CSCI 680 Assignment 1: T5-small checkpoints

Checkpoints for https://github.com/aura-lab-wm/teaching-genai4se-2026-tsun04. All models are T5-small (60.4M parameters) trained from random weights in fp32 with the tokenizer in tokenizer/.

Folder	Model
`pretrained-300k-per-epoch/`	Pre-trained on the 300k corpus with per-epoch span masks; lowest held-out loss (used by B and C)
`finetuned-A-nopt/`	Condition A: fine-tuned from random weights on both tasks
`finetuned-B-pt/`	Condition B: fine-tuned from the pre-trained checkpoint on both tasks
`finetuned-C-pt-st/`	Condition C (bonus): fine-tuned from the pre-trained checkpoint on code&comment-to-code only
`tokenizer/`	Byte-level BPE tokenizer, 32,000 entries

Load a model with T5ForConditionalGeneration.from_pretrained("csci680-tsun04/Assignment1_checkpoint", subfolder="finetuned-B-pt") and the tokenizer with tokenizers.Tokenizer.from_file on tokenizer/tokenizer.json.

Downloads last month: -; Downloads are not tracked for this model. How to track

Inference Providers NEW

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support