Hyperparameters

• 3 epoch
• 1e-4 -> 1e-5 with cosine lr decay
• batch size 128
• max sequence length 2048
• AdamW(weigth decay=0.01, b1=0.9, b2=0.99, grad_clip=1.0)
• no warmup
• BF16

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("heegyu/WizardVicuna-pythia-1.4b-deduped")
model = AutoModelForCausalLM.from_pretrained("heegyu/WizardVicuna-pythia-1.4b-deduped")

inputs = tokenizer(["Human: Hi\n\nAssistant: "], return_tensors="pt")
outputs = model.generate(**inputs, max_new_tokens=16)
print(tokenizer.batch_decode(outputs, skip_special_tokens=False))

output: ['Human: Hi\n\nAssistant: Hello! How can I assist you today?<|endoftext|>']