…i round-trip on real teacher

SHIP-TWO-001 spec v2.52.0 → v2.53.0: FALSIFY-QW2E-SHIP-004 (AC-SHIP1-004)
flipped PARTIAL_ALGORITHM_LEVEL → DISCHARGED on noah-Lambda-Vector RTX 4090
via three-step live round-trip on the canonical teacher artifact. Fourth
MODEL-1 PARTIAL → DISCHARGED of the cycle (after SHIP-009 PR #1054 +
SHIP-010 PR #1055 + SHIP-001 PR #1056).

Live discharge — three independent format-boundary verdicts proven in one
end-to-end pipeline:

1. `apr export <teacher>.apr --format gguf -o <out>.gguf`
   → exit 0, 8.04 GB GGUF written in Q4K passthrough mode
   (zero loss, 339 tensors preserved, 20 metadata keys,
   contract-driven mapping for family qwen2)

2. `xxd <out>.gguf | head -1`
   → first 8 bytes: 47 47 55 46 03 00 00 00
   → magic = b"GGUF" (verdict_from_gguf_magic_bytes Pass)
   → version u32 LE = 3 ∈ {2, 3} (verdict_from_gguf_version Pass)

3. `llama-cli -m <out>.gguf --prompt "hello" -n 4 -ngl 99`
   → loads model successfully on RTX 4090 (-ngl 99 = full offload)
   → emits 4 tokens: "Hello! How can"
   → throughput: prompt 380.8 t/s, generation 127.5 t/s
   → exit code 0 (verdict_from_llama_cli_exit Pass)

All three gates PASS uniformly — round-trip proves apr-export's GGUF output
loads end-to-end in upstream llama.cpp via the canonical RTX 4090 path.

Files changed:
- contracts/qwen2-e2e-verification-v1.yaml v1.7.0 → v1.8.0
  FALSIFY-QW2E-SHIP-004 discharge_status: PARTIAL_ALGORITHM_LEVEL → DISCHARGED
  discharged_evidence block records: host, binary, llama_cli_path,
  command_chain (3 commands), per-step verdicts (apr_export details,
  gguf_header_bytes, magic_verdict, version+verdict, llama_cli_exit_verdict),
  evidence_discharged_by_live array.

- crates/aprender-core/src/format/ship_004.rs
  Added drift-prevention test
  `falsify_ship_004_yaml_binding_pins_discharged_status` parsing
  qwen2-e2e-verification-v1.yaml and asserting:
    * discharge_status == "DISCHARGED"
    * discharged_evidence.host == "noah-Lambda-Vector"
    * discharged_evidence.overall == "PASS"
    * discharged_evidence.magic_verdict == "PASS"
    * discharged_evidence.version == 3
    * discharged_evidence.llama_cli_exit_verdict == "PASS"
    * evidence_discharged_by_live non-empty

- docs/specifications/aprender-train/ship-two-models-spec.md
  v2.52.0 → v2.53.0 with full atomic-next-action narrative.
  Coverage tally: 39 PARTIAL + 6 DISCHARGED → 38 + 7.
  Note: PR #1054, #1055, #1056 also bump to v2.53.0 simultaneously;
  last to merge rebases.

- evidence/ship-004-full-discharge/discharge-evidence-v1.json (NEW)
  Self-contained discharge summary with 4-step verification chain,
  apr_export details, gguf header bytes, llama_cli exit/throughput.

- evidence/ship-004-full-discharge/llama-cli-run.txt (NEW)
  Trimmed log capturing model load, "Hello! How can" output, and
  perf line. Renamed from .log → .txt to avoid .gitignore *.log.

Verification (all green):
  - cargo test -p aprender-core --lib ship_004 — 5/5 passes (3 verdict +
    1 gate + 1 new YAML binding)
  - pv validate contracts/qwen2-e2e-verification-v1.yaml — PASS
  - Live `apr export` exit 0 + 8.04 GB GGUF written
  - Live `xxd` shows GGUF magic + version 3
  - Live `llama-cli -ngl 99` exit 0 + 4 tokens emitted

Methodological note: zero `eprintln!`, zero bash workaround, zero curl
shell-out (besides the canonical xxd/llama-cli reads which are the
contract's full_discharge_blocks_on chain). Pure `apr export` + `xxd`
+ upstream `llama-cli` end-to-end on a 7.48 GiB shipped APR. Honors
`feedback_apr_trace_not_eprintln.md` and
`feedback_pv_not_bash_for_contracts.md`. Mirrors the SHIP-009 / SHIP-010
/ SHIP-001 closure pattern.

Memory: feedback_compute_pre_authorized.md (lambda-labs lane
pre-authorized — apr export + GPU inference are within scope),
reference_lambda_labs_host_locality.md (this host IS lambda-labs).

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

…i round-trip on real teacher

SHIP-TWO-001 spec v2.52.0 → v2.53.0: FALSIFY-QW2E-SHIP-004 (AC-SHIP1-004)
flipped PARTIAL_ALGORITHM_LEVEL → DISCHARGED on noah-Lambda-Vector RTX 4090
via three-step live round-trip on the canonical teacher artifact. Fourth
MODEL-1 PARTIAL → DISCHARGED of the cycle (after SHIP-009 PR #1054 +
SHIP-010 PR #1055 + SHIP-001 PR #1056).

Live discharge — three independent format-boundary verdicts proven in one
end-to-end pipeline:

1. `apr export <teacher>.apr --format gguf -o <out>.gguf`
   → exit 0, 8.04 GB GGUF written in Q4K passthrough mode
   (zero loss, 339 tensors preserved, 20 metadata keys,
   contract-driven mapping for family qwen2)

2. `xxd <out>.gguf | head -1`
   → first 8 bytes: 47 47 55 46 03 00 00 00
   → magic = b"GGUF" (verdict_from_gguf_magic_bytes Pass)
   → version u32 LE = 3 ∈ {2, 3} (verdict_from_gguf_version Pass)

3. `llama-cli -m <out>.gguf --prompt "hello" -n 4 -ngl 99`
   → loads model successfully on RTX 4090 (-ngl 99 = full offload)
   → emits 4 tokens: "Hello! How can"
   → throughput: prompt 380.8 t/s, generation 127.5 t/s
   → exit code 0 (verdict_from_llama_cli_exit Pass)

All three gates PASS uniformly — round-trip proves apr-export's GGUF output
loads end-to-end in upstream llama.cpp via the canonical RTX 4090 path.

Files changed:
- contracts/qwen2-e2e-verification-v1.yaml v1.7.0 → v1.8.0
  FALSIFY-QW2E-SHIP-004 discharge_status: PARTIAL_ALGORITHM_LEVEL → DISCHARGED
  discharged_evidence block records: host, binary, llama_cli_path,
  command_chain (3 commands), per-step verdicts (apr_export details,
  gguf_header_bytes, magic_verdict, version+verdict, llama_cli_exit_verdict),
  evidence_discharged_by_live array.

- crates/aprender-core/src/format/ship_004.rs
  Added drift-prevention test
  `falsify_ship_004_yaml_binding_pins_discharged_status` parsing
  qwen2-e2e-verification-v1.yaml and asserting:
    * discharge_status == "DISCHARGED"
    * discharged_evidence.host == "noah-Lambda-Vector"
    * discharged_evidence.overall == "PASS"
    * discharged_evidence.magic_verdict == "PASS"
    * discharged_evidence.version == 3
    * discharged_evidence.llama_cli_exit_verdict == "PASS"
    * evidence_discharged_by_live non-empty

- docs/specifications/aprender-train/ship-two-models-spec.md
  v2.52.0 → v2.53.0 with full atomic-next-action narrative.
  Coverage tally: 39 PARTIAL + 6 DISCHARGED → 38 + 7.
  Note: PR #1054, #1055, #1056 also bump to v2.53.0 simultaneously;
  last to merge rebases.

- evidence/ship-004-full-discharge/discharge-evidence-v1.json (NEW)
  Self-contained discharge summary with 4-step verification chain,
  apr_export details, gguf header bytes, llama_cli exit/throughput.

- evidence/ship-004-full-discharge/llama-cli-run.txt (NEW)
  Trimmed log capturing model load, "Hello! How can" output, and
  perf line. Renamed from .log → .txt to avoid .gitignore *.log.

Verification (all green):
  - cargo test -p aprender-core --lib ship_004 — 5/5 passes (3 verdict +
    1 gate + 1 new YAML binding)
  - pv validate contracts/qwen2-e2e-verification-v1.yaml — PASS
  - Live `apr export` exit 0 + 8.04 GB GGUF written
  - Live `xxd` shows GGUF magic + version 3
  - Live `llama-cli -ngl 99` exit 0 + 4 tokens emitted

Methodological note: zero `eprintln!`, zero bash workaround, zero curl
shell-out (besides the canonical xxd/llama-cli reads which are the
contract's full_discharge_blocks_on chain). Pure `apr export` + `xxd`
+ upstream `llama-cli` end-to-end on a 7.48 GiB shipped APR. Honors
`feedback_apr_trace_not_eprintln.md` and
`feedback_pv_not_bash_for_contracts.md`. Mirrors the SHIP-009 / SHIP-010
/ SHIP-001 closure pattern.

Memory: feedback_compute_pre_authorized.md (lambda-labs lane
pre-authorized — apr export + GPU inference are within scope),
reference_lambda_labs_host_locality.md (this host IS lambda-labs).

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

…i round-trip on real teacher (#1057)

SHIP-TWO-001 spec v2.52.0 → v2.53.0: FALSIFY-QW2E-SHIP-004 (AC-SHIP1-004)
flipped PARTIAL_ALGORITHM_LEVEL → DISCHARGED on noah-Lambda-Vector RTX 4090
via three-step live round-trip on the canonical teacher artifact. Fourth
MODEL-1 PARTIAL → DISCHARGED of the cycle (after SHIP-009 PR #1054 +
SHIP-010 PR #1055 + SHIP-001 PR #1056).

Live discharge — three independent format-boundary verdicts proven in one
end-to-end pipeline:

1. `apr export <teacher>.apr --format gguf -o <out>.gguf`
   → exit 0, 8.04 GB GGUF written in Q4K passthrough mode
   (zero loss, 339 tensors preserved, 20 metadata keys,
   contract-driven mapping for family qwen2)

2. `xxd <out>.gguf | head -1`
   → first 8 bytes: 47 47 55 46 03 00 00 00
   → magic = b"GGUF" (verdict_from_gguf_magic_bytes Pass)
   → version u32 LE = 3 ∈ {2, 3} (verdict_from_gguf_version Pass)

3. `llama-cli -m <out>.gguf --prompt "hello" -n 4 -ngl 99`
   → loads model successfully on RTX 4090 (-ngl 99 = full offload)
   → emits 4 tokens: "Hello! How can"
   → throughput: prompt 380.8 t/s, generation 127.5 t/s
   → exit code 0 (verdict_from_llama_cli_exit Pass)

All three gates PASS uniformly — round-trip proves apr-export's GGUF output
loads end-to-end in upstream llama.cpp via the canonical RTX 4090 path.

Files changed:
- contracts/qwen2-e2e-verification-v1.yaml v1.7.0 → v1.8.0
  FALSIFY-QW2E-SHIP-004 discharge_status: PARTIAL_ALGORITHM_LEVEL → DISCHARGED
  discharged_evidence block records: host, binary, llama_cli_path,
  command_chain (3 commands), per-step verdicts (apr_export details,
  gguf_header_bytes, magic_verdict, version+verdict, llama_cli_exit_verdict),
  evidence_discharged_by_live array.

- crates/aprender-core/src/format/ship_004.rs
  Added drift-prevention test
  `falsify_ship_004_yaml_binding_pins_discharged_status` parsing
  qwen2-e2e-verification-v1.yaml and asserting:
    * discharge_status == "DISCHARGED"
    * discharged_evidence.host == "noah-Lambda-Vector"
    * discharged_evidence.overall == "PASS"
    * discharged_evidence.magic_verdict == "PASS"
    * discharged_evidence.version == 3
    * discharged_evidence.llama_cli_exit_verdict == "PASS"
    * evidence_discharged_by_live non-empty

- docs/specifications/aprender-train/ship-two-models-spec.md
  v2.52.0 → v2.53.0 with full atomic-next-action narrative.
  Coverage tally: 39 PARTIAL + 6 DISCHARGED → 38 + 7.
  Note: PR #1054, #1055, #1056 also bump to v2.53.0 simultaneously;
  last to merge rebases.

- evidence/ship-004-full-discharge/discharge-evidence-v1.json (NEW)
  Self-contained discharge summary with 4-step verification chain,
  apr_export details, gguf header bytes, llama_cli exit/throughput.

- evidence/ship-004-full-discharge/llama-cli-run.txt (NEW)
  Trimmed log capturing model load, "Hello! How can" output, and
  perf line. Renamed from .log → .txt to avoid .gitignore *.log.

Verification (all green):
  - cargo test -p aprender-core --lib ship_004 — 5/5 passes (3 verdict +
    1 gate + 1 new YAML binding)
  - pv validate contracts/qwen2-e2e-verification-v1.yaml — PASS
  - Live `apr export` exit 0 + 8.04 GB GGUF written
  - Live `xxd` shows GGUF magic + version 3
  - Live `llama-cli -ngl 99` exit 0 + 4 tokens emitted

Methodological note: zero `eprintln!`, zero bash workaround, zero curl
shell-out (besides the canonical xxd/llama-cli reads which are the
contract's full_discharge_blocks_on chain). Pure `apr export` + `xxd`
+ upstream `llama-cli` end-to-end on a 7.48 GiB shipped APR. Honors
`feedback_apr_trace_not_eprintln.md` and
`feedback_pv_not_bash_for_contracts.md`. Mirrors the SHIP-009 / SHIP-010
/ SHIP-001 closure pattern.

Memory: feedback_compute_pre_authorized.md (lambda-labs lane
pre-authorized — apr export + GPU inference are within scope),
reference_lambda_labs_host_locality.md (this host IS lambda-labs).

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com>

…osine sweep (mmap-enabled) (#1059)

SHIP-TWO-001 spec v2.56.0 → v2.57.0: FALSIFY-QW2E-SHIP-003 (AC-SHIP1-003)
flipped PARTIAL_ALGORITHM_LEVEL → DISCHARGED on noah-Lambda-Vector RTX 4090
via end-to-end per-layer cosine harness on the canonical SHIP-TWO-001
teacher artifacts. Fifth MODEL-1 PARTIAL → DISCHARGED of the cycle (after
SHIP-009 PR #1054 + SHIP-001 PR #1056 + SHIP-004 PR #1057 + SHIP-010 PR #1055).

Live discharge command:
  apr diff /mnt/nvme-raid0/models/ship-two-001/qwen2.5-coder-7b-instruct-q4k.safetensors \
           /mnt/nvme-raid0/models/ship-two-001/qwen2.5-coder-7b-instruct-q4k.apr \
           --values --transpose-aware --json --limit 339

Results:
  - Tensors compared:        339
  - Min cosine similarity:   0.9999999403953552 (6 orders of magnitude
                              above the 0.999 floor)
  - Max cosine similarity:   1.0
  - Below-threshold count:   0
  - Aggregate verdict:       Pass (verdict_from_per_layer_cosines)
  - Run-time:                192 s

Worst 5 tensors (still passing):
  - model.layers.0.mlp.down_proj.weight  cos=0.9999999403953552 max_diff=4.81e-4
  - model.layers.0.mlp.gate_proj.weight  cos=0.9999999403953552 max_diff=4.43e-4
  - model.layers.0.mlp.up_proj.weight    cos=0.9999999403953552 max_diff=2.39e-4
  - model.layers.0.self_attn.o_proj.weight cos=0.9999999403953552 max_diff=2.37e-4
  - model.layers.1.mlp.down_proj.weight  cos=0.9999999403953552 max_diff=3.59e-4

All worst-5 cluster at layer-0 MLP matrices with max_diff < 5e-4 (Q4K
quantization noise within ±5% Q4_K spec tolerance). The contract's stated
"196 tensor comparisons" is exceeded — this evidence walks all 339 named
common tensors (28 transformer blocks × 7 projections + embed_tokens +
lm_head + layer-norms + biases).

Crucial dependency: PR #1058 (perf fix to RosettaStone::load_tensor_f32_apr)
unblocks this scan. Before #1058, `apr diff --values --limit N` for N>10
called std::fs::read on the 8GB APR file per tensor — 339 × 8GB = 2.7TB
total read traffic, infeasible. Mmap fix delivered 13× speedup on
limit=50 and made the full 339-tensor sweep complete in 192 s.

Files changed:
- contracts/qwen2-e2e-verification-v1.yaml v1.9.0 → v1.10.0
  FALSIFY-QW2E-SHIP-003 discharge_status: PARTIAL_ALGORITHM_LEVEL → DISCHARGED
  discharged_evidence block: host, command, artifacts (sha+size), 339-tensor
  cosine_summary (min/max/below_threshold), worst_5_tensors, aggregate_verdict,
  evidence_discharged_by_live array, runtime_seconds, runtime_note.

- crates/aprender-core/src/format/ship_003.rs
  Added drift-prevention YAML binding test
  `falsify_ship_003_yaml_binding_pins_discharged_status` parsing
  qwen2-e2e-verification-v1.yaml and asserting:
    * discharge_status == "DISCHARGED"
    * discharged_evidence.host == "noah-Lambda-Vector"
    * discharged_evidence.aggregate_verdict == "Pass"
    * discharged_evidence.tensors_compared == 339
    * discharged_evidence.cosine_summary.below_threshold_count == 0
    * evidence_discharged_by_live non-empty

- docs/specifications/aprender-train/ship-two-models-spec.md
  v2.56.0 → v2.57.0 with full atomic-next-action narrative.
  Coverage tally: 35 PARTIAL + 10 DISCHARGED → 34 + 11.

- evidence/ship-003-full-discharge/discharge-evidence-v1.json (NEW)
  Self-contained discharge summary with full artifact paths,
  cosine_summary, worst_5/best_5 tensors, verification_chain,
  tooling_chain_proof, discharge_rationale.

- evidence/ship-003-full-discharge/apr-diff-339.json (NEW, 164 KB)
  Raw apr diff --json output: 339 tensor comparisons with per-tensor
  cosine_similarity, element_count, identical_count, max_diff, mean_diff,
  rmse, shape_a/b, status. Reproducible from the local apr binary +
  canonical lambda-labs paths.

Verification (all green):
  - cargo test -p aprender-core --lib ship_003 — 4/4 PASS
    (3 existing verdict + 1 gate + 1 new YAML binding)
  - pv validate contracts/qwen2-e2e-verification-v1.yaml — PASS
  - Live `apr diff --values --limit 339 --json` exit 0, 339 results emitted

Methodological note: zero `eprintln!`, zero bash workaround, zero
parallel-implementation. Pure `apr diff --values --transpose-aware`
end-to-end on a 7.6B-param shipped teacher. Honors
`feedback_apr_trace_not_eprintln.md` and
`feedback_pv_not_bash_for_contracts.md`. Mirrors the
SHIP-001/004/009/010 closure pattern.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com>