matthewjohns/clip-multitask-v3

Multitask (CLIP)

Overview

This repository is a compact, custom PyTorch implementation of CLIP for Multitask. The giant configuration is intended for code review, smoke tests, and small controlled experiments rather than as a production-ready pretrained release.

The whole setup fits in a single module on purpose.

---

Repository status

One Python module contains the model and a runnable example. Generated configuration lives in config.json; the recipe in training_args.json. The checkpoint is an initialization artifact and should not be read as a trained model. Scores are deliberately absent.

---

Default experiment recipe

Defaults: adam optimizer, onecycle schedule. Treat these as starting points; a real run should sweep learning rate and batch size before drawing conclusions.

---

Evaluation guidance

A useful first evaluation would use a task-specific held-out set, report the task metric across at least three seeds, and include a matched-capacity baseline. Keep training logs and environment versions with any published result.

---

Citation

If this repository is useful for your work, a link back is appreciated. No formal citation is required yet.

---

Changelog

1. 2026-10-17 — initial upload
2. 2026-10-21 — config and docs sync

---

Limitations

1. weights are random-init, so outputs carry no signal
2. robustness, fairness and domain transfer are untested
3. the API may change without notice while the design settles

---

Architecture

Item	Value
Architecture	CLIP
Scale	giant
Attention	linear
Fusion	concat mlp
Activation	mish
Normalization	instancenorm

---

Files

File	Purpose
finetune.py	primary artifact
README.md	this documentation
config.json	architecture configuration
training_args.json	default experiment settings
model.safetensors	initialization checkpoint
NOTES.md	informal to-do notes
examples/sample_usage.py	usage sketch

---

License

Released under unknown. Review the source-data terms separately when this repository is used with external datasets.