GPT2-Linear-XL

A conversion of gpt2-xl that uses linear layers instead of convolutional layers. This is not an official OpenAI project.

Pretrained model on English language using a causal language modeling (CLM) objective. It was introduced in this paper and first released at this page.

GPT-2 is a transformers model pretrained on a very large corpus of English data in a self-supervised fashion. This means it was pretrained on the raw texts only, with no humans labelling them in any way (which is why it can use lots of publicly available data) with an automatic process to generate inputs and labels from those texts. More precisely, it was trained to guess the next word in sentences. More precisely, inputs are sequences of continuous text of a certain length and the targets are the same sequence, shifted one token (word or piece of word) to the right. The model uses internally a mask-mechanism to make sure the predictions for the token i only uses the inputs from 1 to i but not the future tokens. This way, the model learns an inner representation of the English language that can then be used to extract features useful for downstream tasks. The model is best at what it was pretrained for however, which is generating texts from a prompt.

• Main model: crumbly/gpt2-linear-xl
• Sharded model: crumbly/gpt2-linear-xl-sharded
• Sharded + Brain-float 16bit model: crumbly/gpt2-linear-xl-sharded-bf16

Config:

{
  "n_embd": 1600,
  "n_head": 25,
  "n_layer": 48,
  "n_positions": 1024,
}

Usage

Inference on GPU with 4-bit quantization:

%pip install -qq transformers accelerate bitsandbytes

from transformers import AutoModelForCausalLM, AutoTokenizer
from transformers import BitsAndBytesConfig
import torch

model_id = "crumbly/gpt2-linear-xl-sharded-bf16"
bnb_config = BitsAndBytesConfig(
    load_in_4bit=True,
    bnb_4bit_use_double_quant=True,
    bnb_4bit_quant_type="nf4",
    bnb_4bit_compute_dtype=torch.bfloat16
)

tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(
    model_id,
    trust_remote_code=True,
    device_map={"":0},
    quantization_config=bnb_config
)

inputs = tokenizer("Once upon a time,", return_tensors='pt')
inputs = {
    k:v.cuda() for k,v in inputs.items()
}
outputs = model.generate(
    **inputs,
    max_new_tokens=32,
    temperature=0.7,
    do_sample=True
)
tokenizer.decode(outputs[0])

TODO

• test to see if model works with .from_pretrained
  
  • test fp32, fp16, 8 and 4 bit
• shard model to max 1gb for use in even lower vram settings
  
• safetensors
  
• upload bf16 version of model
  
• upload 8bit model and 4bit model
  
• convert other base gpt2 models
• open orca QLoRA on XL
• ReLoRA continued pretraining on RefinedWeb or RedPajama to reach 1T tokens