You can deploy this model and start earning money today!

Teen-Different/
adaptive-gpts

Name: adaptive-gpts
Availability: InStock

Views

No views yet

Adaptive GPTs

Five small GPT variants trained from scratch for 12k steps on a ~300M token mixed corpus (TinyStories, Cosmopedia WikiHow, WikiText-2, Python code). The core experiment: can a hypernetwork that reads the residual stream generate dynamic LoRA-style updates to the value heads at inference time?

Blog post: Adaptive Attention at Inference Time: Does It Actually Work?

Code: REDDITARUN/a-gpt

Checkpoints

File	Model	Params	Description
`base_best.pt`	Base GPT	28.9M	Vanilla causal GPT. 4 layers, 4 heads, 256 dim.
`matched_best.pt`	Matched GPT	30.5M	6 layers, 4 heads, 256 dim. Parameter-matched to adaptive.
`adaptive_best.pt`	Adaptive GPT	30.5M	Base + per-head TinyHeadTransformer hypernetwork on V.
`diffusion_best.pt`	Diffusion GPT	28.9M	Bidirectional denoising (discrete diffusion).
`adaptive_diffusion_best.pt`	Adaptive Diffusion GPT	30.5M	Diffusion + per-head hypernetwork on V.

Training

All models share: AdamW, lr 3e-4 with cosine decay, batch size 8, 4 gradient accumulation steps, 256 context window, GPT-2 tokenizer, RoPE.

Data mixture: Cosmopedia WikiHow (40%), Python code (30%), WikiText-2 (20%), TinyStories (10%).

Loading

python

import torch
from models.base_gpt import GPT_Base, GPT_Base_Config
from models.a_gpt import GPT_Custom, GPT_Custom_Config, HyperConfig

# Example: load adaptive checkpoint
config = GPT_Custom_Config()
hyper_config = HyperConfig()
model = GPT_Custom(config, hyper_config=hyper_config)

ckpt = torch.load("adaptive_best.pt", map_location="cpu")
model.load_state_dict(ckpt["model"])

Teen-Different/adaptive-gpts

Adaptive GPTs

Checkpoints

Training

Loading

Teen-Different/
adaptive-gpts