Downloads · 30 days
0
siddarthg44/anlp-a1-2023102040-C1
anlp-a1-2023102040-C1 is a machine learning model from siddarthg44. Use it for the machine learning task on the model card, and read the license before you ship it in a product. The card lists the license as mit.
Encoder-decoder transformer built from scratch, trained to decrypt binary cipher sequences into English plaintext. One of a five-way controlled ablation; this repo holds C1.
Downloads · 30 days
0
Access
Public
Updated Sep 2, 2026
Repo size
269 MB
Likes
0
Public
Click a slice to open those files.
.pt51.6 MB · 99%
From the Hugging Face model README
Encoder-decoder transformer built from scratch, trained to decrypt binary cipher sequences into English plaintext. One of a five-way controlled ablation; this repo holds C1.
| Component | Setting |
|---|---|
| Changed from base | None (base) |
| Positional encoding | sinusoidal |
| Attention | mha |
| Normalization | layernorm |
| Tokenization | bpe |
| d_model / heads / layers | 256 / 8 / 4+4 |
| Parameters | 12.89M |
| Best validation loss | 0.1010 (epoch 39) |
import torch
from huggingface_hub import hf_hub_download
ckpt = torch.load(hf_hub_download("siddarthg44/anlp-a1-2023102040-C1", "best.pt"), map_location="cpu")
# rebuild via train.build_model(ModelConfig(**ckpt["model_config"]), ckpt["meta"])
model.load_state_dict(ckpt["state_dict"])
Training and evaluation code is in the submission 2023102040_assignment1.