8xa100m40-gradient-clipping

Model Card for gpjt/8xa100m40-gradient-clipping

This model is gpjt/8xa100m40-gradient-clipping, 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 32b -- Interventions: gradient clipping

How to Get Started with the Model

You can download and run the model for inference directly:

from transformers import pipeline
pipe = pipeline("text-generation", model="gpjt/8xa100m40-gradient-clipping", 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:

>>> from transformers import AutoTokenizer, AutoModel, AutoModelForCausalLM
>>> tokenizer = AutoTokenizer.from_pretrained("gpjt/8xa100m40-gradient-clipping")
>>> model = AutoModel.from_pretrained("gpjt/8xa100m40-gradient-clipping", trust_remote_code=True)
>>> llm_model = AutoModelForCausalLM.from_pretrained("gpjt/8xa100m40-gradient-clipping", 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: 8a A100 with 40GiB per GPU
• Tokens: 3,260,190,720 (Chinchilla-optimal of 20x parameters) rounded up to the nearest batch.
• Dataset: gpjt/fineweb-gpt2-tokens
• Micro-batch size: 12
• Global batch size: 96
• Dropout: 0.1
• Gradient clipping: 3.5
• Learning rate: 0.0004
• Schedule learning rate: False
• Weight decay: 0.1