# Course checkpoint

*Hosted reading edition:* this is a fresh checkpoint template, not shared or
automatically saved learner progress. Download `progress.md` for your notes.

[Journey map](learning-map.html) | [Course reference](study.html) |
[First session](first-session.html).

The course is prepared; no learner session or model training has been completed.
Advance only after the learner can demonstrate the pass gate in `study.md`.
An assistant running code is not evidence that the learner understands it.

## Resume here

**Current stop:** [01 · Trace the whole loop](learning-map.html#session-01).
**Next action:** open the first-session worksheet and begin its vocabulary sort.
**Last demonstrated gate:** none.
**Evidence / notes:** not recorded yet.
**Remaining confusion:** record after the exercise.

Edit this block in `progress.md` in Otty after each visit. The HTML is a static
reading copy, not an automatically saved progress app. The
[map's save instructions](learning-map.html#save-your-place-and-return-without-asking-this-chat)
show how to refresh it.

## Session ledger

| Session | Topic | State |
| --- | --- | --- |
| 01 | [Components, phases and state](learning-map.html#session-01) | Ready: start `first-session.html` |
| 02 | [Tokenizer and chat template](learning-map.html#session-02) | Not started |
| 03 | [Embeddings, shapes and output head](learning-map.html#session-03) | Not started |
| 04 | [Attention, masks, positions and heads](learning-map.html#session-04) | Not started |
| 05 | [Full block; encoder versus decoder](learning-map.html#session-05) | Not started |
| 06 | [Prefill, decode and KV cache](learning-map.html#session-06) | Not started |
| 07 | [Loss, gradients and optimizer update](learning-map.html#session-07) | Not started |
| 08 | [Tiny educational pretraining](learning-map.html#session-08) | Not started |
| 09 | [Base, instruct and post-training stages](learning-map.html#session-09) | Not started |
| 10 | [Labels, splits and baselines](learning-map.html#session-10) | Not started |
| 11 | [LoRA, templates and loss masks](learning-map.html#session-11) | Not started |
| 12 | [Local smoke training and reload](learning-map.html#session-12) | Not started |
| 13 | [Controlled domain adaptation](learning-map.html#session-13) | Not started |
| 14 | [Failure analysis](learning-map.html#session-14) | Not started |
| 15 | [Matched timing and memory comparison](learning-map.html#session-15) | Not started |
| 16 | [Reproduction and teach-back](learning-map.html#session-16) | Not started |

At the end of each session, record: date, material used, prediction before the
experiment, observed result, artifact/configuration, remaining confusion and
the precise next step. Keep machine-generated weights/data/logs under ignored
`runs/` or `models/`, not loose in the repository.