Multi-Layer
g2p_multilingual_byT5_tiny_16_layers_100g2p_multilingual_byT5_tiny_12_layers_100sparse-autoencoder-clip-b-32-sae-vanilla-x64-layer-11-hook_mlp_out-l1-5e-05sparse-autoencoder-clip-b-32-sae-vanilla-x64-layer-10-hook_mlp_out-l1-1e-05sparse-autoencoder-clip-b-32-sae-vanilla-x64-layer-0-hook_mlp_out-l1-0.0001sparse-autoencoder-clip-b-32-sae-vanilla-x64-layer-2-hook_mlp_out-l1-8e-05sparse-autoencoder-clip-b-32-sae-vanilla-x64-layer-5-hook_mlp_out-l1-0.0001sparse-autoencoder-clip-b-32-sae-vanilla-x64-layer-8-hook_mlp_out-l1-0.0001
multilayer-sae-server-backup
multilayer-sae server backup
Public migration backup for /data/caotue/multilayer-sae plus relation-specific Nullu checkpoints from /data/caotue/nullu.
Credential-bearing files and public base-model/dataset/cache content are intentionally excluded.
easynla-qwen3-8b-multilayer-bankprompt-injection-multilayerMultiLayerThinFilms
OptoLlama Dataset
Details
The original dataset can be found in the OptoGPT publication 📝 and here on HuggingFace.
Key Enhancements
Inclusion of an absorption feature in the model ➕📈
Increased the wave length range to 300-2,000nm 💡
Structure
├── materials/
│ ├── Ag.csv
│ ├── Al.csv
│ ├── ...
│ └── ZnSe.csv
├── train/
│ ├── train-0.safetensors
│ ├── train-1.safetensors
│ ├── ...
│ └── train-9.safetensors
├── test/
│ └──… See the full description on the dataset page: https://huggingface.co/datasets/HZBSolarOptics/MultiLayerThinFilms.crosscoder-multilayer-split-activations
Crosscoder Multilayer Split Activations
Raw split activation artifacts for multilayer SPARC-style crosscoder training.
This dataset stores reusable base-only and aligned-only activation tensors. These
are intended to be assembled into matched activations.pt training artifacts
before crosscoder training.
Versions
v1
Source local run: interp_utils/crosscoder/results-multi-v1
Layout:
v1/
base_activations/
smollm3-union/
llama32-3b-union/… See the full description on the dataset page: https://huggingface.co/datasets/MInAlA/crosscoder-multilayer-split-activations.FuXi_MultiLayer
Coupled Earth-System Forecast Sample and Temporal Downscaling Model
本仓库提供一个全球海陆气冰耦合预报样例，以及配套的时间降尺度模型权重。样例包含
2020-01-04 起报的初始场和 48 个集合成员、40 个预报时次的离线结果。
This repository provides a global atmosphere-land-ocean-sea-ice coupled forecast
sample and a temporal downscaling model checkpoint. The sample is initialized on
2020-01-04 and contains 48 ensemble members with 40 forecast steps.
Repository Structure
.
├── model/
│ └── 8000_G.pth
└── data/
├── inp_data/
│… See the full description on the dataset page: https://huggingface.co/datasets/Mrxin/FuXi_MultiLayer.