crumb / gpt2-regular-large

gpt-regular-test

i was stupid and all the newline tokens are replaced with [/n] so be wary if you're using the demo on this page that that just means new line

from transformers import AutoTokenizer
from transformers import AutoModelForCausalLM
model = AutoModelForCausalLM.from_pretrained("crumb/gpt2-regular-large")
tokenizer = AutoTokenizer.from_pretrained("gpt2-large", use_fast=True)

prompt = """(Episode begins with Mordecai and Rigby watching TV)
Mordecai: Dude, what are you doing? I think I'm gonna lose my mind.
Rigby:"""

prompt=prompt.replace("\n","[/n]")
tokenz = tokenizer(prompt,return_tensors='pt')['input_ids']
output = model.generate(
    tokenz, 
    max_length=length,
    num_return_sequences=1,
    top_p=.92,
    temperature=.65,
    do_sample=True,
    top_k=125,
    early_stopping=True,
    pad_token_id=tokenizer.eos_token_id
)
output = tokenizer.decode(output[0]).replace("[/n]","\n")
print(output)

This model is a fine-tuned version of gpt2-large on the entirety of Regular Show. It achieves the following results on the evaluation set (The Power, Death Punchies, Do Me a Solid):

• Loss: 1.6383

Intended uses & limitations

Same as gpt2-large

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 1e-05
• train_batch_size: 1
• eval_batch_size: 1
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 1

Training results

Framework versions

• Transformers 4.18.0
• Pytorch 1.11.0
• Datasets 2.1.0
• Tokenizers 0.12.1