Comma v0.1-1T is a 7 billion parameter language model trained on 1 trillion tokens from
the Comma v0.1 dataset, comprising of openly licensed text from
the Common Pile.
Comma v0.1-1T is a "base model" that can be used a the starting point for finetuning and post-training.
It performs comparably to budget-matched models (7 billion parameters, 1 trillion tokens) trained on unlicensed data.
This repository is a quantization of the aforementioned model using BitsAndBytes 8 Bit.
This requires transformers and bitsandbytes. From our test, the Python process use 15800MB of VRAM.
1from transformers import AutoModelForCausalLM, AutoTokenizer
2
3model_name = "whsinth/comma-v0.1-1t-bnb-8b"
4
5tokenizer = AutoTokenizer.from_pretrained(model_name)
6model = AutoModelForCausalLM.from_pretrained(model_name, device_map="auto")
7inputs = tokenizer("This is how to write hello world in C:\n```\n""", return_tensors="pt", return_token_type_ids=False).to("cuda:0")
8outputs = model.generate(**inputs, max_new_tokens=100)
9print(tokenizer.decode(outputs[0], skip_special_tokens=True))
