dev's Classic Expository Style pass over the websites is the base;
re-applied from this branch only where a fact was missing or wrong: real
volume titles and status, no stale counts or timing claims, the four-volume
teaching guide and slide pages. Lecture lists regenerated by
tools/lectures.py; the PPTX note appears on all four volume pages.
Plain words and formal posture across the landing site, labs, slides, TinyTorch
guide, MLS-im docs and Hardware Kits: removed hype and intensifiers, contractions
and em-dashes in expository prose, announcers and wordy verbs. Wording only; no
code, links, numbers or claims changed.
Landing, navigation, and front pages across the main site, books,
instructors, labs, kits, TinyTorch guide and book, MLSysIM docs, and
MLPerf EDU, plus the shared navbar and ecosystem cards. Volumes use
their book titles and statuses; hard-coded counts (contributors,
universities, decks, slides, domains) are removed; em dashes, emoji,
slogans, and ad copy are rewritten. Lab ranges and kit chapter pairings
are checked against the repo. The You-are-here script recognizes the
Volume III and IV pages.
Release-hardening and transformer-era release of TinyTorch, audited as one
tree. This commit brings tinytorch/ and the TinyTorch CI workflows to the
state of the audit/tinytorch-release branch, where the individual changes
are recorded commit by commit.
- Milestones: transformer-era parts (TinyGPT, sequence routing, TinyCopilot,
conversational Q&A), Milestone 06 "MLPerf to Generative Serving (2018)",
Milestone 07 kernels; per-part recording; NumPy-referenced losses and
causality, attention-content and position probes; try-it prompts after a
pass.
- Framework: dropout eval mode, target gradients, float32 stays float32,
layer validation, LoRA adapters.
- tito: loopback login callback with a state nonce, consented progress sync,
TLS-verified update check, Ctrl-C after a pass keeps the pass.
- Grading: graded tests compare values against NumPy references; tests that
could not fail now check something.
- CI: the publish preflight runs its tests and ships the commit it tested;
workflow inputs pass through env; third-party actions pinned.
- Guide and book: read line by line against the code; every demo is a real
recording.
See tinytorch/CHANGELOG.md, section 0.2.0.
On Milestone 01's separable points a random line usually lands near 0% or
100%. A 0% run was labeled 'Random Guessing!' and the header promised
~50%; 0% now reads 'Boundary points the wrong way!' with a diagnosis that
explains the swapped sides, and the header expectations say 0% to 100%.
The Quick Start capture is re-taken from a real run with the new label.
The dataset prompt treated a missing terminal as 'cancel' with no hint;
it now says why nothing downloaded, and TINYTORCH_AUTO_DOWNLOAD=1 answers
yes for scripted runs.
- Code blocks go back to wrapping: with scroll, trailing `# comments`
were cut off at the right edge with no cue, worst on phones. The
module command blocks already put comments on their own lines.
- Home page on phones: comparison cards stack instead of widening the
page to 531px, and the timeline is one left-aligned column.
- Dark mode: the home lead paragraph was hard-coded dark gray; the
shared navbar stays light, so its title is kept dark; the sidebar logo
gets a light plate so "TORCH" stays visible.
- Tables use smaller type on phones; definition-list entries
(Troubleshooting) get space between them.
- Credits printed a Markdown link as raw text inside an HTML block;
"1998." and "2018." paragraphs on Milestones 04 and 06 rendered as
ordered lists; "** Foundation Tier**" showed literal asterisks; the
Optimization page's nested bullets were flattened and it named its
cycle three ways; Datasets field lines ran together.
- Names: Quick Start is titled Quick Start; "TinyTorch Olympics" is
"Torch Olympics"; Community's educators section points to For
Instructors; milestone pages carry their title in front matter so
their contents list is not nested under it; the Glossary's A-Z
headings appear in its contents list.
- tools/diagrams/STYLE.md, an internal style guide, is no longer
published into the site and its search index.
- Module 01 links Module 02 from What's next.
- Glossary: opens with a "Tools and course terms" section (terminal,
virtual environment, activate, tito, Jupyter Lab, notebook, cell, the
two meanings of kernel, module, export, unit and integration tests,
milestone, Binder), adds weight, bias, and parameter, gives
broadcasting a two-line example, drops the Lab Guide wording, moves
into Getting Started in the sidebar, and is linked from Quick Start.
The newcomer reviewer found none of the tool words it stumbled on.
- Module 01's lead ("saturate the ALUs or stall on cache misses") told
the first page a beginner reads that they were underqualified; it now
says what a tensor is and what they will write.
- Figures on 02, 05, 06, 07, 08 carried a one-line caption inside the
image above the real caption (two of them overclaiming: "correctly
accumulating", "guarantees mathematical equivalence"); removed, canvas
trimmed, renders checked. 05's "drop_last drops remainder" labeled an
option TinyTorch's DataLoader does not have.
Three reviewers playing an undergraduate, a self-taught engineer, and
a newcomer from electrical engineering read the site as students. The
fixes their reports agreed on:
- Quick Start: "Your First Module (15 Minutes)" becomes about 4 to 6
hours; the separate `jupyter lab` step is gone (`tito module start`
opens it); it says what a virtual environment is, that each new
terminal must activate it, what `complete` checks, and that you write
three functions; milestones are listed by the module that unlocks
them; a weeks-at-3/5/10-hours table; the instructor section moves to
For Instructors.
- Lists glued to the preceding line rendered as run-on text on Quick
Start, the three tier pages, datasets, community, and testing.
- Tier pages quoted one "per module" range that the module pages
contradict; they now give the real spread.
- Claims the site's own pages contradict: the perceptron as "the first
trainable neural network" (the milestone is forward-only), "a
complete GPT model generating text" by Module 13, a "competition-
ready" capstone, the unmeasured 65-75% CIFAR-10 figure (four pages),
"RNNs choke" on six-letter reversal, the duplicated "irreversible
understanding" slogan, and the "your first TinyOS" aside.
The Milestone 01 figure did not match what `tito milestone run
perceptron` prints (a six-point "Synthetic Data & Decision Boundary"
step that no longer exists), and nothing in the repo could regenerate
it. guide/tools/capture-terminal.py runs a command with color forced
on and renders its exact output with Rich's SVG exporter, headed by the
command; --lines keeps unedited ranges and marks what was left out.
The new figure is an excerpt of a real run in a scratch checkout where
Modules 01-03 were completed with `tito module complete`. The run is
unseeded; the caption says so and that a run occasionally guesses right
by luck. The PNG is replaced by the SVG.
Phase 1 of the site restructure: known bugs that needed no decision.
- tiers/foundation.qmd "Next Steps" still taught a workflow that no
longer exists: "Read the ABOUT guide", "Implement in *_dev.py",
"Export to *_sol.py". None of those files exist. It is the first page a
student sees after the Quick Start. Replaced with the real two
commands and a link to the Quick Start.
- glossary.qmd was reachable from nowhere. It was a chapter of the Lab
Guide PDF, and nothing on the website linked to it once that was
retired. Added to the Reference section of the sidebar.
- milestones/index.qmd showed `cd milestones/... && python script.py`
beside `tito milestone run 01`. A direct run skips the prerequisite
check and records nothing, so a student who did that would finish the
milestone and still see it as not done.
- getting-started: "benchmark against MLPerf, the industry standard" ->
"using MLPerf's measurement discipline". Milestone 06's README calls
these classroom experiments, not MLPerf submissions.
- Tier pages used `tito module start 09_convolutions` while every other
page uses `09`. Both work; standardised on the number.
Left alone on purpose: getting-started's `tito milestone run perceptron`.
It sits beside a real terminal capture of that exact command, so
changing the text to `run 01` would make the prose and the screenshot
disagree. Both are valid aliases.
Three of the 109 current-generation diagrams carry a caption baked into
the image; fourteen of the twenty-five older ones do. Quarto already
prints a caption underneath, so the reader got the same voice twice in
two typefaces, and the canvas carried dead space to hold it.
Eight of those footers were safe to identify: a single #D1D5DB rule in
the trailing stretch of the file, with only text after it. Their text is
appended to the page's fig-cap before removal, so nothing is lost --
several carried qualifications the caption did not have, such as the
stride/padding/bias of the convolution example and that the
cross-entropy decimals are rounded.
The other six are not touched. In those, #D1D5DB is an internal grid
line rather than a footer rule: in 16_neuron-removal it appears on line
5 of 102. Treating it as a separator would have deleted the diagram's
own labels. They are also book-only files with no guide page to hold a
caption.
Canvases are re-sized to the content. The first attempt measured rect y
without adding rect height and clipped the bottom row of
01_tensor-diag-1; caught by rendering, which is why every one of the
eight was re-rendered and inspected after the fix.
Milestone 06 declares no default_part, so `tito milestone run mlperf`
runs both of its scripts and the prerequisite check uses the union at
milestone.py:143, [1-8, 11, 12, 13, 14-19]. Part 2, "Generation
Speedup", needs the GPT modules: milestone.py:140 lists 11, 12, 13.
Three pages published the requirement as 01-08 + 14-19. A student who
followed them would complete the optimization tier and then be refused
by the prerequisite gate with no idea which modules were missing.
Also aligns two milestone display names with the CLI's own strings
("Transformer Era", "MLPerf Benchmarks").
Three separate overclaims, each checked against the code that would have
to produce the number.
1. --tier challenge emits the full reference solution.
solutions.py::solution_role_action strips a region under the challenge
tier only when it carries role="challenge". src/ contains zero such
annotations (236 regions: 189 scaffold, 6 core, 41 unannotated), so the
tier is a no-op. apply_release_tier output for challenge is byte-identical
to instructor, verified directly.
The code matches its design in NBGRADER_RELEASE_TIERS.md and the
assertions in tests/cli/test_release_tiers.py, so this is unfinished
annotation work, not a code defect, and the tier semantics are left
alone. But an instructor running the command we document gets a fully
solved "challenge" assignment, so `tito nbgrader generate --tier
challenge` now refuses when no region in the requested modules is marked
role="challenge", and says what to annotate.
Note for the author: INSTRUCTOR.md described challenge as "All 236 Blocks
(full implementation from scratch)", which is a different design from
NBGRADER_RELEASE_TIERS.md's "keep for student baseline, strip for
challenge". Those two need reconciling before the tier is annotated.
2. The student tier strips 47 regions, not 41.
Unannotated regions default to core (solutions.py::solution_role), so the
student tier strips 41 unannotated plus 6 explicit role="core" = 47, and
leaves the 189 role="scaffold" regions solved. The documented 41/195 split
is 236 minus 41, which silently reclassifies those 6 core regions as
scaffold. Corrected in getting-started.qmd, tito/overview.qmd, and
INSTRUCTOR.md.
3. The home page advertised optimization gains the milestone disclaims.
index.qmd promised "8-16x smaller, 12-40x faster". milestones/06_2018_mlperf
says compression and speedup "are outcomes to measure, not benefits the
script assumes", that it "does not certify a deployment target or
guarantee a fixed compression/speed" ratio, and that a small workload "may
become slower". The repo's own quantization ceiling is 4x. Reworded, along
with the same claims in ml-timeline.js, whose module range 14-18 was also
wrong (14-19).
Also: the CIFAR-10 milestone target was published as "75%+" in seven
places. milestones/04_1998_cnn states 65-75%, its README expects 70%+, and
02_lecun_cifar10.py passes at >= 65. Corrected to the stated range.
And `tito system jupyter` is documented as starting JupyterLab; it starts
the classic notebook unless given --lab.
Vijay Janapa Reddi