Repository navigation
docs(readme): bring the README back in line with the code - #157
Merged
Merged
Conversation
Fix claims that drifted since the GPU-first rewrite, each checked against the source it describes: - Hardware support: add the native PyTorch `neuron` backend (Trainium2/3, FP8 on Trainium3), label the AOT row, list HunyuanVideo and LTX-2 as experimental on CUDA. - Backend detection order is neuron, trainium, tpu, cuda, rocm, falling back to trainium (difflet/platforms/__init__.py); say so. - Footnotes: GPU notes 1-8 and Trainium notes 7-15 shared superscripts; Trainium notes are now 9-17 so every reference is unique. Rejoin GPU note 8 to its list. - Latest news: ISO dates; "1.0 is released" conflicts with the pre-alpha badge and pyproject's 0.0.0.dev0, so it is now "First public release". - Prerequisites: Trainium3 via the native backend. - CLI flags table: add --backend, --torch-compile, --quant, --dp-degree, --negative-prompt/--true-cfg-scale, --host-vae/--device-vae, which the README already uses. - Observability: describe today's logs; drop the maintainer plan. - Fold the MiniMax-H3 sentence into the tagline; tie the torch pin to requirements-cuda.lock. - pyproject: add the 3.13 classifier (3.11, 3.12 and 3.14 were listed). Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Summary
Docs-only. After the logo PR (#155) I read the whole README against the code; these are the places where it had drifted. Each item names what it was checked against.
neuronbackend (Trainium2/3, FP8 on Trainium3); AOT row labelled; CUDA row lists HunyuanVideo and LTX-2 as experimentaldifflet/platforms/__init__.pyPLATFORM_FACTORIES,docs/verification/neuron-backend-results.md, the NVIDIA feature table, #147neuron→trainium→tpu→cuda→rocm, fallbacktrainiumdifflet/platforms/__init__.py:33anddetect_platform_name⁷/⁸no longer mean two things; GPU note 8 rejoined to its listpyproject.tomlversion = "0.0.0.dev0", the pre-alpha badge--backend neuron;--quant mxfp8Trainium3-onlydifflet/cli/quant.pyhelp text--backend,--torch-compile,--quant,--dp-degree,--negative-prompt/--true-cfg-scale,--host-vae/--device-vaedifflet/cli/main.py_build_parserand the_add_*_flagshelpers (which subcommands get each flag).claude/skills/plan was a maintainer noterequirements-cuda.lockrequirements-cuda.lock:80pyproject.toml: add the3.13classifier (list had 3.11, 3.12, 3.14)Not changed, flagged for a decision:
requires-python = ">=3.11"whileunit-cpu.ymlstill has a py3.10 matrix entry — one of the two should move.Test plan
](#anchor)in README resolves to a heading; every relative link exists1–17⊆ notes1–17, no duplicate note numbers🤖 Generated with Claude Code