Skip to content

PyTorch port: ELF model, multi-backend, Orbax bridge, Muon, benchmarks#5

Open
tzhazuma wants to merge 11 commits into
lillian039:mainfrom
tzhazuma:pytorch-port
Open

PyTorch port: ELF model, multi-backend, Orbax bridge, Muon, benchmarks#5
tzhazuma wants to merge 11 commits into
lillian039:mainfrom
tzhazuma:pytorch-port

Conversation

@tzhazuma
Copy link
Copy Markdown

Complete PyTorch port of ELF (arXiv:2605.10938). Includes full ELF-B/M/L models, 5 converted pretrained checkpoints, Muon optimizer, PPL eval, multi-backend device detection, Orbax checkpoint bridge, LaTeX report.

tzhazuma and others added 11 commits May 15, 2026 04:20
Ultraworked with [Sisyphus](https://github.com/code-yeongyu/oh-my-openagent)

Co-authored-by: Sisyphus <clio-agent@sisyphuslabs.ai>
Ultraworked with [Sisyphus](https://github.com/code-yeongyu/oh-my-openagent)

Co-authored-by: Sisyphus <clio-agent@sisyphuslabs.ai>
Ultraworked with [Sisyphus](https://github.com/code-yeongyu/oh-my-openagent)

Co-authored-by: Sisyphus <clio-agent@sisyphuslabs.ai>
Ultraworked with [Sisyphus](https://github.com/code-yeongyu/oh-my-openagent)

Co-authored-by: Sisyphus <clio-agent@sisyphuslabs.ai>
Ultraworked with [Sisyphus](https://github.com/code-yeongyu/oh-my-openagent)

Co-authored-by: Sisyphus <clio-agent@sisyphuslabs.ai>
Ultraworked with [Sisyphus](https://github.com/code-yeongyu/oh-my-openagent)

Co-authored-by: Sisyphus <clio-agent@sisyphuslabs.ai>
Ultraworked with [Sisyphus](https://github.com/code-yeongyu/oh-my-openagent)

Co-authored-by: Sisyphus <clio-agent@sisyphuslabs.ai>
Python 3.14 compat and transformers attention mask fix

Co-authored-by: Sisyphus <clio-agent@sisyphuslabs.ai>
Muon optimizer with Newton-Schulz orthogonalization

Co-authored-by: Sisyphus <clio-agent@sisyphuslabs.ai>
LaTeX PyTorch port report PDF

Co-authored-by: Sisyphus <clio-agent@sisyphuslabs.ai>
PPL eval script and updated report with benchmark results, ELF-M/L conversion
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant