Downloads · 30 days
0
Llama1BxFWx1024x4pct
aflah/Llama1BxFWx1024x4pct
Llama1BxFWx1024x4pct is a machine learning model from aflah. Use it for the machine learning task on the model card, and read the license before you ship it in a product.
This repository contains a raw GPT-NeoX training checkpoint from the Partial RoPE experiments accompanying Fractional Rotation, Full Potential? Investigating Performance and Convergence of Partial RoPE. The paper was…
Access
Public
Updated Aug 26, 2026
Repo size
16.5 GB
Likes
0
Public
Repo makeup
Click a slice to open those files.
.pt16.5 GB · 100%
At a glance
- Access
- Public
- Created
- Aug 26, 2026
- Updated
- Aug 26, 2026
- SHA
- 0f1fb6c8
Model card
From the Hugging Face model README
Llama1BxFWx1024x4pct
This repository contains a raw GPT-NeoX training checkpoint from the Partial RoPE experiments accompanying Fractional Rotation, Full Potential? Investigating Performance and Convergence of Partial RoPE. The paper was accepted to EMNLP 2026.
Checkpoint details
| Field | Value |
|---|---|
| Architecture | Llama 3.2 1B |
| Dataset | FineWeb (FW) |
| Training sequence length | 1,024 tokens |
| Partial RoPE | 4% |
| Checkpoint | Global step 12,000 |
| Format | Raw GPT-NeoX checkpoint (not Transformers format) |
The files are preserved in their original GPT-NeoX checkpoint format and have not been converted to the Hugging Face Transformers format.
Resources
- Created
- Aug 26, 2026
- Updated
- Aug 26, 2026