docs: bump install requirement to ~> 1.0 (and fix FastKernels moduledoc drift) - #204
Conversation
README and the livebooks pinned `{:emily, "~> 0.7"}`. A `~> 0.7`
requirement won't resolve the upcoming 1.0.0 release, so move the
install snippets to `~> 1.0`. Other deps (Bumblebee/Axon/Nx/…) are
unchanged.
The moduledoc told consumers to add `{:bumblebee, "~> 0.6"}` and
`{:axon, "~> 0.7"}`, but mix.exs has required `~> 0.7` and `~> 0.8`.
Align the example with the actual optional-dep requirements.
|
@codex review |
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 47266a46f6
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
The README quantization bullet and the qwen3_quantized livebook still described the quantized path as running through `dequantize_defn/1` (dequantize-then-dot). Since the fused quantized_matmul work (61b36f4) the native quantized path lowers to the fused `mx::quantized_matmul` kernel, streaming the packed low-bit weights; dequantize_defn is now only the non-MLX fallback. Update the prose to match.
|
@codex review |
|
Codex Review: Didn't find any major issues. Nice work! Reviewed commit: ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
If Codex has suggestions, it will comment; otherwise it will react with 👍. Codex can also answer questions or update the PR. Try commenting "@codex address that feedback". |
1 participant
Prep for the upcoming 1.0.0 release: point the install docs at
~> 1.0, plus a small unrelated doc-drift fix found along the way.Changes
docs: bump install requirement to ~> 1.0— README (deps snippet + "As a hex consumer" step) and all 8 livebooks pinned{:emily, "~> 0.7"}. A~> 0.7requirement won't resolve1.0.0(Hex excludes it), so the install snippets move to~> 1.0.docs: fix stale bumblebee/axon versions in FastKernels moduledoc— the moduledoc example told consumers to add{:bumblebee, "~> 0.6"}/{:axon, "~> 0.7"}, butmix.exshas required~> 0.7/~> 0.8. Aligned the example with the real optional-dep requirements. (Pre-existing drift, not related to the 1.0 bump.)Deliberately not changed
mix.exs@version "0.7.2"— the version source of truth; it's bumped bymix publisho major(→1.0.0) at release time, not by hand.bench/emily_vs_exla_report.md"Emily 0.7.0 local checkout" — historical benchmark provenance; rewriting it would misrepresent which version produced the numbers.{:bumblebee, "~> 0.7"}deps, "Bumblebee 0.7" prose,## 0.7.xCHANGELOG entries — those are Bumblebee's version / release history, not emily's.Testing
mix precommitgreen (737 tests / 40 doctests / 79 properties, 0 failures; credo, docs, format, compile-warnings-as-errors all clean) — the moduledoc edit compiles and renders.