flashstep's picture
Upload README.md with huggingface_hub
fd1a09f verified
|
Raw History Blame Contribute Delete
524 Bytes
# ANLP Assignment 2 — Part 1
This repository contains the trained model checkpoints for Part 1 of ANLP Assignment 2.
## Model Variants
- `model_variant_1.pt` — Dense 2-layer MLP
- `model_variant_2.pt` — MoE, 4 experts, Top-1 routing
- `model_variant_3.pt` — MoE, 4 experts, Top-2 routing
- `model_variant_4.pt` — Shared + Routed MoE
- `model_variant_5.pt` — Matched Active-Parameter MoE
The checkpoints are provided in PyTorch `.pt` format.
See the accompanying report for experimental details and results.