26385 Commits
Author SHA1 Message Date
Vijay Janapa ReddiandGitHub 5b6b3c8d82 Merge pull request #2197 from Mjrovai/kits-raspi-slm-ollama-update
Kits / Raspberry Pi: update SLM lab (Gemma 3, MoonDream, native tool calling)
2026-11-14 18:13:39 +01:00
Vijay Janapa ReddiandGitHub 5006a5b55d Merge pull request #2196 from Mjrovai/kits-raspi-object-detection-yolo26
Kits / Raspberry Pi: update Object Detection lab (LiteRT, YOLO11/YOLO26)
2026-11-14 18:13:29 +01:00
github-actions[bot] 26b1034cfd docs: add @Mjrovai as contributor for code, doc, test (kits) 2026-11-14 17:10:52 +00:00
Vijay Janapa ReddiandGitHub db9e6a2fdd Merge pull request #2195 from Mjrovai/kits-raspi-image-classification-litert
Kits / Raspberry Pi: move Image Classification lab to LiteRT
2026-11-14 18:07:02 +01:00
Vijay Janapa ReddiandGitHub f2e8b264cf Merge pull request #2194 from Mjrovai/kits-raspi-setup-trixie
Kits / Raspberry Pi: update Setup lab for Raspberry Pi OS Trixie
2026-11-14 18:06:58 +01:00
Vijay Janapa Reddi 4c6daca773 Kits / Raspberry Pi: drop the raw alt text that rendered as a caption under the Ollama logo 2026-11-14 17:58:27 +01:00
Vijay Janapa Reddi b76a6131b3 Kits / Raspberry Pi: match Object Detection setup checklist to the renamed Image Classification steps
Name the setup steps as the updated Image Classification lab does, call
the environment LiteRT, and drop a leftover image filename that rendered
as a caption under the training output.
2026-11-14 17:58:17 +01:00
Andrea 2f56a3a648 Add labs skeleton and first lab 2026-11-13 19:30:20 +01:00
Marcelo RovaiandClaude Opus 5.5 6cf7d6d84c Kits / Raspberry Pi: mark RAG working-directory commands as bash
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-12 10:23:49 -03:00
Marcelo RovaiandClaude Opus 5.5 bd7c34f2d8 Kits / Raspberry Pi: use a venv instead of removing EXTERNALLY-MANAGED
Removing /usr/lib/python3.11/EXTERNALLY-MANAGED does not apply on
Trixie (Python 3.13) and contradicts the venv rule stated right after
it. Also rename the sampling variable in avg_temp_power.sh from
SECONDS, which bash reserves and increments automatically, to SAMPLES.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-12 10:18:27 -03:00
Marcelo RovaiandClaude Opus 5.5 a55c805015 Kits / Raspberry Pi: allow 'boxs' in codespell
Ultralytics prints the plural of the custom class 'box' as 'boxs'
('3 boxs, 4 wheels'); the lab quotes that output verbatim.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-12 10:18:27 -03:00
Marcelo RovaiandClaude Opus 5.5 d24278190a Kits / Raspberry Pi: allow 'nothink' in codespell
'/set nothink' is an Ollama session command that turns off the
thinking mode; it is not a misspelling of 'nothing'.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-12 10:18:27 -03:00
Marcelo RovaiandClaude Opus 5.5 2d3872284d Kits / Raspberry Pi: move Image Classification lab to LiteRT
Sync with the author's EdgeML Made Easy book: ai-edge-litert replaces
tflite_runtime, venv with system site packages, Picamera2 test, and
recaptured Edge Impulse figures.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-12 10:11:47 -03:00
Marcelo RovaiandClaude Opus 5.5 051e7dea9f Kits / Raspberry Pi: update Setup lab for Raspberry Pi OS Trixie
Sync with the author's EdgeML Made Easy book: Trixie and Imager 2.0,
rpi-swap and zram (with Zero 2 W notes), rpicam-hello, and a new
temperature and power measurement section.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-12 10:11:47 -03:00
Marcelo RovaiandClaude Opus 5.5 b82cc73ce8 Kits / Raspberry Pi: update Object Detection lab (LiteRT, YOLO11/YOLO26)
Sync with the author's EdgeML Made Easy book: LiteRT for SSD and the
live-stream app, expanded IoU/mAP, and a rewritten YOLO section with
YOLO11 and YOLO26, Trixie installation, NCNN/ONNX export measurements,
and the imgsz rule for exported models.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-12 10:11:47 -03:00
Marcelo RovaiandClaude Opus 5.5 73397fbb3b Kits / Raspberry Pi: update SLM lab (Gemma 3, MoonDream, native tools)
Sync with the author's EdgeML Made Easy book: Gemma 3/3n/4 replace
Gemma 2, MoonDream, Ollama commands and model inspection, an expanded
Python library section, direct REST API, and function calling with
Ollama's native tools instead of instructor/OpenAI.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-12 10:11:47 -03:00
github-actions[bot] 1db21e632b docs: add @Shakhtar-Sankur as contributor for bug, code, test, doc (labs, mlsysim) 2026-11-12 12:58:36 +00:00
github-actions[bot] 3c159b42e4 docs: add @prawalpokharel as contributor for code, test, doc (labs, mlsysim) 2026-11-12 12:57:54 +00:00
github-actions[bot] 531f2423c4 docs: add @Nishitha-Rajesh as contributor for code, doc, tool (mlsysim) 2026-11-12 12:55:34 +00:00
github-actions[bot] e706a1681c docs: add @SatnamCodes as contributor for code, doc, test, tool (tinytorch) 2026-11-12 12:53:35 +00:00
Vijay Janapa Reddi 2c8f1297d6 chore(mlsysim): refresh paper values, diagrams, and timing receipt after #2192, #2193, and #2186 2026-11-12 07:35:23 -05:00
Vijay Janapa Reddi 3b5997d98c Merge pull request #2173 from SatnamCodes/feat/tinytorch-ext8-cuda-gemm 2026-11-12 07:28:57 -05:00
Vijay Janapa Reddi d713722591 Merge pull request #2186 from Nishitha-Rajesh/mlsysim-benchmark2-decode-itl 2026-11-12 07:28:41 -05:00
Vijay Janapa Reddi e578d76871 Merge pull request #2193 from prawalpokharel/feat/mlsysim-temporal-shift 2026-11-12 07:25:44 -05:00
Vijay Janapa Reddi 927971ef4f Merge pull request #2192 from Shakhtar-Sankur/mlsysim-t4-fp32 2026-11-12 07:25:20 -05:00
Vijay Janapa Reddi 36087583ea Merge pull request #2189 from Mjrovai/kits-xiaoml-sensecraft-audio-course-links 2026-11-12 07:25:20 -05:00
Vijay Janapa Reddi b5516e8d1f Merge pull request #2191 from Nishitha-Rajesh/staffml-exercises-prototype 2026-11-12 07:25:20 -05:00
Vijay Janapa Reddi 959e4d9262 Merge pull request #2190 from Nishitha-Rajesh/staffml-moe-mobile 2026-11-12 07:25:19 -05:00
Marcelo RovaiandGitHub be0e8920af Merge branch 'dev' into kits-xiaoml-sensecraft-audio-course-links 2026-11-12 09:22:30 -03:00
github-actions[bot] 0ec5033289 Update contributors list [skip ci] 2026-11-12 11:28:41 +00:00
Vijay Janapa Reddi 7d8bf15995 fix(ci): refresh Vol I source_sha1 bindings and define accent-dark fallback 2026-11-12 05:55:24 -05:00
github-actions[bot] d8e00c9dc7 Update contributors list [skip ci] 2026-11-12 10:44:55 +00:00
Prawal Pokharel 94d14b584e mlsysim: add TemporalShiftModel for deadline-constrained carbon-aware time shifting 2026-11-11 23:28:33 -06:00
Vijay Janapa Reddi be9ce34557 Merge branch 'fix/socratiq-real-screenshots' into dev 2026-11-11 10:30:58 -05:00
Vijay Janapa Reddi b78fb8c0ff Merge dev into fix/socratiq-real-screenshots for retire 2026-11-11 10:29:58 -05:00
Vijay Janapa Reddi b694465f2c Merge branch 'fix/vol1-stack-alt-text' into dev 2026-11-11 10:29:50 -05:00
Vijay Janapa Reddi 91e33f42e7 feat(socratiq): harden BYOK key validation, context carryover, and guide screenshots 2026-11-11 10:28:12 -05:00
Vijay Janapa Reddi f3e7965178 Merge dev into fix/vol1-stack-alt-text for retire 2026-11-11 10:26:41 -05:00
Vijay Janapa Reddi 3226110cf8 Merge branch 'fix/visual-step-polish' into dev 2026-11-11 09:36:52 -05:00
Vijay Janapa Reddi 809a7445a1 fix(explore): polish step-by-step visual state, typography, and layout across all 111 figures 2026-11-11 09:36:20 -05:00
Sankur Kundu de3aed4866 labs: carry the T4 FP32 and INT4 rates in the browser wheel
labs/tools/check_browser_wheels.py requires the committed mlsysim wheel to match the runtime source. Replace mlsysim/hardware/data/cloud/T4.yaml inside mlsysim-0.3.1-py3-none-any.whl with the updated file and its RECORD entry; every other entry is unchanged, and all 508 runtime files now match the source.
2026-11-11 04:31:10 +00:00
Sankur Kundu 588f613e45 fix(mlsysim): give the T4 its FP32 and INT4 rates
The T4 entry stored only an int8 rate, so Engine.solve with precision='fp32' fell through to peak_flops, the 65 TFLOP/s FP16 tensor-core rate, and warned "Results may be inaccurate". FP32 work on the T4 was modelled about 8x too fast, and INT4 4x too slow.

Add the datasheet rates: FP32 8.1 TFLOP/s on the CUDA cores and INT4 260 TOPS. peak_flops (FP16 tensor, 65) and int8 (130) are unchanged, so FP16 and INT8 results, the appendix table and the hardware chapter's ridge point stay the same.

test_precision_keys_use_canonical_vocabulary catches an aliased key such as 'fp32_cuda', but not a chip with no fp32 key at all, which is how the T4 slipped through. Add a T4 regression test and an Engine check that a compute-bound FP32 solve is limited by the FP32 rate without the fallback warning.
2026-11-11 03:41:28 +00:00
Claude 6b965264c7 StaffML: prototype KV-cache decode and per-channel INT8 exercises (#2155)
Option B from #2155: the vault schema is unchanged, and each exercise lives
under staffml/exercises/<exercise>/ with README.md (task and napkin math),
starter.py, solution.py and a test file. The tests grade the reference
solution by default, or the starter with STAFFML_IMPL=starter.

- kv-cache-decode: incremental single-head attention with a pre-allocated
  KV cache. Tests check the outputs against full recompute to 1e-5, that
  each token is projected once (3T rows, not 3T(T+1)/2), that the cache is
  written in place, and the KV-cache size of Llama-2-7B.
- per-channel-int8: per-channel weight scales, a per-tensor activation
  scale, and an INT8 x INT8 -> INT32 matmul with one rescale per channel.
  Tests check dtypes and scale shapes, the half-step round-trip error, the
  analytic matmul error bound, per-channel vs per-tensor accuracy, and
  INT32 headroom.

NumPy and pytest only, CPU only. 13 tests pass on the solutions.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01SxFzWDUc3UeWQ1QBGjw6UM
2026-11-11 01:22:27 +00:00
Claude cfdcc9cfd0 StaffML: three mixture-of-experts questions on the mobile track (#2156)
The three drafts reviewed in #2156, with topic: mixture-of-experts on the
mobile track (status: draft, provenance: llm-draft):

- mobile-202611-77dc (L4, memory): sizing a phone for Qwen3-30B-A3B by
  active vs total parameters.
- mobile-202611-1e6f (L5, memory): expert offloading to UFS 4.0 for
  Mixtral-8x7B at 0% and 90% expert-cache hit rates.
- mobile-202611-2872 (L5, latency): why prefill touches every expert in an
  offloaded MoE, and the time-to-first-token that follows.

IDs allocated with vault new; vault check --strict passes on the full
corpus.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01SxFzWDUc3UeWQ1QBGjw6UM
2026-11-11 01:17:56 +00:00
Vijay Janapa Reddi 405a1074a3 Merge branch 'fix/socratiq-real-screenshots' into dev 2026-11-10 18:06:14 -05:00
Vijay Janapa Reddi f2f6e34214 feat(socratiq): add configurable selection clip, BYOK/OpenRouter PKCE SSO, and figure context
- Add configurable PopClip-style selection bar (Explain this, Ask, Check me, Taxonomy, Simplify, Example, Study tools) with gear menu and Learning preferences sync
- Add security-hardened BYOK and 1-Click OpenRouter OAuth PKCE SSO module (src_shared/byok-provider.js) with strict HTTPS host allowlist, secret redaction, session-only default storage, and shared .github.io origin protection
- Add Free shared model status bar and Connect your own Gemini / Claude controls in both Focused Reader and Study Tools
- Extract interactive figure explorer ([data-mlsys-explore]) control parameters in reader and study context extractors
- Update SocratiQ user guide prose and 2x HiDPI screenshots and add unit and Playwright regression coverage
2026-11-10 18:04:19 -05:00
Vijay Janapa Reddi d5e2ae21d1 Merge branch 'feat/explore-clean-4part-player' into dev 2026-11-10 17:38:41 -05:00
Vijay Janapa Reddi 12edb776c7 feat(explore): adopt unanimous student-voted Clean 4-Part Player layout across all 111 figures
- Wrap secondary live meters, calculation tables, and step-math workspaces inside the collapsible 'Inspect the calculation' drawer by default across original-player.js and convolution.js so every figure presents a uniform 4-part layout:
  1. Guiding Question + 1-sentence Intro + compact [Assumptions] button
  2. Connected 3-Tier Player Frame (Canvas -> Tinted Controls Bar -> Flush Blue Step Narration)
  3. Compact [Inspect the calculation] dropdown-button box at the bottom
- Refresh runtime_sha1 in figure-registry.json for all 111 figures.
2026-11-10 17:38:31 -05:00
Vijay Janapa Reddi c8eb854df6 Merge branch 'feat/explore-student-voted-layout' into dev 2026-11-10 17:18:20 -05:00
Vijay Janapa Reddi dd31b60f55 style(explore): adopt student-panel winning Option B+ box layout across all 111 figures
- Style closed Assumptions and Inspect-the-calculation disclosures as compact fit-content dropdown buttons that expand into full-width bordered cards when opened.
- Unify the Stage Canvas, tinted Transport Controls Dock, and flush Step Narration into a single connected 3-tier player frame.
- Replace the outer 4-sided box on .mlsys-original-details with a top hairline divider to eliminate nested box-in-box clutter.
2026-11-10 17:18:11 -05:00