You can deploy this model and start earning money today!

gpjt/
1xrtx3090-stacked-interventions

Name: 1xrtx3090-stacked-interventions
Availability: InStock

gpjt-llm-from-scratch

custom_code

gpjt/fineweb-gpt2-tokens

apache-2.0

Views

No views yet

Model Card for gpjt/1xrtx3090-stacked-interventions

This model is gpjt/1xrtx3090-stacked-interventions, a trained-from-scratch base model using the GPT-2-style architecture from Sebastian Raschka's book "Build a Large Language Model (from Scratch)".

Model Details

Model Description

Developed by: Giles Thomas, based on code by Sebastian Raschka
Model type: GPT-2 style transformers-based causal LLM.
License: Apache 2
Parameters: 163,009,536
Context length: 1,024
Embedding dimensions: 768
MHA heads: 12
Layers: 12
QKV bias: False
Weight tying: False

Don't have high expectations for the model! It has only 163M parameters (the GPT-2 "small" size) and was trained on roughly the Chinchilla-optimal number of tokens (~20x the number of parameters), which means that it doesn't know many facts and is not terribly smart. If you want to do serious work, use a serious model (I like Qwen's). But if you want to build on this and see what you can do with a 2020-vintage LLM, please do feel free to play with it!

Model Sources

Repository: gpjt/ddp-base-model-from-scratch
Blog post: Writing an LLM from scratch, part 32k -- Interventions: training a better model locally with gradient accumulation

How to Get Started with the Model

You can download and run the model for inference directly:

python

from transformers import pipeline
pipe = pipeline("text-generation", model="gpjt/1xrtx3090-stacked-interventions", trust_remote_code=True)
out = pipe(
    "Every effort moves you",
    max_new_tokens=20,
    do_sample=True,
    temperature=1.4,
    top_k=25,
)
print(out[0]["generated_text"])

Note that because it uses custom code, you'll need to set trust_remote_code to True.

It supports AutoTokenizer, AutoModel and AutoModelForCausalLM:

python

>>> from transformers import AutoTokenizer, AutoModel, AutoModelForCausalLM
>>> tokenizer = AutoTokenizer.from_pretrained("gpjt/1xrtx3090-stacked-interventions")
>>> model = AutoModel.from_pretrained("gpjt/1xrtx3090-stacked-interventions", trust_remote_code=True)
>>> llm_model = AutoModelForCausalLM.from_pretrained("gpjt/1xrtx3090-stacked-interventions", trust_remote_code=True)

You can also fine-tune it; this notebook has an example.

Again, don't expect too much from this model! It's a 163M-parameter GPT-2 one, trained on a limited number of tokens. It's both dumb and ignorant ;-)

Training Details

Machine type: Local machine with an RTX 3090
Tokens: 3,260,190,720 (Chinchilla-optimal of 20x parameters) rounded up to the nearest batch.
Dataset: gpjt/fineweb-gpt2-tokens
Micro-batch size: 6
Global batch size: 96 (using 12 gradient accumulation steps)
Dropout: 0.0
Gradient clipping: 3.5
Learning rate: 0.0014
Schedule learning rate: True
Weight decay: 0.01