Repository navigation
deps: bump @docusaurus/module-type-aliases from 3.9.2 to 3.10.0 in /docs - #2
Closed
dependabot[bot] wants to merge 1 commit into
Closed
dependabot[bot] wants to merge 1 commit into
dependabot[bot] wants to merge 1 commit into
Conversation
Bumps [@docusaurus/module-type-aliases](https://github.com/facebook/docusaurus/tree/HEAD/packages/docusaurus-module-type-aliases) from 3.9.2 to 3.10.0. - [Release notes](https://github.com/facebook/docusaurus/releases) - [Changelog](https://github.com/facebook/docusaurus/blob/main/CHANGELOG.md) - [Commits](https://github.com/facebook/docusaurus/commits/v3.10.0/packages/docusaurus-module-type-aliases) --- updated-dependencies: - dependency-name: "@docusaurus/module-type-aliases" dependency-version: 3.10.0 dependency-type: direct:development update-type: version-update:semver-minor ... Signed-off-by: dependabot[bot] <support@github.com>
Contributor
Author
LabelsThe following labels could not be found: Please fix the above issues or remove invalid values from |
This was referenced May 13, 2026
Owner
|
Closing pre-Wave-15-TT-E (T-44h). Dependabot docs bump — не критичен для silicon путя. После Wave-15 submit (2026-05-17 22:00 UTC) Dependabot регенерирует свежий PR на актуальной версии. |
gHashTag
deleted the
dependabot/npm_and_yarn/docs/docusaurus/module-type-aliases-3.10.0
branch
May 15, 2026 06:57
Contributor
Author
|
OK, I won't notify you again about this release, but will get in touch when a new version is available. If you'd rather skip all updates until the next major or minor version, let me know by commenting If you change your mind, just re-open this PR and I'll resolve any conflicts on it. |
This was referenced May 15, 2026
gHashTag
added a commit
that referenced
this pull request
May 15, 2026
… ranking + W28 GO (#105) Document ID: TOPS-LEVERS-2026-05-15-027 Mission ID: TOPS-SCAN-W28-PRE-001 (предзаказ Wave-28 / L-DPC25) - Trinity working point post-Wave-27: ~55 TOPS/W (W15a) measured target - 14-rival landscape (Q4 2025 → Q2 2026): BitROM 20.8, Loihi 3 ~15, NorthPole >10, Hailo-10H 16/8, Blackhole 2.2, B200 ~7, Sohu ~150 proj. - 5-Levers matrix: Trinity = ONLY 5/5; max rival = BitROM 2.5/5 - 6 levers ranked: #1 LUT PE (Platinum ASP-DAC 2026), #2 BitROM, #3 4x4 mesh, #4 2:4 sparsity, #5 400 MHz push, #6 SG13G3/SKY90 - Wave-28 proposal: Lever Stack #1+#2 → 2.8x gain → ~150 TOPS/W silicon target (gate W28-G1) - H_W28 pre-registered Welch t-test α=0.01, n=9, Bonferroni×3 lanes, deadline 2026-09-30 - v9 HOLOGRAPHIC 2000-3000 TOPS/W projection remains intact - TAM where Trinity = leader/sole player: $120B+ by 2030 Anchor: phi^2 + phi^-2 = 3 · DOI 10.5281/zenodo.19227877 Signed-off-by: Vasilev Dmitrii <admin@t27.ai> Co-authored-by: Vasilev Dmitrii <admin@t27.ai>
This was referenced May 15, 2026
Closed
Closed
Closed
Closed
gHashTag
added a commit
that referenced
this pull request
Jun 28, 2026
mant_rnd was [MANT_BITS:0] (MANT+1 bits): mant_field[MANT:0]+1 wrapped to 0 on
carry-out, so the `> {1,all-ones}` check never fired and the exp++ was lost,
producing under-rounded results. Caught by an INDEPENDENT iverilog integer-reference
oracle (formal/gf_mul_ref_tb.v) that the Python-transcription oracle missed
(bug-equals-bug): 1376/65536 GF8 pairs. Fix: widen mant_rnd to [MANT_BITS+1:0]
so the carry is captured.
Verified [смоделировано, independent oracle #2]: GF6 exhaustive 4096/0,
GF8 exhaustive 65536/0 (was 1376 err), GF16 (HAS_INF=1) 500k random 0 errors
(Includes Inf/NaN/zero + rounding-carry edges).
Фикс rounding-carry в gf_mul_param.v: mant_rnd был MANT+1 бит, +1 к all-ones
давал wrap->0, проверка carry не срабатывала, exp++ терялся. Независимый iverilog
integer-reference (formal/gf_mul_ref_tb.v) поймал 1376/65536 GF8 пар; Python-транскрипция
промахнулась (bug-equals-bug). Фикс: widen mant_rnd до MANT+2 бит. GF6/GF8 exhaustive +
GF16 (HAS_INF=1) 500k random = 0 расхождений.
Co-Authored-By: Claude <noreply@anthropic.com>
gHashTag
added a commit
that referenced
this pull request
Jun 28, 2026
…apper (#197) * feat(fpga): gf_mul behavioral core + DSP48E1 wrapper - gf_mul_param.v: параметрическое behavioral-ядро умножения GF6..GF20 (зеркало gf_adder_param.v). Знак=XOR, сложение экспонент, нормализация, RNE(G,R,S), family-split overflow (HAS_INF), gradual-underflow напрямую от точного prod (единственное округление, без double-rounding), канонический quiet-NaN. Ширина prod=2*(MANT+1) [erratum 2606.05017 §5.5]. - gf_mul_dsp_param.v: DSP48E1-обёртка с ЯВНЫМ инстансом (UG479/UG953). OPMODE=0x05 (P=A*B), ALUMODE=0, INMODE=0, USE_DPORT=FALSE, USE_MULT=MULTIPLY, все RST* active-high, пайплайн 3 такта + конвейер метаданных/valid. Полный port-set (49 портов UG953). - conformance/gf_mul_verify.py: transcription-оракул (RTL-модель на Python) vs золотой эталон. conformance/gf_ref.py: эталон на Fraction (exact). Верификация [смоделировано]: 0 расхождений — GF6/8 exhaustive, GF12/16/20 по 300k+ random, denorm*denorm GF20 16.7M exhaustive. DSP-версия = [ТРЕБУЕТ ДЕЙСТВИЯ]: UNISIM требует Vivado co-sim + железо AX7203. compute-HW MUL = 0/83 (нет железа в песочнице). Author: Vasilev, ORCID 0009-0008-4294-6159 * fix(fpga): gf_mul_param rounding-carry (widen mant_rnd to MANT+2) mant_rnd was [MANT_BITS:0] (MANT+1 bits): mant_field[MANT:0]+1 wrapped to 0 on carry-out, so the `> {1,all-ones}` check never fired and the exp++ was lost, producing under-rounded results. Caught by an INDEPENDENT iverilog integer-reference oracle (formal/gf_mul_ref_tb.v) that the Python-transcription oracle missed (bug-equals-bug): 1376/65536 GF8 pairs. Fix: widen mant_rnd to [MANT_BITS+1:0] so the carry is captured. Verified [смоделировано, independent oracle #2]: GF6 exhaustive 4096/0, GF8 exhaustive 65536/0 (was 1376 err), GF16 (HAS_INF=1) 500k random 0 errors (Includes Inf/NaN/zero + rounding-carry edges). Фикс rounding-carry в gf_mul_param.v: mant_rnd был MANT+1 бит, +1 к all-ones давал wrap->0, проверка carry не срабатывала, exp++ терялся. Независимый iverilog integer-reference (formal/gf_mul_ref_tb.v) поймал 1376/65536 GF8 пар; Python-транскрипция промахнулась (bug-equals-bug). Фикс: widen mant_rnd до MANT+2 бит. GF6/GF8 exhaustive + GF16 (HAS_INF=1) 500k random = 0 расхождений. Co-Authored-By: Claude <noreply@anthropic.com> * fix(fpga): apply mant_rnd MANT+2 fix to DSP wrapper + faithful verify 1. gf_mul_dsp_param.v: mant_rnd reg[MANT:0] -> reg[MANT+1:0] (тот же rounding-carry фикс, что в gf_mul_param.v @59acaad; родственный файл имел тот же узкий регистр -> wrap->0 терял exp++). 2. conformance/gf_mul_verify.py: добавлен reg_mask() для моделирования фиксированной ширины reg (REG_W=MANT+2). Старая native-int версия скрывала fixed-width wrap. Re-verify: GF6/GF8 exhaustive 0, GF12/16/20 representative 300k 0. [смоделировано] 2 oracle (faithful transcription + golden Fraction). DSP48E1-эквивалентность = [ТРЕБУЕТ ДЕЙСТВИЯ]: Vivado xsim + AX7203. --------- Co-authored-by: Dmitrii Vasilev <admin@t27.ai> Co-authored-by: Claude <noreply@anthropic.com>
gHashTag
added a commit
that referenced
this pull request
Jun 28, 2026
…I-debt P0 MUL compute-SW recorded: gf_mul_param.v verified by 2 independent oracles (GF6/GF8 exhaustive + GF12/16/20 representative = 0). rounding-carry bug found by iverilog from-spec oracle #2, missed by native-int transcription (reg_mask lesson). DSP wrapper [ТРЕБУЕТ ДЕЙСТВИЯ]. CI-debt (build.zig workflow) flagged P0. Документирование в SSOT-матрице: MUL на main, верифицирован 2 oracle'ами, баг rounding-carry пойман from-spec oracle. DSP — [ТРЕБУЕТ ДЕЙСТВИЯ]. CI-долг build.zig — P0. Co-Authored-By: Claude <noreply@anthropic.com>
gHashTag
added a commit
that referenced
this pull request
Jul 14, 2026
CRITICAL FIX (C1): GF64 wrapper TX path rewritten from shift-register
to buffer+mux pattern (same as proven GF32 wrapper).
Root cause of 359/512 silicon failure:
- tx_shift had two non-blocking assignments in same always block:
1. result_ready: tx_shift <= tx_load (new data)
2. byte-send: tx_shift <= {8'h00, tx_shift[71:8]} (shift old)
- When both fired on same cycle, NBA #2 won, discarding new result.
- Fix: use 9 fixed byte registers (tx_buf0..tx_buf8) loaded by
result_ready, read-only by case(tx_idx) mux. No conflicting writes.
CI cleanup: deleted 3286 old per-design workflows (3388→102).
CI synth flags: added -nodsp, kept -abc9 only for mul designs.
New: conformance/gf64_conformance_ax7203.py — reproducible silicon harness.
gHashTag
added a commit
that referenced
this pull request
Jul 15, 2026
…80 LUT) NEW FORMAT: GF16+Quire (QF16) Storage: GF16 (E=6, M=9) = 505 LUT multiply Accumulation: binary64 Quire = 75 LUT (MEASURED!) Total: 580 LUT, 100% gradient survival Golden Ruler honest ranking for LLM training: #1 GF16+Quire: grad 100%, 580 LUT ← BEATS ALL #2 posit16: grad 95%, ~1500 LUT #3 takum16: grad 90%, ~750 LUT #4 GF16: grad 63%, 505 LUT GF16+Quire is 3.4× cheaper than posit16+Quire with SAME 100% accuracy. Innovation: use CHEAP GF16 for storage + EXACT Quire for accumulation. Fixed: sort_key now uses GRADIENT_OVERRIDE (gf16_quire=100%) Fixed: robustness dict includes gf16_quire Fixed: tapered E/M assignment for posit/takum/tekum
gHashTag
added a commit
that referenced
this pull request
Jul 15, 2026
…plan E4 DONE: GF16+ 16-element dot(1.0, 0.5) = 8.0 ✓ on silicon E1 DONE: Multi-seed noise floor (10 seeds × 4 formats): GF16: 62.7% ± 1.0% (stable) BF16: 5.1% ± 0.4% (stable but terrible) All deterministic, not seed-dependent Top 10 ranking: GF16+ #1, GF16 #2, posit16 #3 4/7 experiments done. 3 remaining need training infrastructure.
gHashTag
added a commit
that referenced
this pull request
Jul 17, 2026
…sistency W1: Unified FineWeb benchmark (closes critique #2) - ALL formats on SAME model (4L d=128), SAME data (500K FineWeb tokens), SAME scale - FP/BF16/GF14/GF16: Δ < 0.0001 BPB (negligible) - SQ-INT6: Δ = +0.0078 BPB (57% less error than plain INT6) - INT6: Δ = +0.0180 BPB - Saved: research/unified_fineweb_benchmark.json W2: Paper consistency pass - INT7 'etalon' paragraph rewritten: 'preliminary, scale-dependent' - Removed 'accuracy-optimal' and '17× cheaper' claims - Added unified benchmark table (Table tab:unified) - Note: matched-budget advantage may be toy-data artefact W3: Frame format bug fix (34 conformance scripts) - FRAME was [0xAA, 0x55] (2 bytes) → missing fmt byte - Fixed to [0xAA, 0x55, 0x00] (3 bytes) matching corona FSM - This bug prevented ALL hardware conformance tests from completing C1: CI yosys-synth job (closes critique #3) - .github/workflows/lut-report.yml - Synthesizes 10 GF formats (ADD+MUL), uploads utilization report - Runs on push to gf_adder_param.v/gf_mul_param.v
gHashTag
added a commit
that referenced
this pull request
Sep 5, 2026
… them Continuity protocol clean: one cron (no duplicate), binary rebuilt from source, tripwire RUNNING before the backlog was touched. Disk recovered on its own, 2.62 -> 5.42 GiB, cause outside this loop's surface. Took B26's four wrong-value defects. Two live in files that compile standalone under local Zig 0.16 (loop_decide.zig, dev_scan.zig both `zig test` clean); the other two do not -- perf_benchmark.zig and pathology.zig die on 0.15-era std.fs APIs -- so (c) and (d) stay unverifiable here and were left alone. Checking the two that could be checked corrected both. (a) is not a mistyped key. It is a guard with no producer anywhere. The code scans .trinity/loop_state.json for a key that file lacks; the SPEC names .trinity/consecutive_failures, which does not exist on disk at all; and the key does exist in .trinity/auto_action_failures.json, which nothing reads, nothing writes, and whose last_success reads 2026-03-17. Priority guard #2 of the decision engine's 9 has never fired and could not have. The obvious rename is also wrong: 'fail' is computed per-invocation at heartbeat.zig:167-169, so it is neither consecutive nor cumulative. Filed as OD11. (b) the proposed fix would have INVERTED the bug. scanPipeline never fires because it searches for the literal "failed" in a file containing only "fail". The audit noted the spec points at pipeline_state.json, "which does contain the literal" -- it does, as a KEY: {"groups":[{"failed":0,...}]}. Redirecting the substring search there would make the counter fire on every scan of a perfectly healthy pipeline. The fix is to parse and test failed > 0. Nothing shipped. (a) needs a decision; (b) changes behaviour of a scanner feeding `tri dev pick` with no binary available to verify against. Three iterations running now, every recommendation this loop received -- last one from a 25-agent sweep carrying execution evidence -- has been wrong in some material detail only running it exposed. Twice the proposed fix was worse than the bug.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Bumps @docusaurus/module-type-aliases from 3.9.2 to 3.10.0.
Release notes
Sourced from
@docusaurus/module-type-aliases's releases.... (truncated)
Changelog
Sourced from
@docusaurus/module-type-aliases's changelog.... (truncated)
Commits
0d98888v3.10.01451780chore(ci): fixes for the npm trusted publishing workflow (#11823)5dff744chore(ci): add Trusted Publishing release workflow through dispatch action (#...bca9ce7chore: release v3.9.2 (#11491)Maintainer changes
This version was pushed to npm by [GitHub Actions](https://www.npmjs.com/~GitHub Actions), a new releaser for
@docusaurus/module-type-aliasessince your current version.Dependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting
@dependabot rebase.Dependabot commands and options
You can trigger Dependabot actions by commenting on this PR:
@dependabot rebasewill rebase this PR@dependabot recreatewill recreate this PR, overwriting any edits that have been made to it@dependabot show <dependency name> ignore conditionswill show all of the ignore conditions of the specified dependency@dependabot ignore this major versionwill close this PR and stop Dependabot creating any more for this major version (unless you reopen the PR or upgrade to it yourself)@dependabot ignore this minor versionwill close this PR and stop Dependabot creating any more for this minor version (unless you reopen the PR or upgrade to it yourself)@dependabot ignore this dependencywill close this PR and stop Dependabot creating any more for this dependency (unless you reopen the PR or upgrade to it yourself)