -0
- package/dist/backend/select/prism.js +212 -0
- package/dist/backend/select/prism.js.map +1 -0
- package/dist/backend/select/recovery.d.ts +4 -2
- package/dist/backend/select/recovery.js +17 -5
- package/dist/backend/select/recovery.js.map +1 -1
- package/dist/backend/select/tiers.d.ts +20 -11
- package/dist/backend/select/tiers.js +49 -11
- package/dist/backend/select/tiers.js.map +1 -1
- package/dist/backend/select/turboquant-tiers.d.ts +45 -0
- package/dist/backend/select/turboquant-tiers.js +114 -0
- package/dist/backend/select/turboquant-tiers.js.map +1 -0
- package/dist/backend/types.d.ts +6 -69
- package/dist/cli/commands/backends.d.ts +10 -0
- package/dist/cli/commands/backends.js +142 -0
- package/dist/cli/commands/backends.js.map +1 -0
- package/dist/cli/commands/hardware.d.ts +13 -0
- package/dist/cli/commands/hardware.js +71 -0
- package/dist/cli/commands/hardware.js.map +1 -0
- package/dist/cli/commands/host-step.d.ts +28 -0
- package/dist/cli/commands/host-step.js +67 -0
- package/dist/cli/commands/host-step.js.map +1 -0
- package/dist/cli/commands/index.d.ts +4 -1
- package/dist/cli/commands/index.js +4 -1
- package/dist/cli/commands/index.js.map +1 -1
- package/dist/cli/commands/serve.js +4 -4
- package/dist/cli/commands/serve.js.map +1 -1
- package/dist/cli/commands/shared.d.ts +6 -1
- package/dist/cli/commands/shared.js +7 -0
- package/dist/cli/commands/shared.js.map +1 -1
- package/dist/cli/io.d.ts +11 -0
- package/dist/cli/io.js +28 -3
- package/dist/cli/io.js.map +1 -1
- package/dist/cli/launch.js +2 -2
- package/dist/cli/launch.js.map +1 -1
- package/dist/cli/main.d.ts +1 -1
- package/dist/cli/main.js +11 -1
- package/dist/cli/main.js.map +1 -1
- package/dist/client/control-client.d.ts +128 -7
- package/dist/client/control-client.js +211 -12
- package/dist/client/control-client.js.map +1 -1
- package/dist/cloud/registry.js +9 -1
- package/dist/cloud/registry.js.map +1 -1
- package/dist/config/paths.d.ts +137 -0
- package/dist/config/paths.js +163 -0
- package/dist/config/paths.js.map +1 -1
- package/dist/contracts/backend-advisor.d.ts +225 -0
- package/dist/contracts/backend-advisor.js +14 -0
- package/dist/contracts/backend-advisor.js.map +1 -0
- package/dist/contracts/control-api.d.ts +7 -1
- package/dist/contracts/control-api.js +7 -1
- package/dist/contracts/control-api.js.map +1 -1
- package/dist/contracts/decision.d.ts +394 -0
- package/dist/contracts/decision.js +69 -0
- package/dist/contracts/decision.js.map +1 -0
- package/dist/contracts/diffusion.d.ts +62 -9
- package/dist/contracts/embedding.d.ts +103 -0
- package/dist/contracts/embedding.js +24 -0
- package/dist/contracts/embedding.js.map +1 -0
- package/dist/contracts/engine-builds.d.ts +108 -0
- package/dist/contracts/engine-builds.js +11 -0
- package/dist/contracts/engine-builds.js.map +1 -0
- package/dist/contracts/engines.d.ts +157 -0
- package/dist/contracts/engines.js +50 -0
- package/dist/contracts/engines.js.map +1 -0
- package/dist/contracts/environment.d.ts +774 -0
- package/dist/contracts/environment.js +116 -0
- package/dist/contracts/environment.js.map +1 -0
- package/dist/contracts/errors.d.ts +43 -2
- package/dist/contracts/errors.js +41 -0
- package/dist/contracts/errors.js.map +1 -1
- package/dist/contracts/events.d.ts +82 -4
- package/dist/contracts/hardware.d.ts +79 -6
- package/dist/contracts/hardware.js +9 -5
- package/dist/contracts/hardware.js.map +1 -1
- package/dist/contracts/index.d.ts +8 -0
- package/dist/contracts/index.js +8 -0
- package/dist/contracts/index.js.map +1 -1
- package/dist/contracts/model-compatibility.d.ts +97 -0
- package/dist/contracts/model-compatibility.js +6 -0
- package/dist/contracts/model-compatibility.js.map +1 -0
- package/dist/contracts/model-setup.d.ts +115 -0
- package/dist/contracts/model-setup.js +33 -0
- package/dist/contracts/model-setup.js.map +1 -0
- package/dist/contracts/model-yml.d.ts +14 -0
- package/dist/contracts/model-yml.js +1 -0
- package/dist/contracts/model-yml.js.map +1 -1
- package/dist/contracts/session.d.ts +45 -2
- package/dist/contracts/session.js +14 -1
- package/dist/contracts/session.js.map +1 -1
- package/dist/contracts/settings.d.ts +44 -1
- package/dist/contracts/settings.js.map +1 -1
- package/dist/core/atomic-core.d.ts +46 -5
- package/dist/core/atomic-core.js +25 -0
- package/dist/core/atomic-core.js.map +1 -1
- package/dist/core/create.js +523 -23
- package/dist/core/create.js.map +1 -1
- package/dist/core/gpu/index.d.ts +6 -0
- package/dist/core/gpu/index.js +7 -0
- package/dist/core/gpu/index.js.map +1 -0
- package/dist/core/gpu/policy.d.ts +68 -0
- package/dist/core/gpu/policy.js +99 -0
- package/dist/core/gpu/policy.js.map +1 -0
- package/dist/core/gpu/residency.d.ts +44 -0
- package/dist/core/gpu/residency.js +101 -0
- package/dist/core/gpu/residency.js.map +1 -0
- package/dist/core/gpu-residency.d.ts +39 -0
- package/dist/core/gpu-residency.js +32 -0
- package/dist/core/gpu-residency.js.map +1 -0
- package/dist/core/index.d.ts +1 -1
- package/dist/core/managed-engines.d.ts +207 -0
- package/dist/core/managed-engines.js +579 -0
- package/dist/core/managed-engines.js.map +1 -0
- package/dist/core/managed-environment.d.ts +72 -0
- package/dist/core/managed-environment.js +218 -0
- package/dist/core/managed-environment.js.map +1 -0
- package/dist/core/model-setup/index.d.ts +4 -0
- package/dist/core/model-setup/index.js +5 -0
- package/dist/core/model-setup/index.js.map +1 -0
- package/dist/core/model-setup/plan.d.ts +56 -0
- package/dist/core/model-setup/plan.js +158 -0
- package/dist/core/model-setup/plan.js.map +1 -0
- package/dist/core/model-setup/service.d.ts +92 -0
- package/dist/core/model-setup/service.js +329 -0
- package/dist/core/model-setup/service.js.map +1 -0
- package/dist/core/model-setup/store.d.ts +29 -0
- package/dist/core/model-setup/store.js +116 -0
- package/dist/core/model-setup/store.js.map +1 -0
- package/dist/core/model-setup/wiring.d.ts +59 -0
- package/dist/core/model-setup/wiring.js +132 -0
- package/dist/core/model-setup/wiring.js.map +1 -0
- package/dist/core/prism.d.ts +38 -0
- package/dist/core/prism.js +43 -0
- package/dist/core/prism.js.map +1 -0
- package/dist/core/public-server.d.ts +7 -0
- package/dist/core/public-server.js +5 -2
- package/dist/core/public-server.js.map +1 -1
- package/dist/core/sessions.d.ts +6 -1
- package/dist/core/sessions.js +18 -14
- package/dist/core/sessions.js.map +1 -1
- package/dist/core/tensorrt-llm.d.ts +169 -0
- package/dist/core/tensorrt-llm.js +428 -0
- package/dist/core/tensorrt-llm.js.map +1 -0
- package/dist/core/types.d.ts +43 -1
- package/dist/core/types.js.map +1 -1
- package/dist/decision/args.d.ts +110 -0
- package/dist/decision/args.js +126 -0
- package/dist/decision/args.js.map +1 -0
- package/dist/decision/backoff.d.ts +25 -0
- package/dist/decision/backoff.js +33 -0
- package/dist/decision/backoff.js.map +1 -0
- package/dist/decision/engine-candidates.d.ts +52 -0
- package/dist/decision/engine-candidates.js +138 -0
- package/dist/decision/engine-candidates.js.map +1 -0
- package/dist/decision/engine-fallback.d.ts +41 -0
- package/dist/decision/engine-fallback.js +57 -0
- package/dist/decision/engine-fallback.js.map +1 -0
- package/dist/decision/engine-version.d.ts +27 -0
- package/dist/decision/engine-version.js +54 -0
- package/dist/decision/engine-version.js.map +1 -0
- package/dist/decision/engine.d.ts +114 -0
- package/dist/decision/engine.js +253 -0
- package/dist/decision/engine.js.map +1 -0
- package/dist/decision/http.d.ts +36 -0
- package/dist/decision/http.js +73 -0
- package/dist/decision/http.js.map +1 -0
- package/dist/decision/index.d.ts +20 -0
- package/dist/decision/index.js +21 -0
- package/dist/decision/index.js.map +1 -0
- package/dist/decision/model-facts.d.ts +32 -0
- package/dist/decision/model-facts.js +40 -0
- package/dist/decision/model-facts.js.map +1 -0
- package/dist/decision/outcome.d.ts +35 -0
- package/dist/decision/outcome.js +105 -0
- package/dist/decision/outcome.js.map +1 -0
- package/dist/decision/process.d.ts +71 -0
- package/dist/decision/process.js +142 -0
- package/dist/decision/process.js.map +1 -0
- package/dist/decision/readiness.d.ts +74 -0
- package/dist/decision/readiness.js +165 -0
- package/dist/decision/readiness.js.map +1 -0
- package/dist/decision/request.d.ts +11 -0
- package/dist/decision/request.js +56 -0
- package/dist/decision/request.js.map +1 -0
- package/dist/decision/service.d.ts +202 -0
- package/dist/decision/service.js +711 -0
- package/dist/decision/service.js.map +1 -0
- package/dist/decision/upstream-version.d.ts +23 -0
- package/dist/decision/upstream-version.js +34 -0
- package/dist/decision/upstream-version.js.map +1 -0
- package/dist/decision/wiring.d.ts +67 -0
- package/dist/decision/wiring.js +101 -0
- package/dist/decision/wiring.js.map +1 -0
- package/dist/diffusion/args.d.ts +7 -4
- package/dist/diffusion/args.js +31 -7
- package/dist/diffusion/args.js.map +1 -1
- package/dist/diffusion/compat.d.ts +1 -1
- package/dist/diffusion/compat.js +6 -2
- package/dist/diffusion/compat.js.map +1 -1
- package/dist/diffusion/image-job.js +1 -1
- package/dist/diffusion/image-job.js.map +1 -1
- package/dist/diffusion/install.d.ts +35 -11
- package/dist/diffusion/install.js +74 -49
- package/dist/diffusion/install.js.map +1 -1
- package/dist/diffusion/job-kind.d.ts +33 -7
- package/dist/diffusion/jobs.d.ts +6 -1
- package/dist/diffusion/jobs.js +64 -18
- package/dist/diffusion/jobs.js.map +1 -1
- package/dist/diffusion/parse.d.ts +1 -2
- package/dist/diffusion/parse.js +1 -13
- package/dist/diffusion/parse.js.map +1 -1
- package/dist/diffusion/progress.js +3 -0
- package/dist/diffusion/progress.js.map +1 -1
- package/dist/diffusion/server-process.d.ts +6 -0
- package/dist/diffusion/server-process.js +3 -2
- package/dist/diffusion/server-process.js.map +1 -1
- package/dist/diffusion/service.d.ts +52 -6
- package/dist/diffusion/service.js +215 -40
- package/dist/diffusion/service.js.map +1 -1
- package/dist/diffusion/session.d.ts +27 -3
- package/dist/diffusion/session.js +108 -17
- package/dist/diffusion/session.js.map +1 -1
- package/dist/diffusion/state.d.ts +19 -1
- package/dist/diffusion/state.js +10 -0
- package/dist/diffusion/state.js.map +1 -1
- package/dist/diffusion/tracker.d.ts +21 -1
- package/dist/diffusion/tracker.js +43 -6
- package/dist/diffusion/tracker.js.map +1 -1
- package/dist/diffusion/types.d.ts +7 -0
- package/dist/diffusion/types.js +11 -0
- package/dist/diffusion/types.js.map +1 -1
- package/dist/diffusion/validate.js +2 -2
- package/dist/diffusion/validate.js.map +1 -1
- package/dist/diffusion/video-estimate.d.ts +214 -0
- package/dist/diffusion/video-estimate.js +402 -0
- package/dist/diffusion/video-estimate.js.map +1 -0
- package/dist/diffusion/video-eta.d.ts +90 -0
- package/dist/diffusion/video-eta.js +248 -0
- package/dist/diffusion/video-eta.js.map +1 -0
- package/dist/diffusion/video-history.d.ts +45 -0
- package/dist/diffusion/video-history.js +121 -0
- package/dist/diffusion/video-history.js.map +1 -0
- package/dist/diffusion/video-job.d.ts +13 -5
- package/dist/diffusion/video-job.js +42 -15
- package/dist/diffusion/video-job.js.map +1 -1
- package/dist/diffusion/video-recipe.js +3 -0
- package/dist/diffusion/video-recipe.js.map +1 -1
- package/dist/diffusion/video-tiling.d.ts +54 -0
- package/dist/diffusion/video-tiling.js +70 -0
- package/dist/diffusion/video-tiling.js.map +1 -0
- package/dist/diffusion/wiring.d.ts +14 -1
- package/dist/diffusion/wiring.js +7 -0
- package/dist/diffusion/wiring.js.map +1 -1
- package/dist/diffusion/workflow.d.ts +8 -2
- package/dist/diffusion/workflow.js +13 -5
- package/dist/diffusion/workflow.js.map +1 -1
- package/dist/downloads/downloader.d.ts +22 -0
- package/dist/downloads/downloader.js +57 -2
- package/dist/downloads/downloader.js.map +1 -1
- package/dist/downloads/protocol.d.ts +3 -8
- package/dist/downloads/protocol.js.map +1 -1
- package/dist/downloads/segmented.d.ts +40 -0
- package/dist/downloads/segmented.js +170 -0
- package/dist/downloads/segmented.js.map +1 -0
- package/dist/embedding/args.d.ts +39 -0
- package/dist/embedding/args.js +46 -0
- package/dist/embedding/args.js.map +1 -0
- package/dist/embedding/engine.d.ts +50 -0
- package/dist/embedding/engine.js +100 -0
- package/dist/embedding/engine.js.map +1 -0
- package/dist/embedding/ffmpeg.d.ts +18 -0
- package/dist/embedding/ffmpeg.js +63 -0
- package/dist/embedding/ffmpeg.js.map +1 -0
- package/dist/embedding/index.d.ts +16 -0
- package/dist/embedding/index.js +17 -0
- package/dist/embedding/index.js.map +1 -0
- package/dist/embedding/model-facts.d.ts +51 -0
- package/dist/embedding/model-facts.js +84 -0
- package/dist/embedding/model-facts.js.map +1 -0
- package/dist/embedding/process.d.ts +70 -0
- package/dist/embedding/process.js +140 -0
- package/dist/embedding/process.js.map +1 -0
- package/dist/embedding/readiness.d.ts +56 -0
- package/dist/embedding/readiness.js +114 -0
- package/dist/embedding/readiness.js.map +1 -0
- package/dist/embedding/request.d.ts +43 -0
- package/dist/embedding/request.js +285 -0
- package/dist/embedding/request.js.map +1 -0
- package/dist/embedding/service.d.ts +152 -0
- package/dist/embedding/service.js +602 -0
- package/dist/embedding/service.js.map +1 -0
- package/dist/embedding/wiring.d.ts +51 -0
- package/dist/embedding/wiring.js +79 -0
- package/dist/embedding/wiring.js.map +1 -0
- package/dist/engine-builds/builds.d.ts +54 -0
- package/dist/engine-builds/builds.js +151 -0
- package/dist/engine-builds/builds.js.map +1 -0
- package/dist/engine-builds/index.d.ts +10 -0
- package/dist/engine-builds/index.js +11 -0
- package/dist/engine-builds/index.js.map +1 -0
- package/dist/engine-builds/install.d.ts +42 -0
- package/dist/engine-builds/install.js +100 -0
- package/dist/engine-builds/install.js.map +1 -0
- package/dist/engine-builds/manifest.d.ts +121 -0
- package/dist/engine-builds/manifest.js +254 -0
- package/dist/engine-builds/manifest.js.map +1 -0
- package/dist/engine-builds/order.d.ts +24 -0
- package/dist/engine-builds/order.js +44 -0
- package/dist/engine-builds/order.js.map +1 -0
- package/dist/engine-builds/select.d.ts +123 -0
- package/dist/engine-builds/select.js +223 -0
- package/dist/engine-builds/select.js.map +1 -0
- package/dist/engine-builds/service.d.ts +123 -0
- package/dist/engine-builds/service.js +392 -0
- package/dist/engine-builds/service.js.map +1 -0
- package/dist/engines/descriptor-order.d.ts +28 -0
- package/dist/engines/descriptor-order.js +60 -0
- package/dist/engines/descriptor-order.js.map +1 -0
- package/dist/engines/engine-build.d.ts +23 -0
- package/dist/engines/engine-build.js +58 -0
- package/dist/engines/engine-build.js.map +1 -0
- package/dist/engines/index.d.ts +10 -0
- package/dist/engines/index.js +11 -0
- package/dist/engines/index.js.map +1 -0
- package/dist/engines/llamacpp.d.ts +71 -0
- package/dist/engines/llamacpp.js +176 -0
- package/dist/engines/llamacpp.js.map +1 -0
- package/dist/engines/managed.d.ts +49 -0
- package/dist/engines/managed.js +88 -0
- package/dist/engines/managed.js.map +1 -0
- package/dist/engines/service.d.ts +46 -0
- package/dist/engines/service.js +61 -0
- package/dist/engines/service.js.map +1 -0
- package/dist/engines/versions.d.ts +65 -0
- package/dist/engines/versions.js +259 -0
- package/dist/engines/versions.js.map +1 -0
- package/dist/hardware/cpu-flags.d.ts +44 -0
- package/dist/hardware/cpu-flags.js +206 -0
- package/dist/hardware/cpu-flags.js.map +1 -0
- package/dist/hardware/drm-sysfs.d.ts +51 -0
- package/dist/hardware/drm-sysfs.js +67 -0
- package/dist/hardware/drm-sysfs.js.map +1 -0
- package/dist/hardware/facts.d.ts +42 -0
- package/dist/hardware/facts.js +92 -0
- package/dist/hardware/facts.js.map +1 -0
- package/dist/hardware/index.d.ts +19 -3
- package/dist/hardware/index.js +19 -3
- package/dist/hardware/index.js.map +1 -1
- package/dist/hardware/merge.d.ts +55 -0
- package/dist/hardware/merge.js +141 -0
- package/dist/hardware/merge.js.map +1 -0
- package/dist/hardware/nvidia-smi.d.ts +43 -0
- package/dist/hardware/nvidia-smi.js +118 -0
- package/dist/hardware/nvidia-smi.js.map +1 -0
- package/dist/hardware/probe-common.d.ts +72 -0
- package/dist/hardware/probe-common.js +106 -0
- package/dist/hardware/probe-common.js.map +1 -0
- package/dist/hardware/probe-darwin.d.ts +20 -0
- package/dist/hardware/probe-darwin.js +76 -0
- package/dist/hardware/probe-darwin.js.map +1 -0
- package/dist/hardware/probe-linux.d.ts +9 -0
- package/dist/hardware/probe-linux.js +133 -0
- package/dist/hardware/probe-linux.js.map +1 -0
- package/dist/hardware/probe-windows.d.ts +7 -0
- package/dist/hardware/probe-windows.js +50 -0
- package/dist/hardware/probe-windows.js.map +1 -0
- package/dist/hardware/probe.d.ts +16 -0
- package/dist/hardware/probe.js +65 -0
- package/dist/hardware/probe.js.map +1 -0
- package/dist/hardware/service.d.ts +60 -0
- package/dist/hardware/service.js +114 -0
- package/dist/hardware/service.js.map +1 -0
- package/dist/hardware/vulkan.d.ts +47 -0
- package/dist/hardware/vulkan.js +148 -0
- package/dist/hardware/vulkan.js.map +1 -0
- package/dist/hardware/windows-video.d.ts +97 -0
- package/dist/hardware/windows-video.js +214 -0
- package/dist/hardware/windows-video.js.map +1 -0
- package/dist/host/app-daemon.d.ts +47 -0
- package/dist/host/app-daemon.js +146 -0
- package/dist/host/app-daemon.js.map +1 -0
- package/dist/host/index.d.ts +5 -3
- package/dist/host/index.js +5 -3
- package/dist/host/index.js.map +1 -1
- package/dist/host/log-file.d.ts +74 -0
- package/dist/host/log-file.js +214 -0
- package/dist/host/log-file.js.map +1 -0
- package/dist/host/recipes/enable-wsl.d.ts +63 -0
- package/dist/host/recipes/enable-wsl.js +73 -0
- package/dist/host/recipes/enable-wsl.js.map +1 -0
- package/dist/host/recipes/executor-io-windows.d.ts +69 -0
- package/dist/host/recipes/executor-io-windows.js +235 -0
- package/dist/host/recipes/executor-io-windows.js.map +1 -0
- package/dist/host/recipes/executor-io.d.ts +80 -0
- package/dist/host/recipes/executor-io.js +211 -0
- package/dist/host/recipes/executor-io.js.map +1 -0
- package/dist/host/recipes/executor.d.ts +81 -0
- package/dist/host/recipes/executor.js +678 -0
- package/dist/host/recipes/executor.js.map +1 -0
- package/dist/host/recipes/guest-executor.d.ts +29 -0
- package/dist/host/recipes/guest-executor.js +76 -0
- package/dist/host/recipes/guest-executor.js.map +1 -0
- package/dist/host/recipes/index.d.ts +16 -0
- package/dist/host/recipes/index.js +17 -0
- package/dist/host/recipes/index.js.map +1 -0
- package/dist/host/recipes/install-container-runtime.d.ts +389 -0
- package/dist/host/recipes/install-container-runtime.js +796 -0
- package/dist/host/recipes/install-container-runtime.js.map +1 -0
- package/dist/host/recipes/openpgp.d.ts +23 -0
- package/dist/host/recipes/openpgp.js +153 -0
- package/dist/host/recipes/openpgp.js.map +1 -0
- package/dist/host/recipes/request-file.d.ts +96 -0
- package/dist/host/recipes/request-file.js +103 -0
- package/dist/host/recipes/request-file.js.map +1 -0
- package/dist/lock/process-journal.d.ts +5 -2
- package/dist/lock/process-journal.js.map +1 -1
- package/dist/models/capabilities.d.ts +7 -0
- package/dist/models/capabilities.js +17 -1
- package/dist/models/capabilities.js.map +1 -1
- package/dist/models/compatibility/index.d.ts +10 -0
- package/dist/models/compatibility/index.js +11 -0
- package/dist/models/compatibility/index.js.map +1 -0
- package/dist/models/compatibility/inspect.d.ts +28 -0
- package/dist/models/compatibility/inspect.js +90 -0
- package/dist/models/compatibility/inspect.js.map +1 -0
- package/dist/models/compatibility/resolve.d.ts +59 -0
- package/dist/models/compatibility/resolve.js +219 -0
- package/dist/models/compatibility/resolve.js.map +1 -0
- package/dist/models/compatibility/rules-baseline.d.ts +2 -0
- package/dist/models/compatibility/rules-baseline.js +10 -0
- package/dist/models/compatibility/rules-baseline.js.map +1 -0
- package/dist/models/compatibility/rules-baseline.json +292 -0
- package/dist/models/compatibility/rules-service.d.ts +36 -0
- package/dist/models/compatibility/rules-service.js +109 -0
- package/dist/models/compatibility/rules-service.js.map +1 -0
- package/dist/models/compatibility/rules.d.ts +81 -0
- package/dist/models/compatibility/rules.js +164 -0
- package/dist/models/compatibility/rules.js.map +1 -0
- package/dist/models/compatibility/service.d.ts +62 -0
- package/dist/models/compatibility/service.js +138 -0
- package/dist/models/compatibility/service.js.map +1 -0
- package/dist/models/decision-checkpoint.d.ts +22 -0
- package/dist/models/decision-checkpoint.js +33 -0
- package/dist/models/decision-checkpoint.js.map +1 -0
- package/dist/models/gguf/classify.d.ts +27 -1
- package/dist/models/gguf/classify.js +78 -2
- package/dist/models/gguf/classify.js.map +1 -1
- package/dist/models/gguf/index.d.ts +1 -0
- package/dist/models/gguf/index.js +1 -0
- package/dist/models/gguf/index.js.map +1 -1
- package/dist/models/gguf/tensors.d.ts +57 -0
- package/dist/models/gguf/tensors.js +239 -0
- package/dist/models/gguf/tensors.js.map +1 -0
- package/dist/models/index.d.ts +2 -0
- package/dist/models/index.js +2 -0
- package/dist/models/index.js.map +1 -1
- package/dist/router/resolve.d.ts +7 -1
- package/dist/router/resolve.js +14 -1
- package/dist/router/resolve.js.map +1 -1
- package/dist/runtime/container/argv.d.ts +88 -0
- package/dist/runtime/container/argv.js +406 -0
- package/dist/runtime/container/argv.js.map +1 -0
- package/dist/runtime/container/docker-binary.d.ts +4 -0
- package/dist/runtime/container/docker-binary.js +34 -0
- package/dist/runtime/container/docker-binary.js.map +1 -0
- package/dist/runtime/container/env.d.ts +17 -0
- package/dist/runtime/container/env.js +52 -0
- package/dist/runtime/container/env.js.map +1 -0
- package/dist/runtime/container/exec.d.ts +27 -0
- package/dist/runtime/container/exec.js +79 -0
- package/dist/runtime/container/exec.js.map +1 -0
- package/dist/runtime/container/execution-journal.d.ts +55 -0
- package/dist/runtime/container/execution-journal.js +167 -0
- package/dist/runtime/container/execution-journal.js.map +1 -0
- package/dist/runtime/container/heartbeat.d.ts +47 -0
- package/dist/runtime/container/heartbeat.js +78 -0
- package/dist/runtime/container/heartbeat.js.map +1 -0
- package/dist/runtime/container/index.d.ts +30 -0
- package/dist/runtime/container/index.js +31 -0
- package/dist/runtime/container/index.js.map +1 -0
- package/dist/runtime/container/operations.d.ts +40 -0
- package/dist/runtime/container/operations.js +233 -0
- package/dist/runtime/container/operations.js.map +1 -0
- package/dist/runtime/container/pull.d.ts +48 -0
- package/dist/runtime/container/pull.js +280 -0
- package/dist/runtime/container/pull.js.map +1 -0
- package/dist/runtime/container/reconcile.d.ts +34 -0
- package/dist/runtime/container/reconcile.js +79 -0
- package/dist/runtime/container/reconcile.js.map +1 -0
- package/dist/runtime/container/types.d.ts +124 -0
- package/dist/runtime/container/types.js +2 -0
- package/dist/runtime/container/types.js.map +1 -0
- package/dist/runtime/container/watchdog.d.ts +114 -0
- package/dist/runtime/container/watchdog.js +418 -0
- package/dist/runtime/container/watchdog.js.map +1 -0
- package/dist/runtime/container/wiring.d.ts +78 -0
- package/dist/runtime/container/wiring.js +92 -0
- package/dist/runtime/container/wiring.js.map +1 -0
- package/dist/runtime/container/wsl-exec.d.ts +13 -0
- package/dist/runtime/container/wsl-exec.js +18 -0
- package/dist/runtime/container/wsl-exec.js.map +1 -0
- package/dist/runtime/environment/cached-document.d.ts +128 -0
- package/dist/runtime/environment/cached-document.js +218 -0
- package/dist/runtime/environment/cached-document.js.map +1 -0
- package/dist/runtime/environment/canonical-json.d.ts +78 -0
- package/dist/runtime/environment/canonical-json.js +86 -0
- package/dist/runtime/environment/canonical-json.js.map +1 -0
- package/dist/runtime/environment/descriptor-provider.d.ts +121 -0
- package/dist/runtime/environment/descriptor-provider.js +159 -0
- package/dist/runtime/environment/descriptor-provider.js.map +1 -0
- package/dist/runtime/environment/descriptor.d.ts +35 -0
- package/dist/runtime/environment/descriptor.js +239 -0
- package/dist/runtime/environment/descriptor.js.map +1 -0
- package/dist/runtime/environment/diagnostics.d.ts +50 -0
- package/dist/runtime/environment/diagnostics.js +114 -0
- package/dist/runtime/environment/diagnostics.js.map +1 -0
- package/dist/runtime/environment/document-fields.d.ts +39 -0
- package/dist/runtime/environment/document-fields.js +96 -0
- package/dist/runtime/environment/document-fields.js.map +1 -0
- package/dist/runtime/environment/environment-manifest-provider.d.ts +65 -0
- package/dist/runtime/environment/environment-manifest-provider.js +89 -0
- package/dist/runtime/environment/environment-manifest-provider.js.map +1 -0
- package/dist/runtime/environment/environment-manifest.d.ts +35 -0
- package/dist/runtime/environment/environment-manifest.js +131 -0
- package/dist/runtime/environment/environment-manifest.js.map +1 -0
- package/dist/runtime/environment/guest-host.d.ts +43 -0
- package/dist/runtime/environment/guest-host.js +88 -0
- package/dist/runtime/environment/guest-host.js.map +1 -0
- package/dist/runtime/environment/host-exec.d.ts +38 -0
- package/dist/runtime/environment/host-exec.js +145 -0
- package/dist/runtime/environment/host-exec.js.map +1 -0
- package/dist/runtime/environment/host-step-failure.d.ts +33 -0
- package/dist/runtime/environment/host-step-failure.js +34 -0
- package/dist/runtime/environment/host-step-failure.js.map +1 -0
- package/dist/runtime/environment/index.d.ts +56 -0
- package/dist/runtime/environment/index.js +57 -0
- package/dist/runtime/environment/index.js.map +1 -0
- package/dist/runtime/environment/installations.d.ts +25 -0
- package/dist/runtime/environment/installations.js +70 -0
- package/dist/runtime/environment/installations.js.map +1 -0
- package/dist/runtime/environment/inventory.d.ts +42 -0
- package/dist/runtime/environment/inventory.js +57 -0
- package/dist/runtime/environment/inventory.js.map +1 -0
- package/dist/runtime/environment/linux-blockers.d.ts +130 -0
- package/dist/runtime/environment/linux-blockers.js +251 -0
- package/dist/runtime/environment/linux-blockers.js.map +1 -0
- package/dist/runtime/environment/linux-docker-facts.d.ts +200 -0
- package/dist/runtime/environment/linux-docker-facts.js +342 -0
- package/dist/runtime/environment/linux-docker-facts.js.map +1 -0
- package/dist/runtime/environment/linux-docker-network.d.ts +80 -0
- package/dist/runtime/environment/linux-docker-network.js +195 -0
- package/dist/runtime/environment/linux-docker-network.js.map +1 -0
- package/dist/runtime/environment/linux-host.d.ts +22 -0
- package/dist/runtime/environment/linux-host.js +98 -0
- package/dist/runtime/environment/linux-host.js.map +1 -0
- package/dist/runtime/environment/linux-plan.d.ts +126 -0
- package/dist/runtime/environment/linux-plan.js +392 -0
- package/dist/runtime/environment/linux-plan.js.map +1 -0
- package/dist/runtime/environment/linux-probe.d.ts +245 -0
- package/dist/runtime/environment/linux-probe.js +315 -0
- package/dist/runtime/environment/linux-probe.js.map +1 -0
- package/dist/runtime/environment/linux-provisioner.d.ts +102 -0
- package/dist/runtime/environment/linux-provisioner.js +734 -0
- package/dist/runtime/environment/linux-provisioner.js.map +1 -0
- package/dist/runtime/environment/provisioner-descriptors.d.ts +17 -0
- package/dist/runtime/environment/provisioner-descriptors.js +62 -0
- package/dist/runtime/environment/provisioner-descriptors.js.map +1 -0
- package/dist/runtime/environment/recovery.d.ts +74 -0
- package/dist/runtime/environment/recovery.js +121 -0
- package/dist/runtime/environment/recovery.js.map +1 -0
- package/dist/runtime/environment/rootfs-download.d.ts +2 -0
- package/dist/runtime/environment/rootfs-download.js +47 -0
- package/dist/runtime/environment/rootfs-download.js.map +1 -0
- package/dist/runtime/environment/service.d.ts +280 -0
- package/dist/runtime/environment/service.js +776 -0
- package/dist/runtime/environment/service.js.map +1 -0
- package/dist/runtime/environment/state.d.ts +163 -0
- package/dist/runtime/environment/state.js +479 -0
- package/dist/runtime/environment/state.js.map +1 -0
- package/dist/runtime/environment/store-models.d.ts +19 -0
- package/dist/runtime/environment/store-models.js +30 -0
- package/dist/runtime/environment/store-models.js.map +1 -0
- package/dist/runtime/environment/store.d.ts +210 -0
- package/dist/runtime/environment/store.js +438 -0
- package/dist/runtime/environment/store.js.map +1 -0
- package/dist/runtime/environment/windows-environment-record.d.ts +48 -0
- package/dist/runtime/environment/windows-environment-record.js +89 -0
- package/dist/runtime/environment/windows-environment-record.js.map +1 -0
- package/dist/runtime/environment/windows-forwarding.d.ts +27 -0
- package/dist/runtime/environment/windows-forwarding.js +84 -0
- package/dist/runtime/environment/windows-forwarding.js.map +1 -0
- package/dist/runtime/environment/windows-guest-setup.d.ts +48 -0
- package/dist/runtime/environment/windows-guest-setup.js +157 -0
- package/dist/runtime/environment/windows-guest-setup.js.map +1 -0
- package/dist/runtime/environment/windows-host.d.ts +46 -0
- package/dist/runtime/environment/windows-host.js +163 -0
- package/dist/runtime/environment/windows-host.js.map +1 -0
- package/dist/runtime/environment/windows-plan.d.ts +78 -0
- package/dist/runtime/environment/windows-plan.js +216 -0
- package/dist/runtime/environment/windows-plan.js.map +1 -0
- package/dist/runtime/environment/windows-probe.d.ts +135 -0
- package/dist/runtime/environment/windows-probe.js +261 -0
- package/dist/runtime/environment/windows-probe.js.map +1 -0
- package/dist/runtime/environment/windows-provisioner.d.ts +113 -0
- package/dist/runtime/environment/windows-provisioner.js +782 -0
- package/dist/runtime/environment/windows-provisioner.js.map +1 -0
- package/dist/runtime/environment/wiring.d.ts +117 -0
- package/dist/runtime/environment/wiring.js +354 -0
- package/dist/runtime/environment/wiring.js.map +1 -0
- package/dist/runtime/foundation-models/runtime.d.ts +8 -0
- package/dist/runtime/foundation-models/runtime.js +5 -2
- package/dist/runtime/foundation-models/runtime.js.map +1 -1
- package/dist/runtime/llamacpp/args.d.ts +10 -1
- package/dist/runtime/llamacpp/args.js +43 -7
- package/dist/runtime/llamacpp/args.js.map +1 -1
- package/dist/runtime/llamacpp/fit-margin.d.ts +46 -0
- package/dist/runtime/llamacpp/fit-margin.js +47 -0
- package/dist/runtime/llamacpp/fit-margin.js.map +1 -0
- package/dist/runtime/llamacpp/index.d.ts +1 -0
- package/dist/runtime/llamacpp/index.js +1 -0
- package/dist/runtime/llamacpp/index.js.map +1 -1
- package/dist/runtime/llamacpp/load-plan.d.ts +23 -1
- package/dist/runtime/llamacpp/load-plan.js +82 -1
- package/dist/runtime/llamacpp/load-plan.js.map +1 -1
- package/dist/runtime/llamacpp/policy.d.ts +16 -0
- package/dist/runtime/llamacpp/policy.js +20 -0
- package/dist/runtime/llamacpp/policy.js.map +1 -1
- package/dist/runtime/llamacpp/probe.d.ts +21 -1
- package/dist/runtime/llamacpp/probe.js +32 -6
- package/dist/runtime/llamacpp/probe.js.map +1 -1
- package/dist/runtime/llamacpp/runtime.d.ts +56 -1
- package/dist/runtime/llamacpp/runtime.js +142 -15
- package/dist/runtime/llamacpp/runtime.js.map +1 -1
- package/dist/runtime/managed-engines/index.d.ts +8 -0
- package/dist/runtime/managed-engines/index.js +9 -0
- package/dist/runtime/managed-engines/index.js.map +1 -0
- package/dist/runtime/managed-engines/runtime.d.ts +140 -0
- package/dist/runtime/managed-engines/runtime.js +402 -0
- package/dist/runtime/managed-engines/runtime.js.map +1 -0
- package/dist/runtime/managed-engines/spec.d.ts +70 -0
- package/dist/runtime/managed-engines/spec.js +39 -0
- package/dist/runtime/managed-engines/spec.js.map +1 -0
- package/dist/runtime/managed-models/check.d.ts +50 -0
- package/dist/runtime/managed-models/check.js +157 -0
- package/dist/runtime/managed-models/check.js.map +1 -0
- package/dist/runtime/managed-models/compatibility.d.ts +256 -0
- package/dist/runtime/managed-models/compatibility.js +594 -0
- package/dist/runtime/managed-models/compatibility.js.map +1 -0
- package/dist/runtime/managed-models/delete.d.ts +15 -0
- package/dist/runtime/managed-models/delete.js +60 -0
- package/dist/runtime/managed-models/delete.js.map +1 -0
- package/dist/runtime/managed-models/guest-files.d.ts +8 -0
- package/dist/runtime/managed-models/guest-files.js +49 -0
- package/dist/runtime/managed-models/guest-files.js.map +1 -0
- package/dist/runtime/managed-models/index.d.ts +17 -0
- package/dist/runtime/managed-models/index.js +18 -0
- package/dist/runtime/managed-models/index.js.map +1 -0
- package/dist/runtime/managed-models/location.d.ts +22 -0
- package/dist/runtime/managed-models/location.js +53 -0
- package/dist/runtime/managed-models/location.js.map +1 -0
- package/dist/runtime/managed-models/migrate.d.ts +30 -0
- package/dist/runtime/managed-models/migrate.js +119 -0
- package/dist/runtime/managed-models/migrate.js.map +1 -0
- package/dist/runtime/managed-models/model-dir.d.ts +36 -0
- package/dist/runtime/managed-models/model-dir.js +136 -0
- package/dist/runtime/managed-models/model-dir.js.map +1 -0
- package/dist/runtime/managed-models/quant-format.d.ts +71 -0
- package/dist/runtime/managed-models/quant-format.js +299 -0
- package/dist/runtime/managed-models/quant-format.js.map +1 -0
- package/dist/runtime/managed-models/registry.d.ts +26 -0
- package/dist/runtime/managed-models/registry.js +66 -0
- package/dist/runtime/managed-models/registry.js.map +1 -0
- package/dist/runtime/managed-text/adapter.d.ts +283 -0
- package/dist/runtime/managed-text/adapter.js +96 -0
- package/dist/runtime/managed-text/adapter.js.map +1 -0
- package/dist/runtime/managed-text/backend-target.d.ts +3 -0
- package/dist/runtime/managed-text/backend-target.js +52 -0
- package/dist/runtime/managed-text/backend-target.js.map +1 -0
- package/dist/runtime/managed-text/deployment.d.ts +14 -0
- package/dist/runtime/managed-text/deployment.js +32 -0
- package/dist/runtime/managed-text/deployment.js.map +1 -0
- package/dist/runtime/managed-text/engine-cache.d.ts +15 -0
- package/dist/runtime/managed-text/engine-cache.js +64 -0
- package/dist/runtime/managed-text/engine-cache.js.map +1 -0
- package/dist/runtime/managed-text/gateway.d.ts +158 -0
- package/dist/runtime/managed-text/gateway.js +535 -0
- package/dist/runtime/managed-text/gateway.js.map +1 -0
- package/dist/runtime/managed-text/generation-defaults.d.ts +23 -0
- package/dist/runtime/managed-text/generation-defaults.js +51 -0
- package/dist/runtime/managed-text/generation-defaults.js.map +1 -0
- package/dist/runtime/managed-text/index.d.ts +26 -0
- package/dist/runtime/managed-text/index.js +27 -0
- package/dist/runtime/managed-text/index.js.map +1 -0
- package/dist/runtime/managed-text/lifecycle.d.ts +284 -0
- package/dist/runtime/managed-text/lifecycle.js +896 -0
- package/dist/runtime/managed-text/lifecycle.js.map +1 -0
- package/dist/runtime/managed-text/load-policy.d.ts +47 -0
- package/dist/runtime/managed-text/load-policy.js +111 -0
- package/dist/runtime/managed-text/load-policy.js.map +1 -0
- package/dist/runtime/managed-text/mount-source.d.ts +14 -0
- package/dist/runtime/managed-text/mount-source.js +18 -0
- package/dist/runtime/managed-text/mount-source.js.map +1 -0
- package/dist/runtime/managed-text/readiness.d.ts +33 -0
- package/dist/runtime/managed-text/readiness.js +88 -0
- package/dist/runtime/managed-text/readiness.js.map +1 -0
- package/dist/runtime/managed-text/request-rules.d.ts +18 -0
- package/dist/runtime/managed-text/request-rules.js +62 -0
- package/dist/runtime/managed-text/request-rules.js.map +1 -0
- package/dist/runtime/managed-text/types.d.ts +88 -0
- package/dist/runtime/managed-text/types.js +12 -0
- package/dist/runtime/managed-text/types.js.map +1 -0
- package/dist/runtime/managed-text/wsl-deployment.d.ts +24 -0
- package/dist/runtime/managed-text/wsl-deployment.js +70 -0
- package/dist/runtime/managed-text/wsl-deployment.js.map +1 -0
- package/dist/runtime/mlx/errors.d.ts +6 -1
- package/dist/runtime/mlx/errors.js +7 -2
- package/dist/runtime/mlx/errors.js.map +1 -1
- package/dist/runtime/mlx/runtime.d.ts +36 -1
- package/dist/runtime/mlx/runtime.js +85 -4
- package/dist/runtime/mlx/runtime.js.map +1 -1
- package/dist/runtime/shared/backend-output.d.ts +23 -0
- package/dist/runtime/shared/backend-output.js +32 -0
- package/dist/runtime/shared/backend-output.js.map +1 -0
- package/dist/runtime/shared/bounded-output.d.ts +24 -0
- package/dist/runtime/shared/bounded-output.js +63 -0
- package/dist/runtime/shared/bounded-output.js.map +1 -0
- package/dist/runtime/shared/external-sessions.js +9 -1
- package/dist/runtime/shared/external-sessions.js.map +1 -1
- package/dist/runtime/shared/gpu-occupancy.d.ts +60 -0
- package/dist/runtime/shared/gpu-occupancy.js +10 -0
- package/dist/runtime/shared/gpu-occupancy.js.map +1 -0
- package/dist/runtime/shared/index.d.ts +4 -0
- package/dist/runtime/shared/index.js +4 -0
- package/dist/runtime/shared/index.js.map +1 -1
- package/dist/runtime/shared/local-runtime.d.ts +35 -0
- package/dist/runtime/shared/ports.d.ts +16 -0
- package/dist/runtime/shared/ports.js +42 -1
- package/dist/runtime/shared/ports.js.map +1 -1
- package/dist/runtime/shared/process.d.ts +11 -0
- package/dist/runtime/shared/process.js +20 -0
- package/dist/runtime/shared/process.js.map +1 -1
- package/dist/runtime/shared/redact-args.d.ts +7 -0
- package/dist/runtime/shared/redact-args.js +26 -0
- package/dist/runtime/shared/redact-args.js.map +1 -0
- package/dist/runtime/shared/sidecar.d.ts +15 -0
- package/dist/runtime/shared/sidecar.js +43 -8
- package/dist/runtime/shared/sidecar.js.map +1 -1
- package/dist/runtime/tensorrt-llm/adapter.d.ts +247 -0
- package/dist/runtime/tensorrt-llm/adapter.js +1028 -0
- package/dist/runtime/tensorrt-llm/adapter.js.map +1 -0
- package/dist/runtime/tensorrt-llm/check.d.ts +12 -0
- package/dist/runtime/tensorrt-llm/check.js +22 -0
- package/dist/runtime/tensorrt-llm/check.js.map +1 -0
- package/dist/runtime/tensorrt-llm/compatibility.d.ts +124 -0
- package/dist/runtime/tensorrt-llm/compatibility.js +213 -0
- package/dist/runtime/tensorrt-llm/compatibility.js.map +1 -0
- package/dist/runtime/tensorrt-llm/cuda-graphs.d.ts +32 -0
- package/dist/runtime/tensorrt-llm/cuda-graphs.js +33 -0
- package/dist/runtime/tensorrt-llm/cuda-graphs.js.map +1 -0
- package/dist/runtime/tensorrt-llm/delete.d.ts +15 -0
- package/dist/runtime/tensorrt-llm/delete.js +60 -0
- package/dist/runtime/tensorrt-llm/delete.js.map +1 -0
- package/dist/runtime/tensorrt-llm/engine.d.ts +3 -0
- package/dist/runtime/tensorrt-llm/engine.js +30 -0
- package/dist/runtime/tensorrt-llm/engine.js.map +1 -0
- package/dist/runtime/tensorrt-llm/guest-files.d.ts +8 -0
- package/dist/runtime/tensorrt-llm/guest-files.js +49 -0
- package/dist/runtime/tensorrt-llm/guest-files.js.map +1 -0
- package/dist/runtime/tensorrt-llm/host-facts.d.ts +55 -0
- package/dist/runtime/tensorrt-llm/host-facts.js +87 -0
- package/dist/runtime/tensorrt-llm/host-facts.js.map +1 -0
- package/dist/runtime/tensorrt-llm/index.d.ts +30 -0
- package/dist/runtime/tensorrt-llm/index.js +31 -0
- package/dist/runtime/tensorrt-llm/index.js.map +1 -0
- package/dist/runtime/tensorrt-llm/installation.d.ts +23 -0
- package/dist/runtime/tensorrt-llm/installation.js +53 -0
- package/dist/runtime/tensorrt-llm/installation.js.map +1 -0
- package/dist/runtime/tensorrt-llm/kv-cache.d.ts +41 -0
- package/dist/runtime/tensorrt-llm/kv-cache.js +48 -0
- package/dist/runtime/tensorrt-llm/kv-cache.js.map +1 -0
- package/dist/runtime/tensorrt-llm/location.d.ts +20 -0
- package/dist/runtime/tensorrt-llm/location.js +48 -0
- package/dist/runtime/tensorrt-llm/location.js.map +1 -0
- package/dist/runtime/tensorrt-llm/model-dir.d.ts +38 -0
- package/dist/runtime/tensorrt-llm/model-dir.js +137 -0
- package/dist/runtime/tensorrt-llm/model-dir.js.map +1 -0
- package/dist/runtime/tensorrt-llm/prelaunch.d.ts +20 -0
- package/dist/runtime/tensorrt-llm/prelaunch.js +168 -0
- package/dist/runtime/tensorrt-llm/prelaunch.js.map +1 -0
- package/dist/runtime/tensorrt-llm/quant-format.d.ts +65 -0
- package/dist/runtime/tensorrt-llm/quant-format.js +189 -0
- package/dist/runtime/tensorrt-llm/quant-format.js.map +1 -0
- package/dist/runtime/tensorrt-llm/registry.d.ts +26 -0
- package/dist/runtime/tensorrt-llm/registry.js +68 -0
- package/dist/runtime/tensorrt-llm/registry.js.map +1 -0
- package/dist/runtime/tensorrt-llm/route-policy.d.ts +19 -0
- package/dist/runtime/tensorrt-llm/route-policy.js +17 -0
- package/dist/runtime/tensorrt-llm/route-policy.js.map +1 -0
- package/dist/runtime/tensorrt-llm/runtime.d.ts +130 -0
- package/dist/runtime/tensorrt-llm/runtime.js +395 -0
- package/dist/runtime/tensorrt-llm/runtime.js.map +1 -0
- package/dist/runtime/tensorrt-llm/settings.d.ts +7 -0
- package/dist/runtime/tensorrt-llm/settings.js +48 -0
- package/dist/runtime/tensorrt-llm/settings.js.map +1 -0
- package/dist/runtime/vllm/adapter.d.ts +46 -0
- package/dist/runtime/vllm/adapter.js +425 -0
- package/dist/runtime/vllm/adapter.js.map +1 -0
- package/dist/runtime/vllm/engine.d.ts +16 -0
- package/dist/runtime/vllm/engine.js +37 -0
- package/dist/runtime/vllm/engine.js.map +1 -0
- package/dist/runtime/vllm/index.d.ts +11 -0
- package/dist/runtime/vllm/index.js +12 -0
- package/dist/runtime/vllm/index.js.map +1 -0
- package/dist/runtime/vllm/memory.d.ts +69 -0
- package/dist/runtime/vllm/memory.js +123 -0
- package/dist/runtime/vllm/memory.js.map +1 -0
- package/dist/runtime/vllm/settings.d.ts +52 -0
- package/dist/runtime/vllm/settings.js +151 -0
- package/dist/runtime/vllm/settings.js.map +1 -0
- package/dist/runtime/wsl/guest-paths.d.ts +33 -0
- package/dist/runtime/wsl/guest-paths.js +83 -0
- package/dist/runtime/wsl/guest-paths.js.map +1 -0
- package/dist/runtime/wsl/guest-scope.d.ts +14 -0
- package/dist/runtime/wsl/guest-scope.js +98 -0
- package/dist/runtime/wsl/guest-scope.js.map +1 -0
- package/dist/runtime/wsl/index.d.ts +12 -0
- package/dist/runtime/wsl/index.js +13 -0
- package/dist/runtime/wsl/index.js.map +1 -0
- package/dist/runtime/wsl/keeper.d.ts +29 -0
- package/dist/runtime/wsl/keeper.js +48 -0
- package/dist/runtime/wsl/keeper.js.map +1 -0
- package/dist/runtime/wsl/transport.d.ts +106 -0
- package/dist/runtime/wsl/transport.js +181 -0
- package/dist/runtime/wsl/transport.js.map +1 -0
- package/dist/server/control/index.d.ts +1 -1
- package/dist/server/control/router.js +16 -1
- package/dist/server/control/router.js.map +1 -1
- package/dist/server/control/routes/backends.js +26 -0
- package/dist/server/control/routes/backends.js.map +1 -1
- package/dist/server/control/routes/decision.d.ts +10 -0
- package/dist/server/control/routes/decision.js +43 -0
- package/dist/server/control/routes/decision.js.map +1 -0
- package/dist/server/control/routes/diffusion-video.js +4 -0
- package/dist/server/control/routes/diffusion-video.js.map +1 -1
- package/dist/server/control/routes/diffusion.d.ts +1 -1
- package/dist/server/control/routes/diffusion.js +4 -9
- package/dist/server/control/routes/diffusion.js.map +1 -1
- package/dist/server/control/routes/embedding.d.ts +9 -0
- package/dist/server/control/routes/embedding.js +43 -0
- package/dist/server/control/routes/embedding.js.map +1 -0
- package/dist/server/control/routes/engine-builds.d.ts +14 -0
- package/dist/server/control/routes/engine-builds.js +82 -0
- package/dist/server/control/routes/engine-builds.js.map +1 -0
- package/dist/server/control/routes/engines.d.ts +17 -0
- package/dist/server/control/routes/engines.js +152 -0
- package/dist/server/control/routes/engines.js.map +1 -0
- package/dist/server/control/routes/environments.d.ts +16 -0
- package/dist/server/control/routes/environments.js +207 -0
- package/dist/server/control/routes/environments.js.map +1 -0
- package/dist/server/control/routes/hardware.d.ts +1 -1
- package/dist/server/control/routes/hardware.js +11 -6
- package/dist/server/control/routes/hardware.js.map +1 -1
- package/dist/server/control/routes/model-setups.d.ts +11 -0
- package/dist/server/control/routes/model-setups.js +55 -0
- package/dist/server/control/routes/model-setups.js.map +1 -0
- package/dist/server/control/routes/models.d.ts +3 -2
- package/dist/server/control/routes/models.js +45 -2
- package/dist/server/control/routes/models.js.map +1 -1
- package/dist/server/control/types.d.ts +172 -10
- package/dist/server/control/types.js.map +1 -1
- package/dist/server/http.d.ts +9 -1
- package/dist/server/http.js +51 -1
- package/dist/server/http.js.map +1 -1
- package/dist/server/public/decision-preview.d.ts +17 -0
- package/dist/server/public/decision-preview.js +119 -0
- package/dist/server/public/decision-preview.js.map +1 -0
- package/dist/server/public/decision.d.ts +32 -0
- package/dist/server/public/decision.js +135 -0
- package/dist/server/public/decision.js.map +1 -0
- package/dist/server/public/embedding.d.ts +31 -0
- package/dist/server/public/embedding.js +126 -0
- package/dist/server/public/embedding.js.map +1 -0
- package/dist/server/public/errors.d.ts +2 -1
- package/dist/server/public/errors.js +5 -2
- package/dist/server/public/errors.js.map +1 -1
- package/dist/server/public/forward.js +75 -3
- package/dist/server/public/forward.js.map +1 -1
- package/dist/server/public/gates.d.ts +18 -0
- package/dist/server/public/gates.js +41 -14
- package/dist/server/public/gates.js.map +1 -1
- package/dist/server/public/index.d.ts +6 -2
- package/dist/server/public/index.js +23 -7
- package/dist/server/public/index.js.map +1 -1
- package/dist/server/public/listing.d.ts +13 -2
- package/dist/server/public/listing.js +33 -9
- package/dist/server/public/listing.js.map +1 -1
- package/dist/server/public/policy.d.ts +9 -0
- package/dist/server/public/policy.js +68 -0
- package/dist/server/public/policy.js.map +1 -0
- package/dist/server/public/responses.js +10 -0
- package/dist/server/public/responses.js.map +1 -1
- package/dist/server/public/trace.d.ts +2 -2
- package/dist/server/public/trace.js +6 -2
- package/dist/server/public/trace.js.map +1 -1
- package/dist/server/public/types.d.ts +91 -1
- package/dist/settings/decision.d.ts +15 -0
- package/dist/settings/decision.js +88 -0
- package/dist/settings/decision.js.map +1 -0
- package/dist/settings/embedding.d.ts +12 -0
- package/dist/settings/embedding.js +82 -0
- package/dist/settings/embedding.js.map +1 -0
- package/dist/settings/index.d.ts +2 -0
- package/dist/settings/index.js +2 -0
- package/dist/settings/index.js.map +1 -1
- package/dist/settings/schema/atomic-prism.json +433 -0
- package/dist/settings/schema/tensorrt-llm.json +253 -0
- package/dist/settings/schema/vllm.json +273 -0
- package/dist/settings/schema.js +58 -0
- package/dist/settings/schema.js.map +1 -1
- package/dist/settings/store.d.ts +13 -2
- package/dist/settings/store.js +44 -0
- package/dist/settings/store.js.map +1 -1
- package/dist/util/index.d.ts +1 -0
- package/dist/util/index.js +1 -0
- package/dist/util/index.js.map +1 -1
- package/dist/util/result.d.ts +19 -0
- package/dist/util/result.js +15 -0
- package/dist/util/result.js.map +1 -0
- package/dist/version.d.ts +1 -1
- package/dist/version.js +1 -1
- package/dist/version.js.map +1 -1
- package/package.json +2 -2
- package/src/settings/schema/atomic-prism.json +433 -0
- package/src/settings/schema/tensorrt-llm.json +253 -0
- package/src/settings/schema/vllm.json +273 -0
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"adapter.js","sourceRoot":"","sources":["../../../src/runtime/tensorrt-llm/adapter.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;GAuHG;AACH,OAAO,EAAE,eAAe,EAAE,MAAM,0BAA0B,CAAA;AAC1D,OAAO,EACL,sBAAsB,EACtB,sBAAsB,GAEvB,MAAM,wCAAwC,CAAA;AAE/C,OAAO,EACL,qCAAqC,EACrC,qBAAqB,EACrB,uBAAuB,EACvB,YAAY,EACZ,oBAAoB,EACpB,iBAAiB,GAClB,MAAM,0BAA0B,CAAA;AACjC,OAAO,EAAE,uBAAuB,EAA8B,MAAM,kBAAkB,CAAA;AACtF,OAAO,EACL,2CAA2C,EAC3C,mBAAmB,EACnB,6BAA6B,EAC7B,2BAA2B,GAC5B,MAAM,eAAe,CAAA;AAUtB,MAAM,GAAG,GAAG,IAAI,IAAI,CAAC,CAAA;AAErB,gGAAgG;AAChG,WAAW;AACX,gGAAgG;AAEhG;;;wDAGwD;AACxD,MAAM,gBAAgB,GAAG,2BAA2B,CAAA;AAEpD,MAAM,CAAC,MAAM,+BAA+B,GAAG,GAAG,CAAA;AAClD,MAAM,CAAC,MAAM,+BAA+B,GAAG,SAAS,CAAA;AACxD,MAAM,CAAC,MAAM,mCAAmC,GAAG,IAAI,CAAA;AAEvD,MAAM,CAAC,MAAM,kCAAkC,GAAG,CAAC,CAAA;AACnD,MAAM,CAAC,MAAM,kCAAkC,GAAG,SAAS,CAAA;AAC3D,MAAM,CAAC,MAAM,sCAAsC,GAAG,IAAI,CAAA;AAE1D;iGACiG;AACjG,MAAM,CAAC,MAAM,uCAAuC,GAAG,GAAG,CAAA;AAC1D,MAAM,CAAC,MAAM,uCAAuC,GAAG,IAAI,CAAA;AAE3D,MAAM,CAAC,MAAM,qCAAqC,GAAG,CAAC,CAAA;AACtD,MAAM,CAAC,MAAM,qCAAqC,GAAG,IAAI,CAAA;AA8BzD,SAAS,OAAO,CAAC,OAAe,EAAE,KAAc;IAC9C,MAAM,IAAI,eAAe,CAAC,kBAAkB,EAAE,OAAO,EAAE,MAAM,CAAC,KAAK,CAAC,CAAC,CAAA;AACvE,CAAC;AAED,SAAS,aAAa,CAAC,KAAc;IACnC,IAAI,KAAK,KAAK,SAAS,IAAI,KAAK,KAAK,IAAI;QAAE,OAAO,IAAI,CAAA;IACtD,IAAI,OAAO,KAAK,KAAK,QAAQ,IAAI,CAAC,gBAAgB,CAAC,IAAI,CAAC,KAAK,CAAC,EAAE,CAAC;QAC/D,OAAO,CAAC,qEAAqE,EAAE,KAAK,CAAC,CAAA;IACvF,CAAC;IACD,OAAO,KAAK,CAAA;AACd,CAAC;AAED,SAAS,kBAAkB,CACzB,KAAc,EACd,KAAa,EACb,QAAgB,EAChB,GAAW,EACX,GAAW;IAEX,IAAI,KAAK,KAAK,SAAS,IAAI,KAAK,KAAK,IAAI;QAAE,OAAO,QAAQ,CAAA;IAC1D,IAAI,OAAO,KAAK,KAAK,QAAQ,IAAI,CAAC,MAAM,CAAC,SAAS,CAAC,KAAK,CAAC,IAAI,KAAK,GAAG,GAAG,IAAI,KAAK,GAAG,GAAG,EAAE,CAAC;QACxF,OAAO,CAAC,gBAAgB,KAAK,+BAA+B,GAAG,QAAQ,GAAG,GAAG,EAAE,KAAK,CAAC,CAAA;IACvF,CAAC;IACD,OAAO,KAAK,CAAA;AACd,CAAC;AAED,SAAS,gBAAgB,CAAC,KAAc;IACtC,IAAI,KAAK,KAAK,SAAS,IAAI,KAAK,KAAK,IAAI;QAAE,OAAO,2CAA2C,CAAA;IAC7F,MAAM,GAAG,GAAG,uCAAuC,CAAA;IACnD,MAAM,GAAG,GAAG,uCAAuC,CAAA;IACnD,IAAI,OAAO,KAAK,KAAK,QAAQ,IAAI,CAAC,MAAM,CAAC,QAAQ,CAAC,KAAK,CAAC,IAAI,KAAK,GAAG,GAAG,IAAI,KAAK,GAAG,GAAG,EAAE,CAAC;QACvF,OAAO,CAAC,kEAAkE,GAAG,QAAQ,GAAG,GAAG,EAAE,KAAK,CAAC,CAAA;IACrG,CAAC;IACD,OAAO,KAAK,CAAA;AACd,CAAC;AAED,SAAS,cAAc,CACrB,KAAc,EACd,KAAa,EACb,OAAqB,EACrB,QAAW;IAEX,IAAI,KAAK,KAAK,SAAS,IAAI,KAAK,KAAK,IAAI,IAAI,KAAK,KAAK,EAAE;QAAE,OAAO,QAAQ,CAAA;IAC1E,IAAI,OAAO,KAAK,KAAK,QAAQ,IAAI,CAAE,OAA6B,CAAC,QAAQ,CAAC,KAAK,CAAC,EAAE,CAAC;QACjF,OAAO,CAAC,gBAAgB,KAAK,mBAAmB,OAAO,CAAC,IAAI,CAAC,IAAI,CAAC,GAAG,EAAE,KAAK,CAAC,CAAA;IAC/E,CAAC;IACD,OAAO,KAAU,CAAA;AACnB,CAAC;AAED,SAAS,YAAY,CAAC,KAAc,EAAE,KAAa,EAAE,QAAiB;IACpE,IAAI,KAAK,KAAK,SAAS,IAAI,KAAK,KAAK,IAAI;QAAE,OAAO,QAAQ,CAAA;IAC1D,IAAI,KAAK,KAAK,IAAI,IAAI,KAAK,KAAK,MAAM;QAAE,OAAO,IAAI,CAAA;IACnD,IAAI,KAAK,KAAK,KAAK,IAAI,KAAK,KAAK,OAAO;QAAE,OAAO,KAAK,CAAA;IACtD,OAAO,OAAO,CAAC,gBAAgB,KAAK,yBAAyB,EAAE,KAAK,CAAC,CAAA;AACvE,CAAC;AAED,SAAS,2BAA2B,CAAC,KAAc;IACjD,IAAI,KAAK,KAAK,SAAS,IAAI,KAAK,KAAK,IAAI;QAAE,OAAO,IAAI,CAAA;IACtD,MAAM,GAAG,GAAG,qCAAqC,CAAA;IACjD,MAAM,GAAG,GAAG,qCAAqC,CAAA;IACjD,IAAI,OAAO,KAAK,KAAK,QAAQ,IAAI,CAAC,MAAM,CAAC,SAAS,CAAC,KAAK,CAAC,IAAI,KAAK,GAAG,GAAG,IAAI,KAAK,GAAG,GAAG,EAAE,CAAC;QACxF,OAAO,CAAC,gEAAgE,GAAG,QAAQ,GAAG,YAAY,EAAE,KAAK,CAAC,CAAA;IAC5G,CAAC;IACD,OAAO,KAAK,CAAA;AACd,CAAC;AAED,wGAAwG;AACxG,MAAM,UAAU,2BAA2B,CAAC,GAAY;IACtD,IAAI,GAAG,KAAK,SAAS,IAAI,CAAC,OAAO,GAAG,KAAK,QAAQ,IAAI,GAAG,KAAK,IAAI,IAAI,KAAK,CAAC,OAAO,CAAC,GAAG,CAAC,CAAC,EAAE,CAAC;QACzF,OAAO,CAAC,0CAA0C,EAAE,GAAG,CAAC,CAAA;IAC1D,CAAC;IACD,MAAM,CAAC,GAAG,CAAC,GAAG,IAAI,EAAE,CAA4B,CAAA;IAChD,MAAM,cAAc,GAAG,kBAAkB,CACvC,CAAC,CAAC,gBAAgB,CAAC,EACnB,gBAAgB,EAChB,mCAAmC,EACnC,+BAA+B,EAC/B,+BAA+B,CAChC,CAAA;IACD,MAAM,iBAAiB,GAAG,kBAAkB,CAC1C,CAAC,CAAC,mBAAmB,CAAC,EACtB,mBAAmB,EACnB,sCAAsC,EACtC,kCAAkC,EAClC,kCAAkC,CACnC,CAAA;IACD,+FAA+F;IAC/F,gGAAgG;IAChG,6FAA6F;IAC7F,IAAI,iBAAiB,IAAI,cAAc,EAAE,CAAC;QACxC,OAAO,CACL,mCAAmC,iBAAiB,uCAAuC,cAAc,IAAI,EAC7G,iBAAiB,CAClB,CAAA;IACH,CAAC;IACD,OAAO;QACL,MAAM,EAAE,aAAa,CAAC,CAAC,CAAC,QAAQ,CAAC,CAAC;QAClC,cAAc;QACd,iBAAiB;QACjB,iCAAiC,EAAE,gBAAgB,CAAC,CAAC,CAAC,mCAAmC,CAAC,CAAC;QAC3F,cAAc,EAAE,kBAAkB,CAChC,CAAC,CAAC,gBAAgB,CAAC,EACnB,gBAAgB,EAChB,2BAA2B,EAC3B,CAAC,EACD,+BAA+B,CAChC;QACD,mBAAmB,EACjB,CAAC,CAAC,qBAAqB,CAAC,KAAK,SAAS,IAAI,CAAC,CAAC,qBAAqB,CAAC,KAAK,IAAI;YACzE,CAAC,CAAC,IAAI;YACN,CAAC,CAAC,kBAAkB,CAChB,CAAC,CAAC,qBAAqB,CAAC,EACxB,qBAAqB,EACrB,CAAC,EACD,CAAC,EACD,oCAAoC,CACrC;QACP,WAAW,EAAE,cAAc,CAAC,CAAC,CAAC,aAAa,CAAC,EAAE,aAAa,EAAE,CAAC,MAAM,EAAE,IAAI,EAAE,KAAK,CAAU,EAAE,MAAM,CAAC;QACpG,cAAc,EAAE,cAAc,CAAC,CAAC,CAAC,gBAAgB,CAAC,EAAE,gBAAgB,EAAE,CAAC,MAAM,EAAE,KAAK,CAAU,EAAE,MAAM,CAAC;QACvG,oBAAoB,EAAE,2BAA2B,CAAC,CAAC,CAAC,sBAAsB,CAAC,CAAC;QAC5E,qBAAqB,EAAE,YAAY,CAAC,CAAC,CAAC,uBAAuB,CAAC,EAAE,uBAAuB,EAAE,IAAI,CAAC;QAC9F,iBAAiB,EAAE,YAAY,CAAC,CAAC,CAAC,mBAAmB,CAAC,EAAE,mBAAmB,EAAE,IAAI,CAAC;QAClF,yBAAyB,EAAE,cAAc,CACvC,CAAC,CAAC,2BAA2B,CAAC,EAC9B,2BAA2B,EAC3B,CAAC,qBAAqB,EAAE,iBAAiB,CAAU,EACnD,qBAAqB,CACtB;QACD,KAAK,EAAE,cAAc,CAAC,CAAC,CAAC,OAAO,CAAC,EAAE,OAAO,EAAE,CAAC,MAAM,EAAE,SAAS,EAAE,UAAU,EAAE,SAAS,CAAU,EAAE,MAAM,CAAC;QACvG,UAAU,EAAE,sBAAsB,CAAC,cAAc,EAAE,CAAC,CAAC;KACtD,CAAA;AACH,CAAC;AAED,gGAAgG;AAChG,SAAS;AACT,gGAAgG;AAEhG,mDAAmD;AACnD,MAAM,CAAC,MAAM,2BAA2B,GAAG,IAAI,CAAA;AAE/C;;;;;;;;;;;;;;;;;;GAkBG;AACH,MAAM,CAAC,MAAM,6BAA6B,GAAG,sBAAsB,CAAA;AACnE,MAAM,CAAC,MAAM,oCAAoC,GAAG,qCAAqC,CAAA;AAEzF;;;;;;GAMG;AACH,SAAS,aAAa,CACpB,MAAe,EACf,WAA0B,EAC1B,KAAK,GAAG,KAAK,EACb,aAAa,GAAG,KAAK,EACrB,QAGC;IAED,MAAM,KAAK,GAAa,EAAE,CAAA;IAC1B,IAAI,MAAM;QAAE,KAAK,CAAC,IAAI,CAAC,oCAAoC,CAAC,OAAO,EAAE,CAAC,CAAA;IACtE,MAAM,YAAY,GAAG,QAAQ,EAAE,qBAAqB,KAAK,KAAK,CAAA;IAC9D,IAAI,WAAW,KAAK,IAAI,IAAI,KAAK,IAAI,YAAY,EAAE,CAAC;QAClD,KAAK,CAAC,IAAI,CAAC,kBAAkB,CAAC,CAAA;QAC9B,IAAI,WAAW,KAAK,IAAI;YAAE,KAAK,CAAC,IAAI,CAAC,iBAAiB,WAAW,EAAE,CAAC,CAAA;QACpE,IAAI,KAAK;YAAE,KAAK,CAAC,IAAI,CAAC,cAAc,CAAC,CAAA;QACrC,IAAI,YAAY;YAAE,KAAK,CAAC,IAAI,CAAC,6BAA6B,CAAC,CAAA;IAC7D,CAAC;IACD,4FAA4F;IAC5F,IAAI,aAAa;QAAE,KAAK,CAAC,IAAI,CAAC,yBAAyB,CAAC,CAAA;IACxD,IAAI,QAAQ,EAAE,iBAAiB,KAAK,KAAK;QAAE,KAAK,CAAC,IAAI,CAAC,iCAAiC,CAAC,CAAA;IACxF,IAAI,QAAQ,EAAE,yBAAyB,KAAK,iBAAiB,EAAE,CAAC;QAC9D,KAAK,CAAC,IAAI,CAAC,mBAAmB,EAAE,8CAA8C,CAAC,CAAA;IACjF,CAAC;IACD,IAAI,QAAQ,KAAK,SAAS,IAAI,QAAQ,CAAC,KAAK,KAAK,MAAM;QAAE,KAAK,CAAC,IAAI,CAAC,UAAU,QAAQ,CAAC,KAAK,EAAE,CAAC,CAAA;IAC/F,OAAO,KAAK,CAAC,MAAM,KAAK,CAAC,CAAC,CAAC,CAAC,IAAI,CAAC,CAAC,CAAC,GAAG,KAAK,CAAC,IAAI,CAAC,IAAI,CAAC,IAAI,CAAA;AAC5D,CAAC;AAED,OAAO,EAAE,uCAAuC,EAAE,MAAM,kBAAkB,CAAA;AAE1E,gEAAgE;AAChE,SAAS,aAAa,CAAC,iBAA4C;IACjE,IAAI,iBAAiB,KAAK,IAAI,IAAI,iBAAiB,KAAK,SAAS;QAAE,OAAO,KAAK,CAAA;IAC/E,MAAM,CAAC,KAAK,EAAE,KAAK,CAAC,GAAG,iBAAiB,CAAC,KAAK,CAAC,GAAG,CAAC,CAAC,GAAG,CAAC,CAAC,IAAI,EAAE,EAAE,CAAC,MAAM,CAAC,IAAI,CAAC,CAAC,CAAA;IAC/E,IAAI,KAAK,KAAK,SAAS,IAAI,MAAM,CAAC,KAAK,CAAC,KAAK,CAAC;QAAE,OAAO,KAAK,CAAA;IAC5D,OAAO,KAAK,GAAG,CAAC,IAAI,CAAC,KAAK,KAAK,CAAC,IAAI,CAAC,KAAK,IAAI,CAAC,CAAC,IAAI,CAAC,CAAC,CAAA;AACxD,CAAC;AAED;;;;;;;;;GASG;AACH,MAAM,CAAC,MAAM,2BAA2B,GAAG,CAAC,CAAA;AAC5C,MAAM,CAAC,MAAM,+BAA+B,GAAG,GAAG,CAAA;AAClD,MAAM,CAAC,MAAM,oCAAoC,GAAG,UAAU,CAAA;AAE9D;kFACkF;AAClF,MAAM,mBAAmB,GAAG,SAAS,CAAA;AAErC,8FAA8F;AAC9F,SAAS,YAAY,CAAC,IAAa;IACjC,OAAO,OAAO,IAAI,KAAK,QAAQ;QAC7B,IAAI,KAAK,IAAI;QACb,OAAQ,IAA8B,CAAC,UAAU,KAAK,SAAS;QAC/D,CAAC,CAAE,IAA8B;QACjC,CAAC,CAAC,SAAS,CAAA;AACf,CAAC;AAED;;;;;;;;;;;;;;;;qFAgBqF;AACrF,MAAM,UAAU,sBAAsB,CACpC,OAAkD;IAElD,MAAM,EAAE,QAAQ,EAAE,SAAS,EAAE,eAAe,EAAE,mBAAmB,EAAE,MAAM,EAAE,GAAG,OAAO,CAAA;IACrF,MAAM,IAAI,GAAa;QACrB,cAAc;QACd,OAAO;QACP,SAAS;QACT,QAAQ;QACR,mBAAmB;QACnB,QAAQ;QACR,MAAM,CAAC,2BAA2B,CAAC;QACnC,eAAe;QACf,MAAM,CAAC,QAAQ,CAAC,cAAc,CAAC;QAC/B,kBAAkB;QAClB,MAAM,CAAC,QAAQ,CAAC,cAAc,CAAC;QAC/B,kBAAkB;QAClB,MAAM,CAAC,QAAQ,CAAC,cAAc,CAAC;QAC/B,qCAAqC;QACrC,MAAM,CAAC,QAAQ,CAAC,iCAAiC,CAAC;KACnD,CAAA;IACD,IAAI,MAAM,EAAE,WAAW,IAAI,IAAI;QAAE,IAAI,CAAC,IAAI,CAAC,eAAe,EAAE,MAAM,CAAC,WAAW,CAAC,CAAA;IAC/E,IAAI,MAAM,EAAE,gBAAgB,IAAI,IAAI;QAAE,IAAI,CAAC,IAAI,CAAC,oBAAoB,EAAE,MAAM,CAAC,gBAAgB,CAAC,CAAA;IAC9F,kGAAkG;IAClG,iGAAiG;IACjG,sGAAsG;IACtG,wFAAwF;IACxF,qGAAqG;IACrG,wGAAwG;IACxG,sGAAsG;IACtG,mGAAmG;IACnG,2EAA2E;IAC3E,mGAAmG;IACnG,+FAA+F;IAC/F,MAAM,aAAa,GAAG,CAAC,CACrB,YAAY,CAAC,OAAO,CAAC,IAAI,CAAC,EAAE,UAAU;QACtC,uBAAuB,CAAC,QAAQ,CAAC,WAAW,EAAE,OAAO,CAAC,iBAAiB,IAAI,IAAI,CAAC,CACjF,CAAA;IACD,MAAM,KAAK,GAAG,QAAQ,CAAC,cAAc,KAAK,KAAK,IAAI,aAAa,CAAC,OAAO,CAAC,oBAAoB,CAAC,CAAA;IAC9F,kGAAkG;IAClG,qEAAqE;IACrE,MAAM,OAAO,GAAG,aAAa,CAC3B,MAAM,EAAE,iBAAiB,KAAK,IAAI,EAClC,QAAQ,CAAC,mBAAmB,KAAK,IAAI;QACnC,CAAC,CAAC,2BAA2B,CAAC,QAAQ,CAAC,mBAAmB,CAAC;QAC3D,CAAC,CAAC,OAAO,CAAC,aAAa;YACrB,CAAC,CAAC,6BAA6B,CAAC,QAAQ,CAAC,cAAc,CAAC;YACxD,CAAC,CAAC,mBAAmB,CAAC,QAAQ,CAAC,cAAc,EAAE,QAAQ,CAAC,cAAc,CAAC,EAC3E,KAAK,EACL,aAAa,EACb,QAAQ,CACT,CAAA;IACD,IAAI,OAAO,KAAK,IAAI,EAAE,CAAC;QACrB,IAAI,CAAC,IAAI,CAAC,yBAAyB,EAAE,GAAG,mBAAmB,IAAI,6BAA6B,EAAE,CAAC,CAAA;IACjG,CAAC;IAED,OAAO;QACL,MAAM,EAAE,EAAE,cAAc,EAAE,2BAA2B,EAAE;QACvD,IAAI;QACJ,GAAG,CAAC,OAAO,KAAK,IAAI,CAAC,CAAC,CAAC,EAAE,KAAK,EAAE,EAAE,CAAC,6BAA6B,CAAC,EAAE,OAAO,EAAE,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC;QACpF,uFAAuF;QACvF,6FAA6F;QAC7F,+FAA+F;QAC/F,uEAAuE;QACvE,GAAG,EAAE;YACH,uBAAuB,EAAE,GAAG,eAAe,WAAW;YACtD,gBAAgB,EAAE,GAAG,eAAe,SAAS;YAC7C,eAAe,EAAE,GAAG,eAAe,OAAO;YAC1C,8FAA8F;YAC9F,qBAAqB,EAAE,GAAG;SAC3B;KACF,CAAA;AACH,CAAC;AAED,gGAAgG;AAChG,oBAAoB;AACpB,gGAAgG;AAEhG;;;;;;;;;GASG;AACH,MAAM,CAAC,MAAM,8BAA8B,GAAG,OAAO,CAAA;AACrD,+GAA+G;AAC/G,MAAM,CAAC,MAAM,iCAAiC,GAAG,MAAM,CAAA;AACvD,0FAA0F;AAC1F,MAAM,CAAC,MAAM,6BAA6B,GAAG,GAAG,CAAA;AAEhD;;;;;;qGAMqG;AACrG,MAAM,UAAU,6BAA6B,CAAC,WAAmB,EAAE,QAA6B;IAC9F,IAAI,QAAQ,CAAC,oBAAoB,KAAK,IAAI;QAAE,OAAO,QAAQ,CAAC,oBAAoB,GAAG,IAAI,CAAA;IACvF,MAAM,SAAS,GAAG,IAAI,CAAC,GAAG,CAAC,WAAW,EAAE,CAAC,CAAC,GAAG,GAAG,CAAA;IAChD,MAAM,QAAQ,GAAG,8BAA8B,GAAG,iCAAiC,GAAG,SAAS,CAAA;IAC/F,OAAO,IAAI,CAAC,IAAI,CAAC,QAAQ,GAAG,6BAA6B,CAAC,CAAA;AAC5D,CAAC;AAED,gGAAgG;AAChG,sBAAsB;AACtB,gGAAgG;AAEhG;;;;;;;;;;;;qEAYqE;AACrE,MAAM,UAAU,GAAG,IAAI,MAAM,CAC3B;IACE,oCAAoC,CAAC,MAAM;IAC3C,oBAAoB,CAAC,MAAM;IAC3B,yCAAyC,CAAC,MAAM;IAChD,2BAA2B,CAAC,MAAM;IAClC,yDAAyD,CAAC,MAAM;IAChE,2BAA2B,CAAC,MAAM;CACnC,CAAC,IAAI,CAAC,GAAG,CAAC,CACZ,CAAA;AACD,oFAAoF;AACpF,MAAM,qBAAqB,GAAG,CAAC,CAAA;AAC/B,MAAM,0BAA0B,GAAG,GAAG,CAAA;AACtC,MAAM,qBAAqB,GAAG,4CAA4C,CAAA;AAC1E,MAAM,QAAQ,GAAG,iDAAiD,CAAA;AAElE;;;;;GAKG;AACH,SAAS,UAAU,CAAC,GAAW;IAC7B,IAAI,QAAQ,GAA6D,IAAI,CAAA;IAC7E,IAAI,SAAS,GAA2B,IAAI,CAAA;IAC5C,IAAI,QAAQ,GAA2B,IAAI,CAAA;IAC3C,KAAK,MAAM,IAAI,IAAI,GAAG,CAAC,KAAK,CAAC,IAAI,CAAC,EAAE,CAAC;QACnC,MAAM,KAAK,GAAG,qBAAqB,CAAC,IAAI,CAAC,IAAI,CAAC,CAAA;QAC9C,MAAM,IAAI,GAAG,QAAQ,CAAC,IAAI,CAAC,IAAI,CAAC,CAAA;QAChC,IAAI,KAAK,IAAI,IAAI;YAAE,QAAQ,GAAG,EAAE,KAAK,EAAE,IAAI,EAAE,CAAA;QAC7C,IAAI,KAAK;YAAE,SAAS,GAAG,KAAK,CAAA;QAC5B,IAAI,IAAI;YAAE,QAAQ,GAAG,IAAI,CAAA;IAC3B,CAAC;IACD,IAAI,QAAQ,KAAK,IAAI;QAAE,OAAO,QAAQ,CAAA;IACtC,OAAO,SAAS,KAAK,IAAI,CAAC,CAAC,CAAC,EAAE,KAAK,EAAE,SAAS,EAAE,IAAI,EAAE,IAAI,EAAE,CAAC,CAAC,CAAC,EAAE,KAAK,EAAE,IAAI,EAAE,IAAI,EAAE,QAAQ,EAAE,CAAA;AAChG,CAAC;AAED;;0DAE0D;AAC1D,MAAM,wBAAwB,GAAG,sDAAsD,CAAA;AACvF;;+DAE+D;AAC/D,MAAM,wBAAwB,GAAG,6BAA6B,CAAA;AAE9D,SAAS,KAAK,CAAC,MAAc,EAAE,IAAY;IACzC,IAAI,IAAI,KAAK,KAAK;QAAE,OAAO,MAAM,CAAA;IACjC,IAAI,IAAI,KAAK,KAAK;QAAE,OAAO,MAAM,GAAG,IAAI,CAAA;IACxC,IAAI,IAAI,KAAK,KAAK;QAAE,OAAO,MAAM,GAAG,CAAC,IAAI,GAAG,IAAI,CAAC,CAAA;IACjD,OAAO,MAAM,GAAG,GAAG,CAAA,CAAC,UAAU;AAChC,CAAC;AAED;;sFAEsF;AACtF,SAAS,UAAU,CAAC,GAAW;IAC7B,MAAM,KAAK,GAAa,EAAE,CAAA;IAC1B,KAAK,MAAM,GAAG,IAAI,GAAG,CAAC,KAAK,CAAC,IAAI,CAAC,EAAE,CAAC;QAClC,IAAI,CAAC,UAAU,CAAC,IAAI,CAAC,GAAG,CAAC;YAAE,SAAQ;QACnC,MAAM,OAAO,GAAG,GAAG,CAAC,IAAI,EAAE,CAAA;QAC1B,MAAM,IAAI,GACR,OAAO,CAAC,MAAM,GAAG,0BAA0B;YACzC,CAAC,CAAC,GAAG,OAAO,CAAC,KAAK,CAAC,CAAC,EAAE,0BAA0B,GAAG,CAAC,CAAC,GAAG;YACxD,CAAC,CAAC,OAAO,CAAA;QACb,IAAI,KAAK,CAAC,QAAQ,CAAC,IAAI,CAAC;YAAE,SAAQ;QAClC,KAAK,CAAC,IAAI,CAAC,IAAI,CAAC,CAAA;QAChB,IAAI,KAAK,CAAC,MAAM,KAAK,qBAAqB;YAAE,MAAK;IACnD,CAAC;IACD,OAAO,KAAK,CAAC,IAAI,CAAC,IAAI,CAAC,CAAA;AACzB,CAAC;AAED,SAAS,WAAW,CAAC,GAAW;IAC9B,IAAI,CAAC,UAAU,CAAC,IAAI,CAAC,GAAG,CAAC;QAAE,OAAO,IAAI,CAAA;IACtC,oGAAoG;IACpG,wFAAwF;IACxF,MAAM,EAAE,KAAK,EAAE,IAAI,EAAE,GAAG,UAAU,CAAC,GAAG,CAAC,CAAA;IACvC,MAAM,OAAO,GAA2B,EAAE,CAAA;IAC1C,IAAI,aAAa,GAAG,sBAAsB,CAAA;IAC1C,IAAI,QAAQ,GAAG,EAAE,CAAA;IACjB,IAAI,KAAK,EAAE,CAAC;QACV,MAAM,YAAY,GAAG,KAAK,CAAC,MAAM,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC,EAAE,KAAK,CAAC,CAAC,CAAW,CAAC,CAAA;QAChE,OAAO,CAAC,eAAe,CAAC,GAAG,YAAY,CAAA;QACvC,aAAa,GAAG,GAAG,YAAY,MAAM,CAAA;IACvC,CAAC;IACD,IAAI,IAAI,EAAE,CAAC;QACT,MAAM,OAAO,GAAG,KAAK,CAAC,MAAM,CAAC,IAAI,CAAC,CAAC,CAAC,CAAC,EAAE,IAAI,CAAC,CAAC,CAAW,CAAC,CAAA;QACzD,OAAO,CAAC,UAAU,CAAC,GAAG,OAAO,CAAA;QAC7B,QAAQ,GAAG,KAAK,OAAO,YAAY,CAAA;IACrC,CAAC;IACD,OAAO;QACL,IAAI,EAAE,eAAe;QACrB,OAAO,EAAE,6DAA6D,aAAa,UAAU,QAAQ,GAAG;QACxG,GAAG,CAAC,MAAM,CAAC,IAAI,CAAC,OAAO,CAAC,CAAC,MAAM,GAAG,CAAC,CAAC,CAAC,CAAC,EAAE,OAAO,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC;QACvD,OAAO,EAAE,UAAU,CAAC,GAAG,CAAC;KACzB,CAAA;AACH,CAAC;AAED,SAAS,+BAA+B,CAAC,GAAW;IAClD,MAAM,IAAI,GAAG,wBAAwB,CAAC,IAAI,CAAC,GAAG,CAAC,CAAA;IAC/C,OAAO,IAAI;QACT,CAAC,CAAC;YACE,IAAI,EAAE,mBAAmB;YACzB,OAAO,EAAE,0BAA0B,IAAI,CAAC,CAAC,CAAC,iDAAiD;SAC5F;QACH,CAAC,CAAC,IAAI,CAAA;AACV,CAAC;AAED,SAAS,+BAA+B,CAAC,GAAW;IAClD,OAAO,wBAAwB,CAAC,IAAI,CAAC,GAAG,CAAC;QACvC,CAAC,CAAC;YACE,IAAI,EAAE,mBAAmB;YACzB,OAAO,EAAE,8EAA8E;SACxF;QACH,CAAC,CAAC,IAAI,CAAA;AACV,CAAC;AAED;;;;;;;;;mDASmD;AACnD,kIAAkI;AAClI,MAAM,gBAAgB,GAAG,qCAAqC,CAAA;AAE9D;;;;;;;GAOG;AACH,MAAM,UAAU,wBAAwB,CACtC,GAAW,EACX,QAA6B;IAE7B,MAAM,KAAK,GAAG,CAAC,GAAG,GAAG,CAAC,QAAQ,CAAC,gBAAgB,CAAC,CAAC,CAAA;IACjD,MAAM,IAAI,GAAG,KAAK,CAAC,EAAE,CAAC,CAAC,CAAC,CAAC,CAAA;IACzB,IAAI,IAAI,KAAK,SAAS;QAAE,OAAO,IAAI,CAAA;IACnC,MAAM,QAAQ,GAAG,MAAM,CAAC,QAAQ,CAAC,IAAI,CAAC,CAAC,CAAW,EAAE,EAAE,CAAC,CAAA;IACvD,MAAM,OAAO,GAAG,QAAQ,CAAC,cAAc,CAAA;IACvC,IAAI,CAAC,MAAM,CAAC,QAAQ,CAAC,QAAQ,CAAC,IAAI,QAAQ,IAAI,OAAO;QAAE,OAAO,IAAI,CAAA;IAClE,OAAO;QACL,IAAI,EAAE,eAAe;QACrB,OAAO,EACL,kEAAkE,QAAQ,yBAAyB;YACnG,qBAAqB,OAAO,oEAAoE;YAChG,6BAA6B;QAC/B,OAAO,EAAE,EAAE,kBAAkB,EAAE,QAAQ,EAAE,cAAc,EAAE,OAAO,EAAE;QAClE,OAAO,EAAE,IAAI,CAAC,CAAC,CAAC,CAAC,IAAI,EAAE;KACxB,CAAA;AACH,CAAC;AAED,+GAA+G;AAC/G,MAAM,kBAAkB,GAAG,kCAAkC,CAAA;AAE7D;;;;;GAKG;AACH,MAAM,UAAU,gCAAgC,CAAC,OAAe;IAC9D,MAAM,IAAI,GAAG,CAAC,GAAG,OAAO,CAAC,QAAQ,CAAC,kBAAkB,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,CAAC,CAAA;IAC7D,OAAO,IAAI,KAAK,SAAS,CAAC,CAAC,CAAC,IAAI,CAAC,CAAC,CAAC,oCAAoC,IAAI,CAAC,CAAC,CAAW,EAAE,CAAA;AAC5F,CAAC;AAED,MAAM,UAAU,uBAAuB,CAAC,GAAW,EAAE,QAAuB;IAC1E,MAAM,cAAc,GAClB,+BAA+B,CAAC,GAAG,CAAC,IAAI,WAAW,CAAC,GAAG,CAAC,IAAI,+BAA+B,CAAC,GAAG,CAAC,CAAA;IAClG,IAAI,cAAc;QAAE,OAAO,cAAc,CAAA;IACzC,MAAM,QAAQ,GAAG,QAAQ,KAAK,IAAI,CAAC,CAAC,CAAC,mBAAmB,CAAC,CAAC,CAAC,kBAAkB,QAAQ,EAAE,CAAA;IACvF,OAAO,EAAE,IAAI,EAAE,OAAO,EAAE,OAAO,EAAE,oCAAoC,QAAQ,0BAA0B,EAAE,CAAA;AAC3G,CAAC;AAED,gGAAgG;AAChG,eAAe;AACf,gGAAgG;AAEhG;;;;;;2FAM2F;AAC3F,MAAM,UAAU,uBAAuB,CAAC,OAGvC;IACC,MAAM,EAAE,MAAM,EAAE,GAAG,OAAO,CAAA;IAC1B,OAAO;QACL,KAAK,EAAE,MAAM,EAAE,WAAW,IAAI,IAAI;QAClC,SAAS,EAAE,MAAM,EAAE,gBAAgB,IAAI,IAAI;QAC3C,iBAAiB,EAAE,MAAM,EAAE,iBAAiB,IAAI,KAAK;QACrD,MAAM,EAAE,KAAK;QACb,UAAU,EAAE,KAAK;QACjB,SAAS,EAAE,KAAK;KACjB,CAAA;AACH,CAAC;AAED,gGAAgG;AAChG,kBAAkB;AAClB,gGAAgG;AAEhG;;;;;;;;;;;;;;oEAcoE;AACpE,MAAM,CAAC,MAAM,mBAAmB,GAA4B;IAC1D,EAAE,MAAM,EAAE,MAAM,EAAE,IAAI,EAAE,sBAAsB,EAAE;IAChD,EAAE,MAAM,EAAE,MAAM,EAAE,IAAI,EAAE,iBAAiB,EAAE;IAC3C,EAAE,MAAM,EAAE,KAAK,EAAE,IAAI,EAAE,YAAY,EAAE;CACtC,CAAA;AAED;sFACsF;AACtF,MAAM,CAAC,MAAM,8BAA8B,GAA4B;IACrE,EAAE,MAAM,EAAE,MAAM,EAAE,IAAI,EAAE,sBAAsB,EAAE;IAChD,EAAE,MAAM,EAAE,MAAM,EAAE,IAAI,EAAE,iBAAiB,EAAE;CAC5C,CAAA;AAmBD,MAAM,GAAG,GAAG,aAAa,CAAA;AACzB;;;;oCAIoC;AACpC,MAAM,0BAA0B,GAAG,IAAI,MAAM,CAC3C,2BAA2B,GAAG,wBAAwB,GAAG,2CAA2C,GAAG,KAAK,CAC7G,CAAA;AACD,qGAAqG;AACrG,MAAM,iCAAiC,GAAG,IAAI,MAAM,CAClD,wBAAwB,GAAG,2CAA2C,GAAG,KAAK,CAC/E,CAAA;AACD;;;;;;gEAMgE;AAChE,MAAM,8BAA8B,GAAG,IAAI,MAAM,CAC/C,0BAA0B;IACxB,GAAG;IACH,sDAAsD;IACtD,GAAG;IACH,uBAAuB;IACvB,GAAG;IACH,gCAAgC;IAChC,GAAG;IACH,6BAA6B;IAC7B,GAAG;IACH,KAAK,CACR,CAAA;AACD,iGAAiG;AACjG,MAAM,uCAAuC,GAAG,IAAI,MAAM,CACxD,0BAA0B;IACxB,GAAG;IACH,sDAAsD;IACtD,GAAG;IACH,uBAAuB;IACvB,GAAG;IACH,gCAAgC;IAChC,GAAG;IACH,KAAK,CACR,CAAA;AAED;;;;eAIe;AACf,SAAS,SAAS,CAAC,IAAY;IAC7B,IAAI,CAAC;QACH,MAAM,MAAM,GAAG,IAAI,CAAC,KAAK,CAAC,IAAI,CAAY,CAAA;QAC1C,IACE,MAAM,KAAK,IAAI;YACf,OAAO,MAAM,KAAK,QAAQ;YAC1B,OAAQ,MAAgC,CAAC,OAAO,KAAK,QAAQ,EAC7D,CAAC;YACD,OAAQ,MAA8B,CAAC,OAAO,CAAA;QAChD,CAAC;IACH,CAAC;IAAC,MAAM,CAAC;QACP,qEAAqE;IACvE,CAAC;IACD,OAAO,IAAI,CAAA;AACb,CAAC;AAED,SAAS,qBAAqB,CAAC,eAAuB,EAAE,SAAiB;IACvE,MAAM,SAAS,GAAG,IAAI,CAAC,KAAK,CAAC,eAAe,CAAC,CAAA;IAC7C,MAAM,KAAK,GAAG,IAAI,CAAC,KAAK,CAAC,SAAS,CAAC,CAAA;IACnC,OAAO;QACL,KAAK,EAAE;YACL,OAAO,EACL,0CAA0C,KAAK,8CAA8C;gBAC7F,GAAG,SAAS,oDAAoD;YAClE,IAAI,EAAE,uBAAuB;YAC7B,KAAK,EAAE,IAAI;YACX,IAAI,EAAE,yBAAyB;SAChC;KACF,CAAA;AACH,CAAC;AAED;;;;;;;GAOG;AACH,MAAM,UAAU,gCAAgC,CAC9C,MAAc,EACd,IAAY;IAEZ,IAAI,MAAM,KAAK,GAAG;QAAE,OAAO,IAAI,CAAA;IAC/B,MAAM,OAAO,GAAG,SAAS,CAAC,IAAI,CAAC,CAAA;IAE/B,MAAM,eAAe,GAAG,0BAA0B,CAAC,IAAI,CAAC,OAAO,CAAC,CAAA;IAChE,IAAI,eAAe,EAAE,CAAC;QACpB,MAAM,CAAC,EAAE,SAAS,EAAE,QAAQ,EAAE,KAAK,CAAC,GAAG,eAA8D,CAAA;QACrG,OAAO,qBAAqB,CAAC,MAAM,CAAC,SAAS,CAAC,GAAG,MAAM,CAAC,QAAQ,CAAC,EAAE,MAAM,CAAC,KAAK,CAAC,CAAC,CAAA;IACnF,CAAC;IAED,MAAM,kBAAkB,GAAG,8BAA8B,CAAC,IAAI,CAAC,OAAO,CAAC,CAAA;IACvE,IAAI,kBAAkB,EAAE,CAAC;QACvB,MAAM,CAAC,EAAE,AAAD,EAAG,AAAD,EAAG,SAAS,EAAE,gBAAgB,EAAE,aAAa,CAAC,GAAG,kBAO1D,CAAA;QACD,OAAO,qBAAqB,CAAC,MAAM,CAAC,gBAAgB,CAAC,GAAG,MAAM,CAAC,aAAa,CAAC,EAAE,MAAM,CAAC,SAAS,CAAC,CAAC,CAAA;IACnG,CAAC;IAED,MAAM,uBAAuB,GAAG,iCAAiC,CAAC,IAAI,CAAC,OAAO,CAAC,CAAA;IAC/E,IAAI,uBAAuB,EAAE,CAAC;QAC5B,MAAM,CAAC,EAAE,SAAS,EAAE,KAAK,CAAC,GAAG,uBAA8D,CAAA;QAC3F,OAAO,qBAAqB,CAAC,MAAM,CAAC,SAAS,CAAC,EAAE,MAAM,CAAC,KAAK,CAAC,CAAC,CAAA;IAChE,CAAC;IAED,MAAM,gBAAgB,GAAG,uCAAuC,CAAC,IAAI,CAAC,OAAO,CAAC,CAAA;IAC9E,IAAI,gBAAgB,EAAE,CAAC;QACrB,MAAM,CAAC,EAAE,AAAD,EAAG,AAAD,EAAG,SAAS,EAAE,gBAAgB,CAAC,GAAG,gBAM3C,CAAA;QACD,OAAO,qBAAqB,CAAC,MAAM,CAAC,gBAAgB,CAAC,EAAE,MAAM,CAAC,SAAS,CAAC,CAAC,CAAA;IAC3E,CAAC;IAED,OAAO,IAAI,CAAA;AACb,CAAC;AAED,gGAAgG;AAChG,8DAA8D;AAC9D,gGAAgG;AAEhG;;;;4BAI4B;AAC5B,MAAM,iBAAiB,GAAG,IAAI,GAAG,CAAC,8BAA8B,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,IAAI,CAAC,CAAC,CAAA;AAapF,SAAS,kBAAkB,CAAC,IAA6B,EAAE,GAAW;IACpE,IAAI,CAAC,CAAC,GAAG,IAAI,IAAI,CAAC,IAAI,IAAI,CAAC,GAAG,CAAC,KAAK,IAAI;QAAE,OAAO,EAAE,OAAO,EAAE,KAAK,EAAE,KAAK,EAAE,KAAK,EAAE,KAAK,EAAE,CAAC,EAAE,CAAA;IAC3F,MAAM,KAAK,GAAG,IAAI,CAAC,GAAG,CAAC,CAAA;IACvB,MAAM,KAAK,GAAG,OAAO,KAAK,KAAK,QAAQ,IAAI,MAAM,CAAC,SAAS,CAAC,KAAK,CAAC,IAAI,KAAK,GAAG,CAAC,CAAA;IAC/E,OAAO,EAAE,OAAO,EAAE,IAAI,EAAE,KAAK,EAAE,KAAK,EAAE,KAAK,CAAC,CAAC,CAAE,KAAgB,CAAC,CAAC,CAAC,CAAC,EAAE,CAAA;AACvE,CAAC;AAED;;;;;;;GAOG;AACH,SAAS,iBAAiB,CAAC,IAA6B,EAAE,YAAqC;IAC7F,MAAM,KAAK,GAAG,OAAO,IAAI,CAAC,OAAO,CAAC,KAAK,QAAQ,CAAC,CAAC,CAAC,cAAc,IAAI,CAAC,OAAO,CAAC,GAAG,CAAC,CAAC,CAAC,YAAY,CAAA;IAC/F,IAAI,CAAC,YAAY,CAAC,KAAK,IAAI,YAAY,CAAC,IAAI,CAAC,EAAE,CAAC;QAC9C,MAAM,IAAI,qBAAqB,CAAC,GAAG,KAAK,iCAAiC,EAAE,wBAAwB,CAAC,CAAA;IACtG,CAAC;IACD,IAAI,CAAC,YAAY,CAAC,iBAAiB,IAAI,uBAAuB,CAAC,IAAI,CAAC,EAAE,CAAC;QACrE,MAAM,IAAI,qBAAqB,CAAC,GAAG,KAAK,sCAAsC,EAAE,wBAAwB,CAAC,CAAA;IAC3G,CAAC;AACH,CAAC;AAED;;;gGAGgG;AAChG,SAAS,gBAAgB,CAAC,KAAa;IACrC,MAAM,IAAI,eAAe,CAAC,kBAAkB,EAAE,GAAG,KAAK,8BAA8B,CAAC,CAAA;AACvF,CAAC;AAED,SAAS,aAAa,CAAC,KAAc;IACnC,OAAO,KAAK,KAAK,IAAI,IAAI,OAAO,KAAK,KAAK,QAAQ,IAAI,CAAC,KAAK,CAAC,OAAO,CAAC,KAAK,CAAC,CAAA;AAC7E,CAAC;AAED,iGAAiG;AACjG,SAAS,iBAAiB;IACxB,MAAM,IAAI,eAAe,CACvB,kBAAkB,EAClB,2FAA2F,CAC5F,CAAA;AACH,CAAC;AAED,kGAAkG;AAClG,SAAS,wBAAwB;IAC/B,MAAM,IAAI,eAAe,CACvB,kBAAkB,EAClB,+GAA+G,CAChH,CAAA;AACH,CAAC;AAED;;;;;;;;;;;;GAYG;AACH,SAAS,sBAAsB,CAAC,MAAe;IAC7C,IAAI,CAAC,aAAa,CAAC,MAAM,CAAC,IAAI,MAAM,CAAC,MAAM,CAAC,KAAK,aAAa;QAAE,OAAO,MAAM,CAAA;IAC7E,MAAM,OAAO,GAAG,MAAM,CAAC,aAAa,CAAC,CAAA;IACrC,IAAI,CAAC,aAAa,CAAC,OAAO,CAAC;QAAE,iBAAiB,EAAE,CAAA;IAChD,MAAM,KAAK,GAAG,OAAO,CAAC,QAAQ,CAAC,CAAA;IAC/B,IAAI,CAAC,aAAa,CAAC,KAAK,CAAC,EAAE,CAAC;QAC1B,IAAI,MAAM,IAAI,OAAO,IAAI,QAAQ,IAAI,OAAO;YAAE,wBAAwB,EAAE,CAAA;QACxE,OAAO,MAAM,CAAA;IACf,CAAC;IACD,OAAO,EAAE,GAAG,MAAM,EAAE,WAAW,EAAE,KAAK,EAAE,CAAA;AAC1C,CAAC;AAED;;;;;;;GAOG;AACH,MAAM,CAAC,MAAM,uCAAuC,GAAwB,IAAI,GAAG,CAAC,CAAC,SAAS,CAAC,CAAC,CAAA;AAEhG,8GAA8G;AAC9G,MAAM,CAAC,MAAM,+BAA+B,GAAG,oBAAoB,CAAA;AAEnE,gGAAgG;AAChG,MAAM,UAAU,6BAA6B,CAC3C,KAAa,EACb,WAAoB,EACpB,MAAiC;IAEjC,IAAI,KAAK,KAAK,sBAAsB;QAAE,OAAO,IAAI,CAAA;IACjD,MAAM,MAAM,GAAG,MAAM,EAAE,gBAAgB,IAAI,IAAI,CAAA;IAC/C,IAAI,MAAM,KAAK,IAAI,IAAI,CAAC,uCAAuC,CAAC,GAAG,CAAC,MAAM,CAAC;QAAE,OAAO,IAAI,CAAA;IACxF,OAAO,iBAAiB,CAAC,WAAW,CAAC,CAAC,CAAC,CAAC,IAAI,CAAC,CAAC,CAAC,oBAAoB,CAAA;AACrE,CAAC;AAED;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;GAsCG;AACH,MAAM,UAAU,6BAA6B,CAC3C,KAAa,EACb,IAAa,EACb,QAA6B,EAC7B,YAAsC;IAEtC,IAAI,CAAC,iBAAiB,CAAC,GAAG,CAAC,KAAK,CAAC;QAAE,OAAO,IAAI,CAAA;IAC9C,IAAI,OAAO,IAAI,KAAK,QAAQ,IAAI,IAAI,KAAK,IAAI,IAAI,KAAK,CAAC,OAAO,CAAC,IAAI,CAAC;QAAE,OAAO,IAAI,CAAA;IACjF,MAAM,MAAM,GAAG,IAA+B,CAAA;IAC9C,IAAI,YAAY,KAAK,SAAS;QAAE,iBAAiB,CAAC,MAAM,EAAE,YAAY,CAAC,CAAA;IACvE,IAAI,GAAG,GACL,iBAAiB,IAAI,MAAM;QACzB,CAAC,CAAC,EAAE,GAAG,MAAM,EAAE,eAAe,EAAE,sBAAsB,CAAC,MAAM,CAAC,iBAAiB,CAAC,CAAC,EAAE;QACnF,CAAC,CAAC,MAAM,CAAA;IACZ,MAAM,GAAG,GAAG,QAAQ,CAAC,iBAAiB,CAAA;IACtC,oGAAoG;IACpG,GAAG,GAAG,sBAAsB,CAAC,GAAG,EAAE,QAAQ,CAAC,UAAU,CAAC,CAAA;IAEtD,IAAI,KAAK,KAAK,iBAAiB,EAAE,CAAC;QAChC,MAAM,KAAK,GAAG,kBAAkB,CAAC,GAAG,EAAE,YAAY,CAAC,CAAA;QACnD,IAAI,KAAK,CAAC,OAAO,IAAI,CAAC,KAAK,CAAC,KAAK;YAAE,gBAAgB,CAAC,YAAY,CAAC,CAAA;QACjE,OAAO,EAAE,GAAG,GAAG,EAAE,UAAU,EAAE,KAAK,CAAC,OAAO,CAAC,CAAC,CAAC,IAAI,CAAC,GAAG,CAAC,KAAK,CAAC,KAAK,EAAE,GAAG,CAAC,CAAC,CAAC,CAAC,GAAG,EAAE,CAAA;IACjF,CAAC;IAED,uBAAuB;IACvB,MAAM,MAAM,GAAG,kBAAkB,CAAC,GAAG,EAAE,YAAY,CAAC,CAAA;IACpD,MAAM,MAAM,GAAG,kBAAkB,CAAC,GAAG,EAAE,uBAAuB,CAAC,CAAA;IAC/D,IAAI,CAAC,MAAM,CAAC,OAAO,IAAI,CAAC,MAAM,CAAC,KAAK,CAAC,IAAI,CAAC,MAAM,CAAC,OAAO,IAAI,CAAC,MAAM,CAAC,KAAK,CAAC,EAAE,CAAC;QAC3E,gBAAgB,CAAC,kCAAkC,CAAC,CAAA;IACtD,CAAC;IACD,MAAM,MAAM,GAAa,EAAE,CAAA;IAC3B,IAAI,MAAM,CAAC,KAAK;QAAE,MAAM,CAAC,IAAI,CAAC,MAAM,CAAC,KAAK,CAAC,CAAA;IAC3C,IAAI,MAAM,CAAC,KAAK;QAAE,MAAM,CAAC,IAAI,CAAC,MAAM,CAAC,KAAK,CAAC,CAAA;IAC3C,MAAM,KAAK,GAAG,MAAM,CAAC,MAAM,GAAG,CAAC,CAAC,CAAC,CAAC,IAAI,CAAC,GAAG,CAAC,GAAG,EAAE,GAAG,MAAM,CAAC,CAAC,CAAC,CAAC,GAAG,CAAA;IAChE,MAAM,MAAM,GAAG,MAAM,CAAC,OAAO,CAAC,CAAC,CAAC,uBAAuB,CAAC,CAAC,CAAC,YAAY,CAAA;IACtE,MAAM,IAAI,GAA4B,EAAE,GAAG,GAAG,EAAE,CAAA;IAChD,OAAO,IAAI,CAAC,YAAY,CAAC,CAAA;IACzB,OAAO,IAAI,CAAC,uBAAuB,CAAC,CAAA;IACpC,IAAI,CAAC,MAAM,CAAC,GAAG,KAAK,CAAA;IACpB,OAAO,IAAI,CAAA;AACb,CAAC;AAED,gGAAgG;AAChG,cAAc;AACd,gGAAgG;AAEhG;;;;;;;;;GASG;AACH,MAAM,aAAa,GAAG;IACpB,EAAE,KAAK,EAAE,qBAA8B,EAAE,OAAO,EAAE,iDAAiD,EAAE;IACrG,EAAE,KAAK,EAAE,qBAA8B,EAAE,OAAO,EAAE,uCAAuC,EAAE;IAC3F,EAAE,KAAK,EAAE,qBAA8B,EAAE,OAAO,EAAE,+BAA+B,EAAE;IACnF,gHAAgH;IAChH;QACE,KAAK,EAAE,qBAA8B;QACrC,OAAO,EAAE,2DAA2D;KACrE;CACF,CAAA;AAED;0GAC0G;AAC1G,MAAM,CAAC,MAAM,kBAAkB,GAA4C;IACzE,EAAE,EAAE,cAAc;IAClB,eAAe,EAAE,qCAAqC;IACtD,SAAS,EAAE,EAAE,IAAI,EAAE,SAAS,EAAE,cAAc,EAAE,GAAG,EAAE;IACnD,mGAAmG;IACnG,kGAAkG;IAClG,eAAe,EAAE;QACf,UAAU,EAAE,YAAY;QACxB,IAAI,EAAE,sBAAsB;QAC5B,IAAI,EAAE,EAAE,QAAQ,EAAE,CAAC,EAAE,IAAI,EAAE,MAAM,EAAE,OAAO,EAAE,IAAI,EAAE,CAAC,EAAE,UAAU,EAAE,CAAC,EAAE,MAAM,EAAE,KAAK,EAAE;KACpF;IACD,MAAM,EAAE,mBAAmB;IAC3B,gBAAgB,EAAE,8BAA8B;IAChD,YAAY,EAAE,aAAa;IAC3B,gBAAgB,EAAE,2BAA2B;IAC7C,WAAW,EAAE,sBAAsB;IACnC,kBAAkB,EAAE,6BAA6B;IACjD,YAAY,EAAE,uBAAuB;IACrC,aAAa,EAAE,wBAAwB;IACvC,qBAAqB,EAAE,gCAAgC;IACvD,YAAY,EAAE,uBAAuB;IACrC,kBAAkB,EAAE,6BAA6B;IACjD,kBAAkB,EAAE,6BAA6B;IACjD,kFAAkF;IAClF,uFAAuF;IACvF,gBAAgB,EAAE,CAAC,MAAM,EAAE,MAAM,EAAE,IAAI,EAAE,EAAE,CAAC,gCAAgC,CAAC,MAAM,EAAE,IAAI,CAAC;IAC1F,mGAAmG;IACnG,kGAAkG;IAClG,gFAAgF;IAChF,UAAU,EAAE,CAAC,QAAQ,EAAE,EAAE,CAAC;QACxB,QAAQ,CAAC,cAAc;QACvB,QAAQ,CAAC,iCAAiC;QAC1C,QAAQ,CAAC,cAAc;QACvB,QAAQ,CAAC,mBAAmB;QAC5B,QAAQ,CAAC,WAAW;QACpB,QAAQ,CAAC,cAAc;QACvB,QAAQ,CAAC,qBAAqB;QAC9B,QAAQ,CAAC,iBAAiB;QAC1B,QAAQ,CAAC,yBAAyB;QAClC,QAAQ,CAAC,KAAK;KACf;CACF,CAAA"}
|
|
@@ -0,0 +1,12 @@
|
|
|
1
|
+
/**
|
|
2
|
+
* `POST /atomic/v1/models/tensorrt-llm/check` (task 2.16 of change `add-tensorrt-llm-linux`, spec
|
|
3
|
+
* `tensorrt-llm-models`): the shared managed check route (`../managed-models/check.ts`) with
|
|
4
|
+
* TensorRT-LLM's settings — its saved card and its `kv_cache_free_gpu_memory_fraction`, which size
|
|
5
|
+
* its memory rule (`tensorrtLlmCheckEngine`).
|
|
6
|
+
*/
|
|
7
|
+
import type { ModelCompatibility } from '../../contracts/index.js';
|
|
8
|
+
import { type ManagedModelCheckEngine, type ModelCheckDeps } from '../managed-models/check.js';
|
|
9
|
+
export { parseModelCheckInput, type ModelCheckDeps, type ModelCheckHostFacts, } from '../managed-models/check.js';
|
|
10
|
+
/** TensorRT-LLM's side of the check route: its card and its memory rule from its stored settings. */
|
|
11
|
+
export declare const TENSORRT_LLM_MODEL_CHECK: ManagedModelCheckEngine;
|
|
12
|
+
export declare function checkTensorrtLlmModel(body: unknown, deps: ModelCheckDeps): Promise<ModelCompatibility>;
|
|
@@ -0,0 +1,22 @@
|
|
|
1
|
+
import { TENSORRT_LLM_ENGINE_ID } from '../environment/index.js';
|
|
2
|
+
import { checkManagedModel, } from '../managed-models/check.js';
|
|
3
|
+
import { tensorrtLlmCheckEngine } from './compatibility.js';
|
|
4
|
+
import { tensorrtLlmSettings } from './settings.js';
|
|
5
|
+
export { parseModelCheckInput, } from '../managed-models/check.js';
|
|
6
|
+
/** TensorRT-LLM's side of the check route: its card and its memory rule from its stored settings. */
|
|
7
|
+
export const TENSORRT_LLM_MODEL_CHECK = {
|
|
8
|
+
engineId: TENSORRT_LLM_ENGINE_ID,
|
|
9
|
+
gpuIdOf: (settings) => tensorrtLlmSettings(settings).gpu_id ?? null,
|
|
10
|
+
checkEngineOf: (settings) => {
|
|
11
|
+
const resolved = tensorrtLlmSettings(settings);
|
|
12
|
+
return tensorrtLlmCheckEngine({
|
|
13
|
+
contextLength: resolved.context_length,
|
|
14
|
+
kvCacheFreeGpuMemoryFraction: resolved.kv_cache_free_gpu_memory_fraction,
|
|
15
|
+
cudaGraphs: resolved.cuda_graphs,
|
|
16
|
+
});
|
|
17
|
+
},
|
|
18
|
+
};
|
|
19
|
+
export function checkTensorrtLlmModel(body, deps) {
|
|
20
|
+
return checkManagedModel(TENSORRT_LLM_MODEL_CHECK, body, deps);
|
|
21
|
+
}
|
|
22
|
+
//# sourceMappingURL=check.js.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"check.js","sourceRoot":"","sources":["../../../src/runtime/tensorrt-llm/check.ts"],"names":[],"mappings":"AAOA,OAAO,EAAE,sBAAsB,EAAE,MAAM,yBAAyB,CAAA;AAChE,OAAO,EACL,iBAAiB,GAGlB,MAAM,4BAA4B,CAAA;AACnC,OAAO,EAAE,sBAAsB,EAAE,MAAM,oBAAoB,CAAA;AAC3D,OAAO,EAAE,mBAAmB,EAAE,MAAM,eAAe,CAAA;AAEnD,OAAO,EACL,oBAAoB,GAGrB,MAAM,4BAA4B,CAAA;AAEnC,qGAAqG;AACrG,MAAM,CAAC,MAAM,wBAAwB,GAA4B;IAC/D,QAAQ,EAAE,sBAAsB;IAChC,OAAO,EAAE,CAAC,QAAQ,EAAE,EAAE,CAAC,mBAAmB,CAAC,QAAQ,CAAC,CAAC,MAAM,IAAI,IAAI;IACnE,aAAa,EAAE,CAAC,QAAQ,EAAE,EAAE;QAC1B,MAAM,QAAQ,GAAG,mBAAmB,CAAC,QAAQ,CAAC,CAAA;QAC9C,OAAO,sBAAsB,CAAC;YAC5B,aAAa,EAAE,QAAQ,CAAC,cAAc;YACtC,4BAA4B,EAAE,QAAQ,CAAC,iCAAiC;YACxE,UAAU,EAAE,QAAQ,CAAC,WAAW;SACjC,CAAC,CAAA;IACJ,CAAC;CACF,CAAA;AAED,MAAM,UAAU,qBAAqB,CAAC,IAAa,EAAE,IAAoB;IACvE,OAAO,iBAAiB,CAAC,wBAAwB,EAAE,IAAI,EAAE,IAAI,CAAC,CAAA;AAChE,CAAC"}
|
|
@@ -0,0 +1,124 @@
|
|
|
1
|
+
/**
|
|
2
|
+
* TensorRT-LLM's part of the managed compatibility check (spec `tensorrt-llm-models`, "Проверка
|
|
3
|
+
* совместимости без сети"; change `add-vllm-runtime`, design D7). The skeleton — GGUF, the curated
|
|
4
|
+
* digest, the shared naming rule, architecture, format matrix, compute capability, the memory line —
|
|
5
|
+
* is `../managed-models/compatibility.ts`, the same for every managed engine. What is TensorRT-LLM's
|
|
6
|
+
* own lives here as the two hooks of `tensorrtLlmCheckEngine`:
|
|
7
|
+
*
|
|
8
|
+
* - `memoryNeed`: weights plus an allowance for the engine's own layout of them, plus the KV
|
|
9
|
+
* reserve divided by `kv_cache_free_gpu_memory_fraction` (`trtllm-serve`'s semantics: it spends
|
|
10
|
+
* only that fraction of what is free after the weights), bounded by tokens on a unified-memory
|
|
11
|
+
* card, plus the runtime and the activation peak;
|
|
12
|
+
* - `checkpointProblems`: shapes of supported architectures this image's loader cannot read
|
|
13
|
+
* (Nemotron-H with dense MLP layers, a DeepseekV3Gate MoE without its correction bias).
|
|
14
|
+
*
|
|
15
|
+
* `checkModelCompatibility`, `checkModelCompatibilityFiles` and `checkModelMemory` keep their
|
|
16
|
+
* signatures — `MemorySizingInputs` is the provider's settings — and are the skeleton with this
|
|
17
|
+
* engine plugged in. The skeleton's helpers are re-exported so this module's callers are unchanged.
|
|
18
|
+
*/
|
|
19
|
+
import type { GpuFacts, ModelCompatibility, RuntimeDescriptor } from '../../contracts/index.js';
|
|
20
|
+
import { type CheckpointMemoryInputs, type CheckpointProblem, type FilesCheckResult, type HostMemory, type KvReserveBasis, type ManagedCheckEngine, type MemoryNeed, type ModelCheckInput, type ResolvedCheckpoint } from '../managed-models/compatibility.js';
|
|
21
|
+
import type { JsonObject } from '../managed-models/quant-format.js';
|
|
22
|
+
import { type TensorrtLlmCudaGraphsSetting, type TensorrtLlmLaunchPlan } from './cuda-graphs.js';
|
|
23
|
+
export * from '../managed-models/compatibility.js';
|
|
24
|
+
/**
|
|
25
|
+
* A checkpoint whose architecture is supported but whose own shape the engine image cannot load —
|
|
26
|
+
* what the architecture name alone cannot tell (Windows acceptance machine, 2026-10-03):
|
|
27
|
+
*
|
|
28
|
+
* - Nemotron-H with dense MLP layers: a `-` in `hybrid_override_pattern`. The `transformers` 5.5.4
|
|
29
|
+
* in the 1.3.0rc29 image maps that pattern through `M`/`E`/`*` only and fails with
|
|
30
|
+
* `KeyError: '-'` before TensorRT-LLM (which does support `-`) is reached
|
|
31
|
+
* (NVIDIA-Nemotron-3-Nano-4B, Nano-9B-v2). Drop this once a descriptor's image ships a
|
|
32
|
+
* `transformers` that knows `-`.
|
|
33
|
+
* - A `DeepseekV3Gate` architecture with MoE layers (`n_routed_experts > 0`) whose tensors have no
|
|
34
|
+
* `e_score_correction_bias` (PrimeIntellect/GLM-0.5B, saved in its training framework's own
|
|
35
|
+
* layout — `mlp.router.gate.weight`, `mlp.expert_bias` — not the Hugging Face one TensorRT-LLM
|
|
36
|
+
* reads): `AssertionError` while the weights load. Needs `weight_names`; without them it is not
|
|
37
|
+
* checked.
|
|
38
|
+
*
|
|
39
|
+
* `null` when none applies.
|
|
40
|
+
*/
|
|
41
|
+
export declare function tensorrtLlmCheckpointProblems(architecture: string, configJson: JsonObject, weightNames: readonly string[] | undefined): CheckpointProblem | null;
|
|
42
|
+
export interface MemoryReserve {
|
|
43
|
+
reserveBytes: number;
|
|
44
|
+
basis: KvReserveBasis;
|
|
45
|
+
}
|
|
46
|
+
/**
|
|
47
|
+
* Bytes reserved on top of the weights when checking whether a checkpoint fits a card's free memory.
|
|
48
|
+
*
|
|
49
|
+
* On a discrete card, `KV_bytes / kv_cache_free_gpu_memory_fraction` (task 2.16w round 1, finding 6
|
|
50
|
+
* (RULING), superseding the placeholder `weights × (1 − fraction)` rule): `trtllm-serve` spends only
|
|
51
|
+
* that fraction of whatever memory remains free *after* weights load on the KV cache, keeping the
|
|
52
|
+
* rest as headroom, so guaranteeing `KV_bytes` of real cache capacity needs `KV_bytes / fraction` of
|
|
53
|
+
* free memory left over once weights are loaded — the ADR this formula documents derives that
|
|
54
|
+
* division in full.
|
|
55
|
+
*
|
|
56
|
+
* On a unified-memory card (`unifiedMemory`, design D13) the launch bounds the KV cache by tokens,
|
|
57
|
+
* `kv_cache_config.max_tokens = tensorrtLlmUnifiedKvMaxTokens(contextLength)` (`kv-cache.ts`), and
|
|
58
|
+
* TensorRT-LLM uses the smaller of that bound and the fraction; the reserve is therefore the KV for
|
|
59
|
+
* exactly that many tokens, not divided by the fraction, so the check and the launch agree
|
|
60
|
+
* (docs/decisions/2026-09-29-tensorrt-llm-unified-memory-kv-cache-bounded-by-tokens.md).
|
|
61
|
+
*
|
|
62
|
+
* `basis: 'config'` when `kvCacheBytes` had what it needed; `basis: 'weight_fraction'` — the older,
|
|
63
|
+
* cruder `weights × (1 − fraction)` rule, the same on either kind of card — only when `config.json`
|
|
64
|
+
* lacked the architecture fields the real formula needs, and the returned `basis` says so, so a
|
|
65
|
+
* verdict computed from the fallback is never silently indistinguishable from one computed from the
|
|
66
|
+
* checkpoint's real shape.
|
|
67
|
+
*/
|
|
68
|
+
export declare function kvCacheReserveBytes(weightBytesTotal: number, configJson: JsonObject, hfQuantConfigJson: JsonObject | null, contextLength: number, kvCacheFreeGpuMemoryFraction: number, unifiedMemory: boolean): MemoryReserve;
|
|
69
|
+
/**
|
|
70
|
+
* What `trtllm-serve` holds outside torch on any card: the CUDA context, cuBLAS/cuDNN workspaces,
|
|
71
|
+
* NCCL buffers. Measured 1.14–1.62 GiB on an RTX 4070 Laptop (Windows live acceptance, 2026-10-03,
|
|
72
|
+
* "Memory used outside torch"), then 1.6–1.8 GiB on an RTX 5090 Laptop (2026-10-09, the logged
|
|
73
|
+
* figure less the 0.32 GiB other processes held); the check takes the high end with headroom.
|
|
74
|
+
*/
|
|
75
|
+
export declare const TENSORRT_LLM_RUNTIME_OVERHEAD_BYTES: number;
|
|
76
|
+
/**
|
|
77
|
+
* The share of the checkpoint's weight bytes added for the engine's own layout of them, which is not
|
|
78
|
+
* the checkpoint's: the 1.3.0rc29 loader keeps Qwen3.5's linear-attention `in_proj_qkvz` in bf16
|
|
79
|
+
* whatever the checkpoint stores (`_add_qkvz_bf16_workaround`), and skips `mtp.*` layers. Measured on
|
|
80
|
+
* the RTX 5090 Laptop (2026-10-09): Qwen3.8-27B-NVFP4 held 21.71 GiB for 20.42 GiB of files (+6.3%),
|
|
81
|
+
* NVIDIA-Nemotron-3.5-Lightning-30B-A3B-NVFP4 18.23 GiB for 20.08 (−9.2%, 2.49 GiB of MTP unread).
|
|
82
|
+
* 5% refuses the first on that card's 23.57 GiB free, which otherwise spilled into shared memory
|
|
83
|
+
* under WDDM and never got ready, and still passes the second, which loads.
|
|
84
|
+
*/
|
|
85
|
+
export declare const TENSORRT_LLM_WEIGHT_LAYOUT_ALLOWANCE = 0.05;
|
|
86
|
+
/** Bytes added to the checkpoint's weights for the engine's own layout of them. */
|
|
87
|
+
export declare function weightLayoutAllowanceBytes(weightBytesTotal: number): number;
|
|
88
|
+
/**
|
|
89
|
+
* The engine's own memory beyond weights and KV cache: the runtime overhead plus the activation
|
|
90
|
+
* peak of a `contextLength`-token prefill. Without it a checkpoint whose weights fit an 8 GB card
|
|
91
|
+
* passed the check and then left the KV cache 224 tokens (Ministral-3b bf16, 8.20 GiB peak on an
|
|
92
|
+
* 8.00 GiB card), and every request failed mid-stream.
|
|
93
|
+
*/
|
|
94
|
+
export declare function engineOverheadBytes(configJson: JsonObject, contextLength: number): number;
|
|
95
|
+
/** What sizes the KV-cache reserve: the provider's settings, resolved by the caller (`check.ts`, `runtime.ts`). */
|
|
96
|
+
export interface MemorySizingInputs {
|
|
97
|
+
/** The session's context length: stored provider settings, a load's own overrides, or the
|
|
98
|
+
* adapter's default. */
|
|
99
|
+
contextLength: number;
|
|
100
|
+
kvCacheFreeGpuMemoryFraction: number;
|
|
101
|
+
/** The provider's `cuda_graphs`. Only `on` adds the graphs' memory: `auto` turns them off where
|
|
102
|
+
* they would not fit (`tensorrtLlmLaunchPlan`), `off` never captures them. Absent is `auto`. */
|
|
103
|
+
cudaGraphs?: TensorrtLlmCudaGraphsSetting;
|
|
104
|
+
}
|
|
105
|
+
/** TensorRT-LLM's memory rule on `gpu`: its weights as the engine lays them out, plus that card's own kind of KV reserve, plus the engine. */
|
|
106
|
+
export declare function tensorrtLlmMemoryNeed(gpu: GpuFacts, checkpoint: {
|
|
107
|
+
weightBytesTotal: number;
|
|
108
|
+
configJson: JsonObject;
|
|
109
|
+
hfQuantConfigJson: JsonObject | null;
|
|
110
|
+
}, memory: MemorySizingInputs): MemoryNeed;
|
|
111
|
+
/**
|
|
112
|
+
* The launch plan `beforeCreate` hands the adapter (`ManagedEngineSpec.launchPlan`), from the card
|
|
113
|
+
* re-probed once the previous sessions stopped and the memory check passed: `auto` captures CUDA
|
|
114
|
+
* graphs only when they fit in what the card has free beyond everything else the check counts.
|
|
115
|
+
*/
|
|
116
|
+
export declare function tensorrtLlmLaunchPlan(memory: MemorySizingInputs, checkpoint: CheckpointMemoryInputs, gpu: GpuFacts, host: HostMemory): TensorrtLlmLaunchPlan;
|
|
117
|
+
/** The skeleton's hooks for TensorRT-LLM, sized by the provider's settings. */
|
|
118
|
+
export declare function tensorrtLlmCheckEngine(memory: MemorySizingInputs): ManagedCheckEngine;
|
|
119
|
+
/** Every check above the memory line, for TensorRT-LLM (see `checkCheckpointFiles`). */
|
|
120
|
+
export declare function checkModelCompatibilityFiles(input: ModelCheckInput, descriptor: RuntimeDescriptor, gpus: readonly GpuFacts[], hostMemory: HostMemory, memory: MemorySizingInputs): FilesCheckResult;
|
|
121
|
+
/** The memory line alone, for TensorRT-LLM (see `checkCheckpointMemory`). */
|
|
122
|
+
export declare function checkModelMemory(resolved: ResolvedCheckpoint, descriptor: RuntimeDescriptor, gpus: readonly GpuFacts[], hostMemory: HostMemory, memory: MemorySizingInputs): ModelCompatibility;
|
|
123
|
+
/** The full verdict for TensorRT-LLM (see `checkCheckpoint`). */
|
|
124
|
+
export declare function checkModelCompatibility(input: ModelCheckInput, descriptor: RuntimeDescriptor, gpus: readonly GpuFacts[], hostMemory: HostMemory, memory: MemorySizingInputs): ModelCompatibility;
|
|
@@ -0,0 +1,213 @@
|
|
|
1
|
+
/**
|
|
2
|
+
* TensorRT-LLM's part of the managed compatibility check (spec `tensorrt-llm-models`, "Проверка
|
|
3
|
+
* совместимости без сети"; change `add-vllm-runtime`, design D7). The skeleton — GGUF, the curated
|
|
4
|
+
* digest, the shared naming rule, architecture, format matrix, compute capability, the memory line —
|
|
5
|
+
* is `../managed-models/compatibility.ts`, the same for every managed engine. What is TensorRT-LLM's
|
|
6
|
+
* own lives here as the two hooks of `tensorrtLlmCheckEngine`:
|
|
7
|
+
*
|
|
8
|
+
* - `memoryNeed`: weights plus an allowance for the engine's own layout of them, plus the KV
|
|
9
|
+
* reserve divided by `kv_cache_free_gpu_memory_fraction` (`trtllm-serve`'s semantics: it spends
|
|
10
|
+
* only that fraction of what is free after the weights), bounded by tokens on a unified-memory
|
|
11
|
+
* card, plus the runtime and the activation peak;
|
|
12
|
+
* - `checkpointProblems`: shapes of supported architectures this image's loader cannot read
|
|
13
|
+
* (Nemotron-H with dense MLP layers, a DeepseekV3Gate MoE without its correction bias).
|
|
14
|
+
*
|
|
15
|
+
* `checkModelCompatibility`, `checkModelCompatibilityFiles` and `checkModelMemory` keep their
|
|
16
|
+
* signatures — `MemorySizingInputs` is the provider's settings — and are the skeleton with this
|
|
17
|
+
* engine plugged in. The skeleton's helpers are re-exported so this module's callers are unchanged.
|
|
18
|
+
*/
|
|
19
|
+
import { checkCheckpoint, checkCheckpointFiles, checkCheckpointMemory, freeMemoryBytes, isUnifiedMemory, kvCacheBytes, positiveNumberField, } from '../managed-models/compatibility.js';
|
|
20
|
+
import { TENSORRT_LLM_ENGINE_ID } from '../environment/index.js';
|
|
21
|
+
import { TENSORRT_LLM_CUDA_GRAPHS_RESERVE_BYTES, tensorrtLlmCudaGraphsOn, } from './cuda-graphs.js';
|
|
22
|
+
import { tensorrtLlmUnifiedKvMaxTokens } from './kv-cache.js';
|
|
23
|
+
export * from '../managed-models/compatibility.js';
|
|
24
|
+
const CORRECTION_BIAS_GATED_ARCHITECTURES = new Set([
|
|
25
|
+
'DeepseekV3ForCausalLM',
|
|
26
|
+
'DeepseekV32ForCausalLM',
|
|
27
|
+
'GlmMoeDsaForCausalLM',
|
|
28
|
+
'Glm4MoeForCausalLM',
|
|
29
|
+
]);
|
|
30
|
+
const CORRECTION_BIAS_TENSOR = 'e_score_correction_bias';
|
|
31
|
+
const NEMOTRON_H_ARCHITECTURE = /^NemotronH/;
|
|
32
|
+
/**
|
|
33
|
+
* A checkpoint whose architecture is supported but whose own shape the engine image cannot load —
|
|
34
|
+
* what the architecture name alone cannot tell (Windows acceptance machine, 2026-10-03):
|
|
35
|
+
*
|
|
36
|
+
* - Nemotron-H with dense MLP layers: a `-` in `hybrid_override_pattern`. The `transformers` 5.5.4
|
|
37
|
+
* in the 1.3.0rc29 image maps that pattern through `M`/`E`/`*` only and fails with
|
|
38
|
+
* `KeyError: '-'` before TensorRT-LLM (which does support `-`) is reached
|
|
39
|
+
* (NVIDIA-Nemotron-3-Nano-4B, Nano-9B-v2). Drop this once a descriptor's image ships a
|
|
40
|
+
* `transformers` that knows `-`.
|
|
41
|
+
* - A `DeepseekV3Gate` architecture with MoE layers (`n_routed_experts > 0`) whose tensors have no
|
|
42
|
+
* `e_score_correction_bias` (PrimeIntellect/GLM-0.5B, saved in its training framework's own
|
|
43
|
+
* layout — `mlp.router.gate.weight`, `mlp.expert_bias` — not the Hugging Face one TensorRT-LLM
|
|
44
|
+
* reads): `AssertionError` while the weights load. Needs `weight_names`; without them it is not
|
|
45
|
+
* checked.
|
|
46
|
+
*
|
|
47
|
+
* `null` when none applies.
|
|
48
|
+
*/
|
|
49
|
+
export function tensorrtLlmCheckpointProblems(architecture, configJson, weightNames) {
|
|
50
|
+
if (NEMOTRON_H_ARCHITECTURE.test(architecture)) {
|
|
51
|
+
const pattern = configJson.hybrid_override_pattern;
|
|
52
|
+
if (typeof pattern === 'string' && pattern.includes('-')) {
|
|
53
|
+
return {
|
|
54
|
+
message: 'This engine release cannot load Nemotron-H checkpoints with dense MLP layers ("-" in hybrid_override_pattern).',
|
|
55
|
+
details: `architecture=${architecture} hybrid_override_pattern=${pattern}`,
|
|
56
|
+
};
|
|
57
|
+
}
|
|
58
|
+
}
|
|
59
|
+
if (CORRECTION_BIAS_GATED_ARCHITECTURES.has(architecture) &&
|
|
60
|
+
weightNames !== undefined &&
|
|
61
|
+
weightNames.length > 0 &&
|
|
62
|
+
(positiveNumberField(configJson, 'n_routed_experts') ?? 0) > 0 &&
|
|
63
|
+
!weightNames.some((name) => name.endsWith(CORRECTION_BIAS_TENSOR))) {
|
|
64
|
+
return {
|
|
65
|
+
message: `This checkpoint has no ${CORRECTION_BIAS_TENSOR} tensors, which ${architecture} needs in its MoE router on this engine.`,
|
|
66
|
+
details: `architecture=${architecture} missing=${CORRECTION_BIAS_TENSOR}`,
|
|
67
|
+
};
|
|
68
|
+
}
|
|
69
|
+
return null;
|
|
70
|
+
}
|
|
71
|
+
/**
|
|
72
|
+
* Bytes reserved on top of the weights when checking whether a checkpoint fits a card's free memory.
|
|
73
|
+
*
|
|
74
|
+
* On a discrete card, `KV_bytes / kv_cache_free_gpu_memory_fraction` (task 2.16w round 1, finding 6
|
|
75
|
+
* (RULING), superseding the placeholder `weights × (1 − fraction)` rule): `trtllm-serve` spends only
|
|
76
|
+
* that fraction of whatever memory remains free *after* weights load on the KV cache, keeping the
|
|
77
|
+
* rest as headroom, so guaranteeing `KV_bytes` of real cache capacity needs `KV_bytes / fraction` of
|
|
78
|
+
* free memory left over once weights are loaded — the ADR this formula documents derives that
|
|
79
|
+
* division in full.
|
|
80
|
+
*
|
|
81
|
+
* On a unified-memory card (`unifiedMemory`, design D13) the launch bounds the KV cache by tokens,
|
|
82
|
+
* `kv_cache_config.max_tokens = tensorrtLlmUnifiedKvMaxTokens(contextLength)` (`kv-cache.ts`), and
|
|
83
|
+
* TensorRT-LLM uses the smaller of that bound and the fraction; the reserve is therefore the KV for
|
|
84
|
+
* exactly that many tokens, not divided by the fraction, so the check and the launch agree
|
|
85
|
+
* (docs/decisions/2026-09-29-tensorrt-llm-unified-memory-kv-cache-bounded-by-tokens.md).
|
|
86
|
+
*
|
|
87
|
+
* `basis: 'config'` when `kvCacheBytes` had what it needed; `basis: 'weight_fraction'` — the older,
|
|
88
|
+
* cruder `weights × (1 − fraction)` rule, the same on either kind of card — only when `config.json`
|
|
89
|
+
* lacked the architecture fields the real formula needs, and the returned `basis` says so, so a
|
|
90
|
+
* verdict computed from the fallback is never silently indistinguishable from one computed from the
|
|
91
|
+
* checkpoint's real shape.
|
|
92
|
+
*/
|
|
93
|
+
export function kvCacheReserveBytes(weightBytesTotal, configJson, hfQuantConfigJson, contextLength, kvCacheFreeGpuMemoryFraction, unifiedMemory) {
|
|
94
|
+
const kvTokens = unifiedMemory ? tensorrtLlmUnifiedKvMaxTokens(contextLength) : contextLength;
|
|
95
|
+
const kvBytes = kvCacheBytes(configJson, hfQuantConfigJson, kvTokens);
|
|
96
|
+
if (kvBytes !== undefined) {
|
|
97
|
+
return {
|
|
98
|
+
reserveBytes: unifiedMemory ? kvBytes : Math.ceil(kvBytes / kvCacheFreeGpuMemoryFraction),
|
|
99
|
+
basis: 'config',
|
|
100
|
+
};
|
|
101
|
+
}
|
|
102
|
+
return {
|
|
103
|
+
reserveBytes: Math.ceil(weightBytesTotal * (1 - kvCacheFreeGpuMemoryFraction)),
|
|
104
|
+
basis: 'weight_fraction',
|
|
105
|
+
};
|
|
106
|
+
}
|
|
107
|
+
/**
|
|
108
|
+
* What `trtllm-serve` holds outside torch on any card: the CUDA context, cuBLAS/cuDNN workspaces,
|
|
109
|
+
* NCCL buffers. Measured 1.14–1.62 GiB on an RTX 4070 Laptop (Windows live acceptance, 2026-10-03,
|
|
110
|
+
* "Memory used outside torch"), then 1.6–1.8 GiB on an RTX 5090 Laptop (2026-10-09, the logged
|
|
111
|
+
* figure less the 0.32 GiB other processes held); the check takes the high end with headroom.
|
|
112
|
+
*/
|
|
113
|
+
export const TENSORRT_LLM_RUNTIME_OVERHEAD_BYTES = 2 * 1024 ** 3;
|
|
114
|
+
/**
|
|
115
|
+
* The share of the checkpoint's weight bytes added for the engine's own layout of them, which is not
|
|
116
|
+
* the checkpoint's: the 1.3.0rc29 loader keeps Qwen3.5's linear-attention `in_proj_qkvz` in bf16
|
|
117
|
+
* whatever the checkpoint stores (`_add_qkvz_bf16_workaround`), and skips `mtp.*` layers. Measured on
|
|
118
|
+
* the RTX 5090 Laptop (2026-10-09): Qwen3.8-27B-NVFP4 held 21.71 GiB for 20.42 GiB of files (+6.3%),
|
|
119
|
+
* NVIDIA-Nemotron-3.5-Lightning-30B-A3B-NVFP4 18.23 GiB for 20.08 (−9.2%, 2.49 GiB of MTP unread).
|
|
120
|
+
* 5% refuses the first on that card's 23.57 GiB free, which otherwise spilled into shared memory
|
|
121
|
+
* under WDDM and never got ready, and still passes the second, which loads.
|
|
122
|
+
*/
|
|
123
|
+
export const TENSORRT_LLM_WEIGHT_LAYOUT_ALLOWANCE = 0.05;
|
|
124
|
+
/** Bytes added to the checkpoint's weights for the engine's own layout of them. */
|
|
125
|
+
export function weightLayoutAllowanceBytes(weightBytesTotal) {
|
|
126
|
+
return Math.ceil(Math.max(weightBytesTotal, 0) * TENSORRT_LLM_WEIGHT_LAYOUT_ALLOWANCE);
|
|
127
|
+
}
|
|
128
|
+
/**
|
|
129
|
+
* Bytes of activation peak per prompt token per MLP intermediate element: the engine profiles a
|
|
130
|
+
* warmup of `max_num_tokens` tokens (the launch sets it to the context length), and the peak grows
|
|
131
|
+
* with both. Measured 0.91 GiB at 8192 tokens and 0.46 GiB at 4096 tokens for an
|
|
132
|
+
* `intermediate_size` of 14336 (ministral/Ministral-3b-instruct, same acceptance machine): ≈ 8.4.
|
|
133
|
+
*/
|
|
134
|
+
const ACTIVATION_BYTES_PER_TOKEN_ELEMENT = 8.5;
|
|
135
|
+
/** The MLP width that sizes the activations: `intermediate_size`, a VLM's `text_config` one, else 4 × hidden. */
|
|
136
|
+
function intermediateSize(rootConfigJson) {
|
|
137
|
+
const textConfig = typeof rootConfigJson.text_config === 'object' && rootConfigJson.text_config !== null
|
|
138
|
+
? rootConfigJson.text_config
|
|
139
|
+
: undefined;
|
|
140
|
+
for (const config of [rootConfigJson, textConfig]) {
|
|
141
|
+
if (config === undefined)
|
|
142
|
+
continue;
|
|
143
|
+
const intermediate = positiveNumberField(config, 'intermediate_size');
|
|
144
|
+
if (intermediate !== undefined)
|
|
145
|
+
return intermediate;
|
|
146
|
+
const hidden = positiveNumberField(config, 'hidden_size');
|
|
147
|
+
if (hidden !== undefined)
|
|
148
|
+
return 4 * hidden;
|
|
149
|
+
}
|
|
150
|
+
return undefined;
|
|
151
|
+
}
|
|
152
|
+
/**
|
|
153
|
+
* The engine's own memory beyond weights and KV cache: the runtime overhead plus the activation
|
|
154
|
+
* peak of a `contextLength`-token prefill. Without it a checkpoint whose weights fit an 8 GB card
|
|
155
|
+
* passed the check and then left the KV cache 224 tokens (Ministral-3b bf16, 8.20 GiB peak on an
|
|
156
|
+
* 8.00 GiB card), and every request failed mid-stream.
|
|
157
|
+
*/
|
|
158
|
+
export function engineOverheadBytes(configJson, contextLength) {
|
|
159
|
+
const intermediate = intermediateSize(configJson) ?? 0;
|
|
160
|
+
return Math.ceil(TENSORRT_LLM_RUNTIME_OVERHEAD_BYTES + contextLength * intermediate * ACTIVATION_BYTES_PER_TOKEN_ELEMENT);
|
|
161
|
+
}
|
|
162
|
+
/** TensorRT-LLM's memory rule on `gpu`: its weights as the engine lays them out, plus that card's own kind of KV reserve, plus the engine. */
|
|
163
|
+
export function tensorrtLlmMemoryNeed(gpu, checkpoint, memory) {
|
|
164
|
+
const unified = isUnifiedMemory(gpu);
|
|
165
|
+
const reserve = kvCacheReserveBytes(checkpoint.weightBytesTotal, checkpoint.configJson, checkpoint.hfQuantConfigJson, memory.contextLength, memory.kvCacheFreeGpuMemoryFraction, unified);
|
|
166
|
+
const overheadBytes = engineOverheadBytes(checkpoint.configJson, memory.contextLength);
|
|
167
|
+
const layoutBytes = weightLayoutAllowanceBytes(checkpoint.weightBytesTotal);
|
|
168
|
+
const cudaGraphsBytes = memory.cudaGraphs === 'on' ? TENSORRT_LLM_CUDA_GRAPHS_RESERVE_BYTES : 0;
|
|
169
|
+
const neededBytes = checkpoint.weightBytesTotal + layoutBytes + reserve.reserveBytes + overheadBytes + cudaGraphsBytes;
|
|
170
|
+
// The token bound the launch writes, on a card where it writes one (adapter.ts).
|
|
171
|
+
const kvTokens = unified && reserve.basis === 'config'
|
|
172
|
+
? ` kv_max_tokens=${tensorrtLlmUnifiedKvMaxTokens(memory.contextLength)}`
|
|
173
|
+
: '';
|
|
174
|
+
return {
|
|
175
|
+
neededBytes,
|
|
176
|
+
kvReserveBasis: reserve.basis,
|
|
177
|
+
details: `weight_bytes=${checkpoint.weightBytesTotal} weight_layout_bytes=${layoutBytes} kv_reserve_bytes=${reserve.reserveBytes} kv_reserve_basis=${reserve.basis}${kvTokens} engine_overhead_bytes=${overheadBytes}${cudaGraphsBytes > 0 ? ` cuda_graphs_bytes=${cudaGraphsBytes}` : ''} needed_bytes=${neededBytes}`,
|
|
178
|
+
};
|
|
179
|
+
}
|
|
180
|
+
/**
|
|
181
|
+
* The launch plan `beforeCreate` hands the adapter (`ManagedEngineSpec.launchPlan`), from the card
|
|
182
|
+
* re-probed once the previous sessions stopped and the memory check passed: `auto` captures CUDA
|
|
183
|
+
* graphs only when they fit in what the card has free beyond everything else the check counts.
|
|
184
|
+
*/
|
|
185
|
+
export function tensorrtLlmLaunchPlan(memory, checkpoint, gpu, host) {
|
|
186
|
+
const setting = memory.cudaGraphs ?? 'auto';
|
|
187
|
+
if (setting !== 'auto')
|
|
188
|
+
return { cudaGraphs: setting === 'on' };
|
|
189
|
+
const withoutGraphs = tensorrtLlmMemoryNeed(gpu, checkpoint, { ...memory, cudaGraphs: 'off' });
|
|
190
|
+
const headroomBytes = freeMemoryBytes(gpu, host) - withoutGraphs.neededBytes;
|
|
191
|
+
return { cudaGraphs: tensorrtLlmCudaGraphsOn('auto', gpu.total_vram_bytes, headroomBytes) };
|
|
192
|
+
}
|
|
193
|
+
/** The skeleton's hooks for TensorRT-LLM, sized by the provider's settings. */
|
|
194
|
+
export function tensorrtLlmCheckEngine(memory) {
|
|
195
|
+
return {
|
|
196
|
+
engineId: TENSORRT_LLM_ENGINE_ID,
|
|
197
|
+
checkpointProblems: tensorrtLlmCheckpointProblems,
|
|
198
|
+
memoryNeed: (gpu, checkpoint) => tensorrtLlmMemoryNeed(gpu, checkpoint, memory),
|
|
199
|
+
};
|
|
200
|
+
}
|
|
201
|
+
/** Every check above the memory line, for TensorRT-LLM (see `checkCheckpointFiles`). */
|
|
202
|
+
export function checkModelCompatibilityFiles(input, descriptor, gpus, hostMemory, memory) {
|
|
203
|
+
return checkCheckpointFiles(input, descriptor, gpus, hostMemory, tensorrtLlmCheckEngine(memory));
|
|
204
|
+
}
|
|
205
|
+
/** The memory line alone, for TensorRT-LLM (see `checkCheckpointMemory`). */
|
|
206
|
+
export function checkModelMemory(resolved, descriptor, gpus, hostMemory, memory) {
|
|
207
|
+
return checkCheckpointMemory(resolved, descriptor, gpus, hostMemory, tensorrtLlmCheckEngine(memory));
|
|
208
|
+
}
|
|
209
|
+
/** The full verdict for TensorRT-LLM (see `checkCheckpoint`). */
|
|
210
|
+
export function checkModelCompatibility(input, descriptor, gpus, hostMemory, memory) {
|
|
211
|
+
return checkCheckpoint(input, descriptor, gpus, hostMemory, tensorrtLlmCheckEngine(memory));
|
|
212
|
+
}
|
|
213
|
+
//# sourceMappingURL=compatibility.js.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"compatibility.js","sourceRoot":"","sources":["../../../src/runtime/tensorrt-llm/compatibility.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;;;GAiBG;AAGH,OAAO,EACL,eAAe,EACf,oBAAoB,EACpB,qBAAqB,EACrB,eAAe,EACf,eAAe,EACf,YAAY,EACZ,mBAAmB,GAUpB,MAAM,oCAAoC,CAAA;AAE3C,OAAO,EAAE,sBAAsB,EAAE,MAAM,yBAAyB,CAAA;AAChE,OAAO,EACL,sCAAsC,EACtC,uBAAuB,GAGxB,MAAM,kBAAkB,CAAA;AACzB,OAAO,EAAE,6BAA6B,EAAE,MAAM,eAAe,CAAA;AAE7D,cAAc,oCAAoC,CAAA;AAElD,MAAM,mCAAmC,GAAwB,IAAI,GAAG,CAAC;IACvE,uBAAuB;IACvB,wBAAwB;IACxB,sBAAsB;IACtB,oBAAoB;CACrB,CAAC,CAAA;AAEF,MAAM,sBAAsB,GAAG,yBAAyB,CAAA;AAExD,MAAM,uBAAuB,GAAG,YAAY,CAAA;AAE5C;;;;;;;;;;;;;;;;GAgBG;AACH,MAAM,UAAU,6BAA6B,CAC3C,YAAoB,EACpB,UAAsB,EACtB,WAA0C;IAE1C,IAAI,uBAAuB,CAAC,IAAI,CAAC,YAAY,CAAC,EAAE,CAAC;QAC/C,MAAM,OAAO,GAAG,UAAU,CAAC,uBAAuB,CAAA;QAClD,IAAI,OAAO,OAAO,KAAK,QAAQ,IAAI,OAAO,CAAC,QAAQ,CAAC,GAAG,CAAC,EAAE,CAAC;YACzD,OAAO;gBACL,OAAO,EACL,gHAAgH;gBAClH,OAAO,EAAE,gBAAgB,YAAY,4BAA4B,OAAO,EAAE;aAC3E,CAAA;QACH,CAAC;IACH,CAAC;IACD,IACE,mCAAmC,CAAC,GAAG,CAAC,YAAY,CAAC;QACrD,WAAW,KAAK,SAAS;QACzB,WAAW,CAAC,MAAM,GAAG,CAAC;QACtB,CAAC,mBAAmB,CAAC,UAAU,EAAE,kBAAkB,CAAC,IAAI,CAAC,CAAC,GAAG,CAAC;QAC9D,CAAC,WAAW,CAAC,IAAI,CAAC,CAAC,IAAI,EAAE,EAAE,CAAC,IAAI,CAAC,QAAQ,CAAC,sBAAsB,CAAC,CAAC,EAClE,CAAC;QACD,OAAO;YACL,OAAO,EAAE,0BAA0B,sBAAsB,mBAAmB,YAAY,0CAA0C;YAClI,OAAO,EAAE,gBAAgB,YAAY,YAAY,sBAAsB,EAAE;SAC1E,CAAA;IACH,CAAC;IACD,OAAO,IAAI,CAAA;AACb,CAAC;AAOD;;;;;;;;;;;;;;;;;;;;;GAqBG;AACH,MAAM,UAAU,mBAAmB,CACjC,gBAAwB,EACxB,UAAsB,EACtB,iBAAoC,EACpC,aAAqB,EACrB,4BAAoC,EACpC,aAAsB;IAEtB,MAAM,QAAQ,GAAG,aAAa,CAAC,CAAC,CAAC,6BAA6B,CAAC,aAAa,CAAC,CAAC,CAAC,CAAC,aAAa,CAAA;IAC7F,MAAM,OAAO,GAAG,YAAY,CAAC,UAAU,EAAE,iBAAiB,EAAE,QAAQ,CAAC,CAAA;IACrE,IAAI,OAAO,KAAK,SAAS,EAAE,CAAC;QAC1B,OAAO;YACL,YAAY,EAAE,aAAa,CAAC,CAAC,CAAC,OAAO,CAAC,CAAC,CAAC,IAAI,CAAC,IAAI,CAAC,OAAO,GAAG,4BAA4B,CAAC;YACzF,KAAK,EAAE,QAAQ;SAChB,CAAA;IACH,CAAC;IACD,OAAO;QACL,YAAY,EAAE,IAAI,CAAC,IAAI,CAAC,gBAAgB,GAAG,CAAC,CAAC,GAAG,4BAA4B,CAAC,CAAC;QAC9E,KAAK,EAAE,iBAAiB;KACzB,CAAA;AACH,CAAC;AAED;;;;;GAKG;AACH,MAAM,CAAC,MAAM,mCAAmC,GAAG,CAAC,GAAG,IAAI,IAAI,CAAC,CAAA;AAEhE;;;;;;;;GAQG;AACH,MAAM,CAAC,MAAM,oCAAoC,GAAG,IAAI,CAAA;AAExD,mFAAmF;AACnF,MAAM,UAAU,0BAA0B,CAAC,gBAAwB;IACjE,OAAO,IAAI,CAAC,IAAI,CAAC,IAAI,CAAC,GAAG,CAAC,gBAAgB,EAAE,CAAC,CAAC,GAAG,oCAAoC,CAAC,CAAA;AACxF,CAAC;AAED;;;;;GAKG;AACH,MAAM,kCAAkC,GAAG,GAAG,CAAA;AAE9C,iHAAiH;AACjH,SAAS,gBAAgB,CAAC,cAA0B;IAClD,MAAM,UAAU,GACd,OAAO,cAAc,CAAC,WAAW,KAAK,QAAQ,IAAI,cAAc,CAAC,WAAW,KAAK,IAAI;QACnF,CAAC,CAAE,cAAc,CAAC,WAA0B;QAC5C,CAAC,CAAC,SAAS,CAAA;IACf,KAAK,MAAM,MAAM,IAAI,CAAC,cAAc,EAAE,UAAU,CAAC,EAAE,CAAC;QAClD,IAAI,MAAM,KAAK,SAAS;YAAE,SAAQ;QAClC,MAAM,YAAY,GAAG,mBAAmB,CAAC,MAAM,EAAE,mBAAmB,CAAC,CAAA;QACrE,IAAI,YAAY,KAAK,SAAS;YAAE,OAAO,YAAY,CAAA;QACnD,MAAM,MAAM,GAAG,mBAAmB,CAAC,MAAM,EAAE,aAAa,CAAC,CAAA;QACzD,IAAI,MAAM,KAAK,SAAS;YAAE,OAAO,CAAC,GAAG,MAAM,CAAA;IAC7C,CAAC;IACD,OAAO,SAAS,CAAA;AAClB,CAAC;AAED;;;;;GAKG;AACH,MAAM,UAAU,mBAAmB,CAAC,UAAsB,EAAE,aAAqB;IAC/E,MAAM,YAAY,GAAG,gBAAgB,CAAC,UAAU,CAAC,IAAI,CAAC,CAAA;IACtD,OAAO,IAAI,CAAC,IAAI,CACd,mCAAmC,GAAG,aAAa,GAAG,YAAY,GAAG,kCAAkC,CACxG,CAAA;AACH,CAAC;AAaD,8IAA8I;AAC9I,MAAM,UAAU,qBAAqB,CACnC,GAAa,EACb,UAAsG,EACtG,MAA0B;IAE1B,MAAM,OAAO,GAAG,eAAe,CAAC,GAAG,CAAC,CAAA;IACpC,MAAM,OAAO,GAAG,mBAAmB,CACjC,UAAU,CAAC,gBAAgB,EAC3B,UAAU,CAAC,UAAU,EACrB,UAAU,CAAC,iBAAiB,EAC5B,MAAM,CAAC,aAAa,EACpB,MAAM,CAAC,4BAA4B,EACnC,OAAO,CACR,CAAA;IACD,MAAM,aAAa,GAAG,mBAAmB,CAAC,UAAU,CAAC,UAAU,EAAE,MAAM,CAAC,aAAa,CAAC,CAAA;IACtF,MAAM,WAAW,GAAG,0BAA0B,CAAC,UAAU,CAAC,gBAAgB,CAAC,CAAA;IAC3E,MAAM,eAAe,GAAG,MAAM,CAAC,UAAU,KAAK,IAAI,CAAC,CAAC,CAAC,sCAAsC,CAAC,CAAC,CAAC,CAAC,CAAA;IAC/F,MAAM,WAAW,GACf,UAAU,CAAC,gBAAgB,GAAG,WAAW,GAAG,OAAO,CAAC,YAAY,GAAG,aAAa,GAAG,eAAe,CAAA;IACpG,iFAAiF;IACjF,MAAM,QAAQ,GACZ,OAAO,IAAI,OAAO,CAAC,KAAK,KAAK,QAAQ;QACnC,CAAC,CAAC,kBAAkB,6BAA6B,CAAC,MAAM,CAAC,aAAa,CAAC,EAAE;QACzE,CAAC,CAAC,EAAE,CAAA;IACR,OAAO;QACL,WAAW;QACX,cAAc,EAAE,OAAO,CAAC,KAAK;QAC7B,OAAO,EAAE,gBAAgB,UAAU,CAAC,gBAAgB,wBAAwB,WAAW,qBAAqB,OAAO,CAAC,YAAY,qBAAqB,OAAO,CAAC,KAAK,GAAG,QAAQ,0BAA0B,aAAa,GAAG,eAAe,GAAG,CAAC,CAAC,CAAC,CAAC,sBAAsB,eAAe,EAAE,CAAC,CAAC,CAAC,EAAE,iBAAiB,WAAW,EAAE;KACxT,CAAA;AACH,CAAC;AAED;;;;GAIG;AACH,MAAM,UAAU,qBAAqB,CACnC,MAA0B,EAC1B,UAAkC,EAClC,GAAa,EACb,IAAgB;IAEhB,MAAM,OAAO,GAAG,MAAM,CAAC,UAAU,IAAI,MAAM,CAAA;IAC3C,IAAI,OAAO,KAAK,MAAM;QAAE,OAAO,EAAE,UAAU,EAAE,OAAO,KAAK,IAAI,EAAE,CAAA;IAC/D,MAAM,aAAa,GAAG,qBAAqB,CAAC,GAAG,EAAE,UAAU,EAAE,EAAE,GAAG,MAAM,EAAE,UAAU,EAAE,KAAK,EAAE,CAAC,CAAA;IAC9F,MAAM,aAAa,GAAG,eAAe,CAAC,GAAG,EAAE,IAAI,CAAC,GAAG,aAAa,CAAC,WAAW,CAAA;IAC5E,OAAO,EAAE,UAAU,EAAE,uBAAuB,CAAC,MAAM,EAAE,GAAG,CAAC,gBAAgB,EAAE,aAAa,CAAC,EAAE,CAAA;AAC7F,CAAC;AAED,+EAA+E;AAC/E,MAAM,UAAU,sBAAsB,CAAC,MAA0B;IAC/D,OAAO;QACL,QAAQ,EAAE,sBAAsB;QAChC,kBAAkB,EAAE,6BAA6B;QACjD,UAAU,EAAE,CAAC,GAAG,EAAE,UAAU,EAAE,EAAE,CAAC,qBAAqB,CAAC,GAAG,EAAE,UAAU,EAAE,MAAM,CAAC;KAChF,CAAA;AACH,CAAC;AAED,wFAAwF;AACxF,MAAM,UAAU,4BAA4B,CAC1C,KAAsB,EACtB,UAA6B,EAC7B,IAAyB,EACzB,UAAsB,EACtB,MAA0B;IAE1B,OAAO,oBAAoB,CAAC,KAAK,EAAE,UAAU,EAAE,IAAI,EAAE,UAAU,EAAE,sBAAsB,CAAC,MAAM,CAAC,CAAC,CAAA;AAClG,CAAC;AAED,6EAA6E;AAC7E,MAAM,UAAU,gBAAgB,CAC9B,QAA4B,EAC5B,UAA6B,EAC7B,IAAyB,EACzB,UAAsB,EACtB,MAA0B;IAE1B,OAAO,qBAAqB,CAAC,QAAQ,EAAE,UAAU,EAAE,IAAI,EAAE,UAAU,EAAE,sBAAsB,CAAC,MAAM,CAAC,CAAC,CAAA;AACtG,CAAC;AAED,iEAAiE;AACjE,MAAM,UAAU,uBAAuB,CACrC,KAAsB,EACtB,UAA6B,EAC7B,IAAyB,EACzB,UAAsB,EACtB,MAA0B;IAE1B,OAAO,eAAe,CAAC,KAAK,EAAE,UAAU,EAAE,IAAI,EAAE,UAAU,EAAE,sBAAsB,CAAC,MAAM,CAAC,CAAC,CAAA;AAC7F,CAAC"}
|
|
@@ -0,0 +1,32 @@
|
|
|
1
|
+
/**
|
|
2
|
+
* Whether a TensorRT-LLM launch captures CUDA graphs, in one module the launch (`adapter.ts`), its
|
|
3
|
+
* plan and the memory check (`compatibility.ts`) all import: the graphs cost memory the check has to
|
|
4
|
+
* count, and `auto` has to know whether that memory is there.
|
|
5
|
+
*/
|
|
6
|
+
/** Below this much card memory `cuda_graphs: auto` leaves CUDA graphs off: they took about 2 GB on
|
|
7
|
+
* an 8 GB card in the Windows live acceptance (Qwen3.5-2B, "Memory used outside torch … 2.12 GiB").
|
|
8
|
+
* A card that reports no size of its own is a unified-memory one (GB10): it keeps them. */
|
|
9
|
+
export declare const TENSORRT_LLM_CUDA_GRAPHS_MIN_VRAM_BYTES: number;
|
|
10
|
+
/**
|
|
11
|
+
* What capturing CUDA graphs adds to the engine's peak, beyond what the memory check already counts.
|
|
12
|
+
* NVIDIA-Nemotron-3.5-Lightning-30B-A3B-NVFP4 on the RTX 5090 Laptop (2026-10-09): 21.47 GiB peak
|
|
13
|
+
* with graphs off, 26.37 GiB with them on (5.15 GiB outside torch instead of 1.92, 2.57 GiB of
|
|
14
|
+
* activations instead of 1.32), and every request then failed with "CUDA out of memory … 0 bytes is
|
|
15
|
+
* free". About 2 GB on an 8 GB card (Qwen3.5-2B). The larger figure, rounded up: too much only turns
|
|
16
|
+
* graphs off where they would have fit, too little starts a model that cannot answer.
|
|
17
|
+
*/
|
|
18
|
+
export declare const TENSORRT_LLM_CUDA_GRAPHS_RESERVE_BYTES: number;
|
|
19
|
+
export type TensorrtLlmCudaGraphsSetting = 'auto' | 'on' | 'off';
|
|
20
|
+
/** What `beforeCreate` hands the TensorRT-LLM launch (`ManagedLaunchContext.plan`), from the card as
|
|
21
|
+
* it stands right before the container is created. */
|
|
22
|
+
export interface TensorrtLlmLaunchPlan {
|
|
23
|
+
cudaGraphs: boolean;
|
|
24
|
+
}
|
|
25
|
+
/**
|
|
26
|
+
* Whether the launch captures CUDA graphs. `on` and `off` are the user's. `auto` turns them on only on
|
|
27
|
+
* a card of at least `TENSORRT_LLM_CUDA_GRAPHS_MIN_VRAM_BYTES` (or a unified-memory one, `vramBytes`
|
|
28
|
+
* null) — and, when the free memory left after everything else the check counts is known
|
|
29
|
+
* (`headroomBytes`), only when the graphs fit in it: a 24 GB card holding an 18 GiB model has the size
|
|
30
|
+
* but not the room.
|
|
31
|
+
*/
|
|
32
|
+
export declare function tensorrtLlmCudaGraphsOn(setting: TensorrtLlmCudaGraphsSetting, vramBytes: number | null, headroomBytes?: number): boolean;
|
|
@@ -0,0 +1,33 @@
|
|
|
1
|
+
/**
|
|
2
|
+
* Whether a TensorRT-LLM launch captures CUDA graphs, in one module the launch (`adapter.ts`), its
|
|
3
|
+
* plan and the memory check (`compatibility.ts`) all import: the graphs cost memory the check has to
|
|
4
|
+
* count, and `auto` has to know whether that memory is there.
|
|
5
|
+
*/
|
|
6
|
+
/** Below this much card memory `cuda_graphs: auto` leaves CUDA graphs off: they took about 2 GB on
|
|
7
|
+
* an 8 GB card in the Windows live acceptance (Qwen3.5-2B, "Memory used outside torch … 2.12 GiB").
|
|
8
|
+
* A card that reports no size of its own is a unified-memory one (GB10): it keeps them. */
|
|
9
|
+
export const TENSORRT_LLM_CUDA_GRAPHS_MIN_VRAM_BYTES = 12 * 1024 ** 3;
|
|
10
|
+
/**
|
|
11
|
+
* What capturing CUDA graphs adds to the engine's peak, beyond what the memory check already counts.
|
|
12
|
+
* NVIDIA-Nemotron-3.5-Lightning-30B-A3B-NVFP4 on the RTX 5090 Laptop (2026-10-09): 21.47 GiB peak
|
|
13
|
+
* with graphs off, 26.37 GiB with them on (5.15 GiB outside torch instead of 1.92, 2.57 GiB of
|
|
14
|
+
* activations instead of 1.32), and every request then failed with "CUDA out of memory … 0 bytes is
|
|
15
|
+
* free". About 2 GB on an 8 GB card (Qwen3.5-2B). The larger figure, rounded up: too much only turns
|
|
16
|
+
* graphs off where they would have fit, too little starts a model that cannot answer.
|
|
17
|
+
*/
|
|
18
|
+
export const TENSORRT_LLM_CUDA_GRAPHS_RESERVE_BYTES = 5 * 1024 ** 3;
|
|
19
|
+
/**
|
|
20
|
+
* Whether the launch captures CUDA graphs. `on` and `off` are the user's. `auto` turns them on only on
|
|
21
|
+
* a card of at least `TENSORRT_LLM_CUDA_GRAPHS_MIN_VRAM_BYTES` (or a unified-memory one, `vramBytes`
|
|
22
|
+
* null) — and, when the free memory left after everything else the check counts is known
|
|
23
|
+
* (`headroomBytes`), only when the graphs fit in it: a 24 GB card holding an 18 GiB model has the size
|
|
24
|
+
* but not the room.
|
|
25
|
+
*/
|
|
26
|
+
export function tensorrtLlmCudaGraphsOn(setting, vramBytes, headroomBytes) {
|
|
27
|
+
if (setting !== 'auto')
|
|
28
|
+
return setting === 'on';
|
|
29
|
+
if (vramBytes !== null && vramBytes < TENSORRT_LLM_CUDA_GRAPHS_MIN_VRAM_BYTES)
|
|
30
|
+
return false;
|
|
31
|
+
return headroomBytes === undefined || headroomBytes >= TENSORRT_LLM_CUDA_GRAPHS_RESERVE_BYTES;
|
|
32
|
+
}
|
|
33
|
+
//# sourceMappingURL=cuda-graphs.js.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"cuda-graphs.js","sourceRoot":"","sources":["../../../src/runtime/tensorrt-llm/cuda-graphs.ts"],"names":[],"mappings":"AAAA;;;;GAIG;AAEH;;4FAE4F;AAC5F,MAAM,CAAC,MAAM,uCAAuC,GAAG,EAAE,GAAG,IAAI,IAAI,CAAC,CAAA;AAErE;;;;;;;GAOG;AACH,MAAM,CAAC,MAAM,sCAAsC,GAAG,CAAC,GAAG,IAAI,IAAI,CAAC,CAAA;AAUnE;;;;;;GAMG;AACH,MAAM,UAAU,uBAAuB,CACrC,OAAqC,EACrC,SAAwB,EACxB,aAAsB;IAEtB,IAAI,OAAO,KAAK,MAAM;QAAE,OAAO,OAAO,KAAK,IAAI,CAAA;IAC/C,IAAI,SAAS,KAAK,IAAI,IAAI,SAAS,GAAG,uCAAuC;QAAE,OAAO,KAAK,CAAA;IAC3F,OAAO,aAAa,KAAK,SAAS,IAAI,aAAa,IAAI,sCAAsC,CAAA;AAC/F,CAAC"}
|
|
@@ -0,0 +1,15 @@
|
|
|
1
|
+
import type { ManagedScopePaths } from '../../config/index.js';
|
|
2
|
+
import type { ModelFileOps } from './guest-files.js';
|
|
3
|
+
export interface DeletedTensorrtLlmModelFiles {
|
|
4
|
+
freedBytes: number;
|
|
5
|
+
engineCachesRemoved: number;
|
|
6
|
+
}
|
|
7
|
+
/**
|
|
8
|
+
* `files` (change `add-tensorrt-llm-windows`, task 2.8): on Windows the model and its caches are in the
|
|
9
|
+
* WSL guest, and are sized and removed by commands there (`guestModelFiles`) — all caches and the model
|
|
10
|
+
* in one call each, the caches listed first. Without it, the walk below, as on Linux.
|
|
11
|
+
*/
|
|
12
|
+
export declare function deleteTensorrtLlmModelFiles(paths: ManagedScopePaths, model: {
|
|
13
|
+
id: string;
|
|
14
|
+
dir: string;
|
|
15
|
+
}, files?: ModelFileOps): Promise<DeletedTensorrtLlmModelFiles>;
|
|
@@ -0,0 +1,60 @@
|
|
|
1
|
+
/**
|
|
2
|
+
* The files of a deleted `tensorrt-llm` model (task 2.24, design D12a, spec `tensorrt-llm-models`
|
|
3
|
+
* "Модель удаляется через core"): every engine cache of the model, from any descriptor it was ever
|
|
4
|
+
* loaded with, then its own folder. Caches go first: a failure there leaves the model listed, so the
|
|
5
|
+
* deletion can simply be retried, instead of leaving gigabytes of cache no model points at.
|
|
6
|
+
*
|
|
7
|
+
* The caller (`tensorrtLlmModelDeleter` in `src/core/tensorrt-llm.ts`) has already stopped the model
|
|
8
|
+
* with Docker's confirmation and holds its loads off; nothing here asks Docker anything.
|
|
9
|
+
*/
|
|
10
|
+
import { lstat, readdir, rm } from 'node:fs/promises';
|
|
11
|
+
import { join } from 'node:path';
|
|
12
|
+
import { engineCacheDirsOf, removeEngineCaches } from '../managed-text/index.js';
|
|
13
|
+
/**
|
|
14
|
+
* Bytes of the regular files under `path`, symlinks not followed, a hard-linked file counted once
|
|
15
|
+
* across one deletion (`seen`). A missing path is 0.
|
|
16
|
+
*/
|
|
17
|
+
async function bytesUnder(path, seen) {
|
|
18
|
+
const info = await lstat(path).catch(() => undefined);
|
|
19
|
+
if (info === undefined)
|
|
20
|
+
return 0;
|
|
21
|
+
if (info.isFile()) {
|
|
22
|
+
const inode = `${info.dev}:${info.ino}`;
|
|
23
|
+
if (seen.has(inode))
|
|
24
|
+
return 0;
|
|
25
|
+
seen.add(inode);
|
|
26
|
+
return info.size;
|
|
27
|
+
}
|
|
28
|
+
if (!info.isDirectory())
|
|
29
|
+
return 0;
|
|
30
|
+
const children = await readdir(path).catch(() => []);
|
|
31
|
+
let total = 0;
|
|
32
|
+
for (const child of children)
|
|
33
|
+
total += await bytesUnder(join(path, child), seen);
|
|
34
|
+
return total;
|
|
35
|
+
}
|
|
36
|
+
/**
|
|
37
|
+
* `files` (change `add-tensorrt-llm-windows`, task 2.8): on Windows the model and its caches are in the
|
|
38
|
+
* WSL guest, and are sized and removed by commands there (`guestModelFiles`) — all caches and the model
|
|
39
|
+
* in one call each, the caches listed first. Without it, the walk below, as on Linux.
|
|
40
|
+
*/
|
|
41
|
+
export async function deleteTensorrtLlmModelFiles(paths, model, files) {
|
|
42
|
+
if (files !== undefined) {
|
|
43
|
+
const caches = await engineCacheDirsOf(paths, model.id);
|
|
44
|
+
const sizes = await files.sizes([...caches, model.dir]);
|
|
45
|
+
await files.remove([...caches, model.dir]);
|
|
46
|
+
const freed = [...sizes.values()].reduce((sum, bytes) => sum + bytes, 0);
|
|
47
|
+
return { freedBytes: freed, engineCachesRemoved: caches.length };
|
|
48
|
+
}
|
|
49
|
+
const seen = new Set();
|
|
50
|
+
const caches = await engineCacheDirsOf(paths, model.id);
|
|
51
|
+
const sizes = new Map();
|
|
52
|
+
for (const dir of caches)
|
|
53
|
+
sizes.set(dir, await bytesUnder(dir, seen));
|
|
54
|
+
const modelBytes = await bytesUnder(model.dir, seen);
|
|
55
|
+
const removed = await removeEngineCaches(paths, { modelId: model.id });
|
|
56
|
+
await rm(model.dir, { recursive: true, force: true });
|
|
57
|
+
const cacheBytes = removed.reduce((sum, dir) => sum + (sizes.get(dir) ?? 0), 0);
|
|
58
|
+
return { freedBytes: cacheBytes + modelBytes, engineCachesRemoved: removed.length };
|
|
59
|
+
}
|
|
60
|
+
//# sourceMappingURL=delete.js.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"delete.js","sourceRoot":"","sources":["../../../src/runtime/tensorrt-llm/delete.ts"],"names":[],"mappings":"AAAA;;;;;;;;GAQG;AACH,OAAO,EAAE,KAAK,EAAE,OAAO,EAAE,EAAE,EAAE,MAAM,kBAAkB,CAAA;AACrD,OAAO,EAAE,IAAI,EAAE,MAAM,WAAW,CAAA;AAEhC,OAAO,EAAE,iBAAiB,EAAE,kBAAkB,EAAE,MAAM,0BAA0B,CAAA;AAQhF;;;GAGG;AACH,KAAK,UAAU,UAAU,CAAC,IAAY,EAAE,IAAiB;IACvD,MAAM,IAAI,GAAG,MAAM,KAAK,CAAC,IAAI,CAAC,CAAC,KAAK,CAAC,GAAG,EAAE,CAAC,SAAS,CAAC,CAAA;IACrD,IAAI,IAAI,KAAK,SAAS;QAAE,OAAO,CAAC,CAAA;IAChC,IAAI,IAAI,CAAC,MAAM,EAAE,EAAE,CAAC;QAClB,MAAM,KAAK,GAAG,GAAG,IAAI,CAAC,GAAG,IAAI,IAAI,CAAC,GAAG,EAAE,CAAA;QACvC,IAAI,IAAI,CAAC,GAAG,CAAC,KAAK,CAAC;YAAE,OAAO,CAAC,CAAA;QAC7B,IAAI,CAAC,GAAG,CAAC,KAAK,CAAC,CAAA;QACf,OAAO,IAAI,CAAC,IAAI,CAAA;IAClB,CAAC;IACD,IAAI,CAAC,IAAI,CAAC,WAAW,EAAE;QAAE,OAAO,CAAC,CAAA;IACjC,MAAM,QAAQ,GAAG,MAAM,OAAO,CAAC,IAAI,CAAC,CAAC,KAAK,CAAC,GAAG,EAAE,CAAC,EAAE,CAAC,CAAA;IACpD,IAAI,KAAK,GAAG,CAAC,CAAA;IACb,KAAK,MAAM,KAAK,IAAI,QAAQ;QAAE,KAAK,IAAI,MAAM,UAAU,CAAC,IAAI,CAAC,IAAI,EAAE,KAAK,CAAC,EAAE,IAAI,CAAC,CAAA;IAChF,OAAO,KAAK,CAAA;AACd,CAAC;AAED;;;;GAIG;AACH,MAAM,CAAC,KAAK,UAAU,2BAA2B,CAC/C,KAAwB,EACxB,KAAkC,EAClC,KAAoB;IAEpB,IAAI,KAAK,KAAK,SAAS,EAAE,CAAC;QACxB,MAAM,MAAM,GAAG,MAAM,iBAAiB,CAAC,KAAK,EAAE,KAAK,CAAC,EAAE,CAAC,CAAA;QACvD,MAAM,KAAK,GAAG,MAAM,KAAK,CAAC,KAAK,CAAC,CAAC,GAAG,MAAM,EAAE,KAAK,CAAC,GAAG,CAAC,CAAC,CAAA;QACvD,MAAM,KAAK,CAAC,MAAM,CAAC,CAAC,GAAG,MAAM,EAAE,KAAK,CAAC,GAAG,CAAC,CAAC,CAAA;QAC1C,MAAM,KAAK,GAAG,CAAC,GAAG,KAAK,CAAC,MAAM,EAAE,CAAC,CAAC,MAAM,CAAC,CAAC,GAAG,EAAE,KAAK,EAAE,EAAE,CAAC,GAAG,GAAG,KAAK,EAAE,CAAC,CAAC,CAAA;QACxE,OAAO,EAAE,UAAU,EAAE,KAAK,EAAE,mBAAmB,EAAE,MAAM,CAAC,MAAM,EAAE,CAAA;IAClE,CAAC;IACD,MAAM,IAAI,GAAG,IAAI,GAAG,EAAU,CAAA;IAC9B,MAAM,MAAM,GAAG,MAAM,iBAAiB,CAAC,KAAK,EAAE,KAAK,CAAC,EAAE,CAAC,CAAA;IACvD,MAAM,KAAK,GAAG,IAAI,GAAG,EAAkB,CAAA;IACvC,KAAK,MAAM,GAAG,IAAI,MAAM;QAAE,KAAK,CAAC,GAAG,CAAC,GAAG,EAAE,MAAM,UAAU,CAAC,GAAG,EAAE,IAAI,CAAC,CAAC,CAAA;IACrE,MAAM,UAAU,GAAG,MAAM,UAAU,CAAC,KAAK,CAAC,GAAG,EAAE,IAAI,CAAC,CAAA;IAEpD,MAAM,OAAO,GAAG,MAAM,kBAAkB,CAAC,KAAK,EAAE,EAAE,OAAO,EAAE,KAAK,CAAC,EAAE,EAAE,CAAC,CAAA;IACtE,MAAM,EAAE,CAAC,KAAK,CAAC,GAAG,EAAE,EAAE,SAAS,EAAE,IAAI,EAAE,KAAK,EAAE,IAAI,EAAE,CAAC,CAAA;IACrD,MAAM,UAAU,GAAG,OAAO,CAAC,MAAM,CAAC,CAAC,GAAG,EAAE,GAAG,EAAE,EAAE,CAAC,GAAG,GAAG,CAAC,KAAK,CAAC,GAAG,CAAC,GAAG,CAAC,IAAI,CAAC,CAAC,EAAE,CAAC,CAAC,CAAA;IAC/E,OAAO,EAAE,UAAU,EAAE,UAAU,GAAG,UAAU,EAAE,mBAAmB,EAAE,OAAO,CAAC,MAAM,EAAE,CAAA;AACrF,CAAC"}
|