knowledge: audit DeepSeek C expert-cache byte validity

docs/reference-matrix.md

@@ -546,3 +546,9 @@ create a repo runtime descriptor, backend port or measured resource effect.
 | Source / revision | Inspected scope | Extraction and qualification boundary |
 | --- | --- | --- |
 | pytorch/pytorch @ `9a90a4fefcaafa63ac984621963a6fcc66aee68c`, source version `2.16.0a0` | Bounded implementation, documentation and test-definition spans across installation/build, tensor/dispatch/autograd, training/data, compiler/export, memory/streams/graphs, distributed/checkpoint, profiling and custom operators. [Exact readset](../src/inferrsi/data/skill_library/frameworks/pytorch/pytorch/references/source-readset.json) retains Git blob and inspected checkout identities separately; [directory coverage](../src/inferrsi/data/skill_library/frameworks/pytorch/pytorch/references/source-directory-coverage.md) distinguishes tree discovery from substantive inspection and unpopulated dependencies | [pytorch candidate graph](../src/inferrsi/data/skill_library/frameworks/pytorch/pytorch/SKILL.md), eight child procedures, bounded public-API helpers and exact LICENSE/NOTICE bytes. Resolver `NOASSERTION` is retained separately from source terms. Preparation and CPU observations use the distinct `2.15.0+cpu` wheel, not this source build. Required accelerator, native extension, distributed-device and deployment checks remain blocked/not_run; no quality/service/performance or efficacy claim, candidate promotion, managed DisCo import or implicit TorchNPU compatibility pairing. |
+
+## DeepSeek streamed expert-cache oracle contracts
+
+| Source | Inspected scope | Canonical extraction and boundary |
+| --- | --- | --- |
+| [ronak-create/deepseek-v4-in-c](https://github.com/ronak-create/deepseek-v4-in-c/tree/d0402056d0da4bf7edfd2d4977de479a0db9d201) @ `d0402056d0da4bf7edfd2d4977de479a0db9d201` (Apache-2.0 root and selected-file declarations; third-party/model rights excluded) | Bounded source inspection, 2026-11-03: byte/fixture geometry, begin/end slot lifetime, actual model/oracle callback wiring and tiny-reference limits; [selected hashes, regions and rights](../src/inferrsi/data/skill_library/methods/weight-loading-optimization/references/deepseek-expert-cache-source.json) | [weight-loading-optimization](../src/inferrsi/data/skill_library/methods/weight-loading-optimization/SKILL.md) ([expert-cache oracle reference](../src/inferrsi/data/skill_library/methods/weight-loading-optimization/references/deepseek-expert-cache-oracles.md)): independent buffered bytes, usable residents during misses, ordered accumulation and oracle reachability. Source-derived candidate/not_run; historical corruption reports are not reproduced tests, and tiny fixtures do not establish released-model quality or performance. |

docs/searchspace-matrix.md

@@ -614,3 +614,9 @@ NOASSERTION and InferRSI's repository license.
 | Accelerator / RAM / Host | Diagnose / configure / verify: tensor/alias/gradient state, training and data ownership, compiler guards and export contracts, allocator accounting, stream dependencies and graph input/output lifetimes | [pytorch operating graph](../src/inferrsi/data/skill_library/frameworks/pytorch/pytorch/SKILL.md), [memory/graph route](../src/inferrsi/data/skill_library/frameworks/pytorch/pytorch/sub-skills/memory-streams-and-graphs/SKILL.md), [compiler/export route](../src/inferrsi/data/skill_library/frameworks/pytorch/pytorch/sub-skills/compiler-and-export/SKILL.md) | Source version `2.16.0a0` and observed CPU wheel `2.15.0+cpu` are distinct. CPU observations do not qualify pinned-source execution, CUDA/ROCm/MPS/XPU, native custom operators, memory fit or model quality. |
 | SSD / Storage / RAM | Diagnose / verify: serialization trust and state identity, checkpoint ownership, distributed restore and capture/staging/upload completion boundaries | `pytorch`: [distributed/checkpoint route](../src/inferrsi/data/skill_library/frameworks/pytorch/pytorch/sub-skills/distributed-and-checkpoint/SKILL.md); existing [DCP async-save owner](../src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/SKILL.md) retains its separate source pin | No source-build checkpoint run, backend durability, distributed recovery or measured storage benefit is established. A state-dict round trip in a separate installed wheel does not qualify those properties. |
 | Communication / Network / Accelerator | Diagnose / verify: process/rank/device mapping, backend and collective schedule, DDP/FSDP2/DTensor state and completion | `pytorch`: [distributed/checkpoint route](../src/inferrsi/data/skill_library/frameworks/pytorch/pytorch/sub-skills/distributed-and-checkpoint/SKILL.md), [source and runtime boundary](../src/inferrsi/data/skill_library/frameworks/pytorch/pytorch/references/repo-provenance.md) | Selected source and CPU controls do not establish NCCL or other accelerator transport, multi-device/multi-node correctness, overlap or throughput. Required backend checks remain `BLOCKED_REQUIRED_BACKEND`; no new measured/green resource cell, OS control or compatible latest PyTorch/TorchNPU stack is asserted. |
+
+## DeepSeek streamed expert-cache verification
+
+| Resource | Stage / scoped question | Canonical owner | Evidence boundary |
+| --- | --- | --- | --- |
+| SSD / Storage + RAM / Host | Diagnose / Verify: independent tensor-byte identity after aligned/coalesced reads, resident versus pending slot lifetime, ordered consumer accumulation and actual oracle callback coverage | [weight-loading-optimization](../src/inferrsi/data/skill_library/methods/weight-loading-optimization/SKILL.md) ([expert-cache oracle reference](../src/inferrsi/data/skill_library/methods/weight-loading-optimization/references/deepseek-expert-cache-oracles.md)) and [source record](../src/inferrsi/data/skill_library/methods/weight-loading-optimization/references/deepseek-expert-cache-source.json); DeepSeek source at `d0402056d0da4bf7edfd2d4977de479a0db9d201` | Source-derived candidate / empirical not_run; byte oracle shares tensor metadata, resident snapshot covers w1, and tiny PyTorch/batched checks are source definitions. No measured resource gain, full-model quality, native/device execution or agent efficacy is established. |

docs/skill-coverage.md

@@ -4,7 +4,7 @@ Generated from the registry, the skill sidecars and the two maintained
 matrices. **Do not edit by hand**; regenerate after any registry, skill,
 matrix or frozen-plan change.
 
-- Input digest: `sha256:7640b171256091efd1514a33d5e434eef3feaed238fc92fe3eeba02fee94e9ff`
+- Input digest: `sha256:b09848a0636bd5f34e2876c351050600b58903b48cc68c476fceffde4836ccec`
 - Regenerate: `python tools/build_skill_coverage.py`
 - Verify without writing: `python tools/build_skill_coverage.py --check`
 
@@ -142,7 +142,7 @@ optimization is fully covered or that any procedure has measured benefit.
 | [training-input-pipeline](../src/inferrsi/data/skill_library/methods/training-input-pipeline/SKILL.md) | common | Diagnose training input stalls and choose bounded queue, shard-cache and cache-ownership changes from DALI, Mosaic ... | Diagnosing training input stalls and choosing bounded queue, shard-cache and ... | failure_modes, input_contract, produced_artifact, refusal_point<br>4 advice | 1 Accelerator · Observe / Configure / Optimize<br>2 RAM/Host · Observe / Diagnose / Configure / Optimize<br>3 SSD · Diagnose / Configure / Optimize<br>4 Network · Configure / Optimize<br>5 OS · Configure / Optimize | 1 · Systems sources (B1 / NVIDIA DALI @ `6fe47c48d3b0ca335f5e1afd3c77c750b2c7e041` ...)<br>1 · Systems sources (B2 / Mosaic Streaming @ `d99bf9c7cdf2dd4ed62a4960ed35270b93337a5f` ...)<br>1 · Systems sources (B3 / FFCV @ `3a12966b3afe3a81733a732e633317d747bfaac7` ...) | candidate · not_run · 1 readset(s) | no-eval |
 | [training-resource-budget](../src/inferrsi/data/skill_library/methods/training-resource-budget/SKILL.md) | common | Choose training offload and resource-budget changes using pinned DeepSpeed configuration, attributed expert diagnosis ... | Choosing training offload and resource-budget changes from pinned DeepSpeed ... | failure_modes, input_contract, produced_artifact, refusal_point<br>4 advice | 1 Accelerator · Diagnose / Configure / Optimize<br>2 RAM/Host · Observe / Diagnose / Configure / Optimize / Verify<br>3 SSD · Configure / Optimize<br>4 Network · Observe / Diagnose / Verify<br>5 OS · Diagnose / Configure | 1 · Systems sources (CR1 / DeepSpeed ZeRO-Infinity @ `5f266e26d216a705d9c64cac88fb813ce ...)<br>1 · Systems sources (CR5 / Ascend MindSpeed adaptive memory — historical `core_r0.8.0` ...)<br>1 · Systems sources (ZeRO-Offload — [2101.06840v1](https://arxiv.org/abs/2101.06840v1))<br>2 · Methodology and corpus references (CR6 / Stas Bekman ML Engineering Open Book @ ...) | candidate · not_run · 9 readset(s) | no-eval |
 | [video-diffusion-optimization](../src/inferrsi/data/skill_library/methods/video-diffusion-optimization/SKILL.md) | common | Run and verify goal-driven video diffusion optimization, backend-agnostic. Use when the user asks for latency, RTF ... | Backend-agnostic video diffusion inference and serving optimization | — | 1 Accelerator † guidance: Video diffusion campaign guidance (unvalidated) | — | candidate · not_run · 0 readset(s) | no-Reference-Matrix-row, no-eval |
-| [weight-loading-optimization](../src/inferrsi/data/skill_library/methods/weight-loading-optimization/SKILL.md) | common | Optimize the serving-level weight load path from storage to device: judge the load path on four dimensions (source ... | Diagnosing the serving-level weight load path from storage to device and choosing changes ... | failure_modes, input_contract, produced_artifact, refusal_point<br>5 advice | 1 Accelerator † guidance: Flare preparation and complete-session source contracts<br>2 RAM/Host † guidance: Flare preparation and complete-session source contracts<br>3 SSD · Observe / Profile / Diagnose / Optimize / Verify † guidance: Flare preparation and complete-session source contracts<br>4 Network · Configure / Optimize / Verify | 1 · Systems sources (Ascend/MindIE-SD @ 8481d39)<br>1 · Systems sources (NVIDIA/TensorRT-LLM @ f33e0e5fd)<br>1 · Systems sources (ascend-ai-coding/awesome-ascend-skills @ 75ec89c)<br>1 · Systems sources (foundation-model-stack/fastsafetensors @ 219e256)<br>1 · Systems sources (ggml-org/llama.cpp @ `3e390af18de1131dda27913c00511725b4274c8a` (MIT ...)<br>1 · Systems sources (mindie-agent/mindie-agent @ 6a81b41) | candidate · not_run · 3 readset(s) | no-eval |
+| [weight-loading-optimization](../src/inferrsi/data/skill_library/methods/weight-loading-optimization/SKILL.md) | common | Optimize the serving-level weight load path from storage to device: judge the load path on four dimensions (source ... | Diagnosing the serving-level weight load path from storage to device and choosing changes ... | failure_modes, input_contract, produced_artifact, refusal_point<br>6 advice | 1 Accelerator † guidance: Flare preparation and complete-session source contracts<br>2 RAM/Host † guidance: Flare preparation and complete-session source contracts; DeepSeek streamed expert-cache verification<br>3 SSD · Observe / Profile / Diagnose / Optimize / Verify † guidance: Flare preparation and complete-session source contracts; DeepSeek streamed expert-cache verification<br>4 Network · Configure / Optimize / Verify | 1 · Systems sources (Ascend/MindIE-SD @ 8481d39)<br>1 · Systems sources (NVIDIA/TensorRT-LLM @ f33e0e5fd)<br>1 · Systems sources (ascend-ai-coding/awesome-ascend-skills @ 75ec89c)<br>1 · Systems sources (foundation-model-stack/fastsafetensors @ 219e256)<br>1 · Systems sources (ggml-org/llama.cpp @ `3e390af18de1131dda27913c00511725b4274c8a` (MIT ...)<br>1 · Systems sources (mindie-agent/mindie-agent @ 6a81b41)<br>DeepSeek streamed expert-cache oracle contracts ([ronak-create/deepseek-v4-in-c](https://github.com/ronak-create/deepse ...) | candidate · not_run · 5 readset(s) | no-eval |
 
 ## Registered: Workflows (6)
 

src/inferrsi/data/repo_packs/inferrsi/knowledge/README.md

@@ -15,7 +15,7 @@ These are knowledge abstracts (`--kind knowledge`), with namespaced package
 IDs distinct from explicitly selected `checkout_ref` skills. Ascend/CUDA
 procedures use their corresponding backend; shared methods use `common`.
 
-Source snapshot: [7aa3344bf84c42467f43b3a927e1c7d13642bd55](https://github.com/InferMatrixLab/InferRSI/tree/7aa3344bf84c42467f43b3a927e1c7d13642bd55/src/inferrsi/data/skill_library).
+Source snapshot: [da3356920acac5bbf6f3ac6a4fd6de636a6d1c1c](https://github.com/InferMatrixLab/InferRSI/tree/da3356920acac5bbf6f3ac6a4fd6de636a6d1c1c/src/inferrsi/data/skill_library).
 
 ```bash
 inferrsi knowledge search 'profiling' --repo-type inferrsi --limit 50
@@ -29,7 +29,7 @@ contain the current input paths and bytes (Git checkout filters are respected).
 
 ```bash
 python tools/build_skill_index.py --source-revision <committed-source-revision>
-python tools/build_skill_index.py --source-revision 7aa3344bf84c42467f43b3a927e1c7d13642bd55 --check
+python tools/build_skill_index.py --source-revision da3356920acac5bbf6f3ac6a4fd6de636a6d1c1c --check
 ```
 
 `--check` detects missing, stale and unexpected generated records without writing.

src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-aclgraph-capture-replay.json

@@ -21,8 +21,8 @@
   "sources": [
     {
       "type": "git",
-      "url": "https://github.com/InferMatrixLab/InferRSI/blob/7aa3344bf84c42467f43b3a927e1c7d13642bd55/src/inferrsi/data/skill_library/hardware/ascend/ascend-aclgraph-capture-replay/SKILL.md",
-      "commit": "7aa3344bf84c42467f43b3a927e1c7d13642bd55",
+      "url": "https://github.com/InferMatrixLab/InferRSI/blob/da3356920acac5bbf6f3ac6a4fd6de636a6d1c1c/src/inferrsi/data/skill_library/hardware/ascend/ascend-aclgraph-capture-replay/SKILL.md",
+      "commit": "da3356920acac5bbf6f3ac6a4fd6de636a6d1c1c",
       "paths": [
         "src/inferrsi/data/skill_library/hardware/ascend/ascend-aclgraph-capture-replay/SKILL.md",
         "src/inferrsi/data/skill_library/hardware/ascend/ascend-aclgraph-capture-replay/metadata.json",

src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-device-health.json

@@ -23,8 +23,8 @@
   "sources": [
     {
       "type": "git",
-      "url": "https://github.com/InferMatrixLab/InferRSI/blob/7aa3344bf84c42467f43b3a927e1c7d13642bd55/src/inferrsi/data/skill_library/hardware/ascend/ascend-device-health/SKILL.md",
-      "commit": "7aa3344bf84c42467f43b3a927e1c7d13642bd55",
+      "url": "https://github.com/InferMatrixLab/InferRSI/blob/da3356920acac5bbf6f3ac6a4fd6de636a6d1c1c/src/inferrsi/data/skill_library/hardware/ascend/ascend-device-health/SKILL.md",
+      "commit": "da3356920acac5bbf6f3ac6a4fd6de636a6d1c1c",
       "paths": [
         "src/inferrsi/data/skill_library/hardware/ascend/ascend-device-health/SKILL.md",
         "src/inferrsi/data/skill_library/hardware/ascend/ascend-device-health/metadata.json",

src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-distributed-parallel.json

@@ -27,8 +27,8 @@
   "sources": [
     {
       "type": "git",
-      "url": "https://github.com/InferMatrixLab/InferRSI/blob/7aa3344bf84c42467f43b3a927e1c7d13642bd55/src/inferrsi/data/skill_library/hardware/ascend/ascend-distributed-parallel/SKILL.md",
-      "commit": "7aa3344bf84c42467f43b3a927e1c7d13642bd55",
+      "url": "https://github.com/InferMatrixLab/InferRSI/blob/da3356920acac5bbf6f3ac6a4fd6de636a6d1c1c/src/inferrsi/data/skill_library/hardware/ascend/ascend-distributed-parallel/SKILL.md",
+      "commit": "da3356920acac5bbf6f3ac6a4fd6de636a6d1c1c",
       "paths": [
         "src/inferrsi/data/skill_library/hardware/ascend/ascend-distributed-parallel/SKILL.md",
         "src/inferrsi/data/skill_library/hardware/ascend/ascend-distributed-parallel/metadata.json",

src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-graph-mode.json

@@ -26,8 +26,8 @@
   "sources": [
     {
       "type": "git",
-      "url": "https://github.com/InferMatrixLab/InferRSI/blob/7aa3344bf84c42467f43b3a927e1c7d13642bd55/src/inferrsi/data/skill_library/hardware/ascend/ascend-graph-mode/SKILL.md",
-      "commit": "7aa3344bf84c42467f43b3a927e1c7d13642bd55",
+      "url": "https://github.com/InferMatrixLab/InferRSI/blob/da3356920acac5bbf6f3ac6a4fd6de636a6d1c1c/src/inferrsi/data/skill_library/hardware/ascend/ascend-graph-mode/SKILL.md",
+      "commit": "da3356920acac5bbf6f3ac6a4fd6de636a6d1c1c",
       "paths": [
         "src/inferrsi/data/skill_library/hardware/ascend/ascend-graph-mode/SKILL.md",
         "src/inferrsi/data/skill_library/hardware/ascend/ascend-graph-mode/metadata.json",

src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-model-adaptation.json

@@ -22,8 +22,8 @@
   "sources": [
     {
       "type": "git",
-      "url": "https://github.com/InferMatrixLab/InferRSI/blob/7aa3344bf84c42467f43b3a927e1c7d13642bd55/src/inferrsi/data/skill_library/hardware/ascend/ascend-model-adaptation/SKILL.md",
-      "commit": "7aa3344bf84c42467f43b3a927e1c7d13642bd55",
+      "url": "https://github.com/InferMatrixLab/InferRSI/blob/da3356920acac5bbf6f3ac6a4fd6de636a6d1c1c/src/inferrsi/data/skill_library/hardware/ascend/ascend-model-adaptation/SKILL.md",
+      "commit": "da3356920acac5bbf6f3ac6a4fd6de636a6d1c1c",
       "paths": [
         "src/inferrsi/data/skill_library/hardware/ascend/ascend-model-adaptation/SKILL.md",
         "src/inferrsi/data/skill_library/hardware/ascend/ascend-model-adaptation/metadata.json",

src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-operator-profiling.json

@@ -23,8 +23,8 @@
   "sources": [
     {
       "type": "git",
-      "url": "https://github.com/InferMatrixLab/InferRSI/blob/7aa3344bf84c42467f43b3a927e1c7d13642bd55/src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-profiling/SKILL.md",
-      "commit": "7aa3344bf84c42467f43b3a927e1c7d13642bd55",
+      "url": "https://github.com/InferMatrixLab/InferRSI/blob/da3356920acac5bbf6f3ac6a4fd6de636a6d1c1c/src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-profiling/SKILL.md",
+      "commit": "da3356920acac5bbf6f3ac6a4fd6de636a6d1c1c",
       "paths": [
         "src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-profiling/SKILL.md",
         "src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-profiling/metadata.json",

src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-operator-tuning.json

@@ -24,8 +24,8 @@
   "sources": [
     {
       "type": "git",
-      "url": "https://github.com/InferMatrixLab/InferRSI/blob/7aa3344bf84c42467f43b3a927e1c7d13642bd55/src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-tuning/SKILL.md",
-      "commit": "7aa3344bf84c42467f43b3a927e1c7d13642bd55",
+      "url": "https://github.com/InferMatrixLab/InferRSI/blob/da3356920acac5bbf6f3ac6a4fd6de636a6d1c1c/src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-tuning/SKILL.md",
+      "commit": "da3356920acac5bbf6f3ac6a4fd6de636a6d1c1c",
       "paths": [
         "src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-tuning/SKILL.md",
         "src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-tuning/metadata.json",

src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-optimization-casebook.json

@@ -22,8 +22,8 @@
   "sources": [
     {
       "type": "git",
-      "url": "https://github.com/InferMatrixLab/InferRSI/blob/7aa3344bf84c42467f43b3a927e1c7d13642bd55/src/inferrsi/data/skill_library/hardware/ascend/ascend-optimization-casebook/SKILL.md",
-      "commit": "7aa3344bf84c42467f43b3a927e1c7d13642bd55",
+      "url": "https://github.com/InferMatrixLab/InferRSI/blob/da3356920acac5bbf6f3ac6a4fd6de636a6d1c1c/src/inferrsi/data/skill_library/hardware/ascend/ascend-optimization-casebook/SKILL.md",
+      "commit": "da3356920acac5bbf6f3ac6a4fd6de636a6d1c1c",
       "paths": [
         "src/inferrsi/data/skill_library/hardware/ascend/ascend-optimization-casebook/SKILL.md",
         "src/inferrsi/data/skill_library/hardware/ascend/ascend-optimization-casebook/metadata.json",

src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-performance-analysis.json

@@ -22,8 +22,8 @@
   "sources": [
     {
       "type": "git",
-      "url": "https://github.com/InferMatrixLab/InferRSI/blob/7aa3344bf84c42467f43b3a927e1c7d13642bd55/src/inferrsi/data/skill_library/hardware/ascend/ascend-performance-analysis/SKILL.md",
-      "commit": "7aa3344bf84c42467f43b3a927e1c7d13642bd55",
+      "url": "https://github.com/InferMatrixLab/InferRSI/blob/da3356920acac5bbf6f3ac6a4fd6de636a6d1c1c/src/inferrsi/data/skill_library/hardware/ascend/ascend-performance-analysis/SKILL.md",
+      "commit": "da3356920acac5bbf6f3ac6a4fd6de636a6d1c1c",
       "paths": [
         "src/inferrsi/data/skill_library/hardware/ascend/ascend-performance-analysis/SKILL.md",
         "src/inferrsi/data/skill_library/hardware/ascend/ascend-performance-analysis/metadata.json",