ANLP Assignment 2 (Monsoon 2026): all checkpoints and raw generations
Every model trained for the assignment lives in this one repository, one folder per run. All training logs are on
W&B: https://wandb.ai/tchaitanya0109-iiit-hyderabad/anlp-a2 (public). Code, README and report: the course GitLab
repository advanced-nlp/m26-assignment2-chaitanya.thogata on code.iiit.ac.in.
| folder | contents |
|---|---|
part1/<variant>/ |
Part 1, seed 42: decoder-only translation models (vi/ja → en) trained from scratch for 725M tokens. Variants: dense, moe-e4k1 (4 experts, top-1), moe-e4k2 (top-2), moe-s1e3k1 (1 shared + 3 routed, top-1), moe-e4k2-active (top-2, active params = dense) |
part1/seed43-<variant>/ |
the same five configurations re-trained with seed 43 (seed-variance replicates) |
part2/<optimizer>/ |
Part 2, seed 42: the Part 1 dense configuration pre-trained from scratch for 1× browndw/human-ai-parallel-corpus (40.3M tokens) with adamw, mars, lion, muon, sophia (bonus); final/ plus intermediate checkpoints tok-0.1x/ … tok-1.0x/ every 0.1× of the dataset |
part2/seed43-<optimizer>/ |
the same five Part 2 runs with seed 43 |
part3/generations/ |
Part 3: raw generations of EleutherAI/pythia-160m on 1,000 ROCStories test stories for greedy, beam (W = 1, 2, 4), top-k and top-p decoding (sampling with 3 seeds), plus per-method metrics |
assets/ |
full-resolution figures and result tables of the report (Part 1 routing heatmaps, load balance, validation curves, benchmarks, significance tests, seed comparison; Part 2 loss / BLEU / chrF curves, LR sweeps, benchmark, significance; Part 3 quality-diversity and beam timing) |
Each checkpoint folder holds model.safetensors, config.json (model hyper-parameters) and tokenizer.json
(byte-level BPE). Every run folder has a model card (README.md); Part 1 folders also hold metrics.json (test
perplexity, BLEU, chrF) and Part 2 final/summary.json holds validation loss / test BLEU / chrF at every 0.1×. The
models are custom PyTorch modules: load them with the code in the GitLab repository (src/part1/model.py,
src.utils.load_model_safetensors).