Download Models

Mistral Checkpoints

The Mistral team has trained 5 GPT-2 Medium models and 5 GPT-2 Small models on the OpenWebText corpus and is making them available to the public.

Each model is available on the Hugging Face Hub and can be accessed via Git LFS.

Checkpoints are branches of each repo for each model. For instance, here is how to get the 300k step checkpoint for battlestar:

# Make sure you have git-lfs installed
# (https://git-lfs.github.com)
git lfs install

# get checkpoint 300000 for battlestar
git clone https://huggingface.co/stanford-crfm/battlestar-gpt2-small-x49 --branch checkpoint-300000 --single-branch
cd battlestar-gpt2-small-x49
git lfs pull

Links to the checkpoints are in the table below.

GPT-2 Medium Models

Run

Type

Size

Link

Arwen

GPT-2 Medium

4.9G

download

Beren

GPT-2 Medium

4.9G

download

Celebrimbor

GPT-2 Medium

4.9G

download

Durin

GPT-2 Medium

4.9G

download

Eowyn

GPT-2 Medium

4.9G

download

GPT-2 Small Models

Run

Type

Size

Link

Alias

GPT-2 Small

1.8G

download

Battlestar

GPT-2 Small

1.8G

download

Caprica

GPT-2 Small

1.8G

download

Darkmatter

GPT-2 Small

1.8G

download

Expanse

GPT-2 Small

1.8G

download