license: mit
LoRA weights only and trained for research - nothing from the foundation model. Trained using Open-Assistant's dataset. Shout-out to Open-Assistant and LAION for giving us early research access to the dataset.
Sample usage
import torch
import os
import transformers
from peft import PeftModel
from transformers import LlamaTokenizer, LlamaForCausalLM
model_path = "decapoda-research/llama-7b-hf"
peft_path = 'serpdotai/llama-oasst-lora-7B'
tokenizer_path = 'decapoda-research/llama-7b-hf'
model = LlamaForCausalLM.from_pretrained(model_path, load_in_8bit=True, device_map="auto") # or something like {"": 0}
model = PeftModel.from_pretrained(model, peft_path, torch_dtype=torch.float16, device_map="auto") # or something like {"": 0}
tokenizer = LlamaTokenizer.from_pretrained(tokenizer_path)
batch = tokenizer("\n\nUser: Are you sentient?\n\nAssistant:", return_tensors="pt")
with torch.no_grad():
out = model.generate(
input_ids=batch["input_ids"].cuda(),
attention_mask=batch["attention_mask"].cuda(),
max_length=100,
do_sample=True,
top_k=50,
top_p=1.0,
temperature=1.0
)
print(tokenizer.decode(out[0]))
The model will continue the conversation between the user and itself. If you want to use as a chatbot you can alter the generate method to include stop sequences for 'User:' and 'Assistant:' or strip off anything past the assistant's original response before returning.
Trained for 4 epochs with a sequence length of 2048 on 8 A6000s with an effective batch size of 120.
Training settings:
lr: 2.0e-04 lr_scheduler_type: linear warmup_ratio: 0.06 weight_decay: 0.1 optimizer: adamw_torch LoRA config:
target_modules: ['q_proj', 'k_proj', 'v_proj', 'o_proj'] r: 64 lora_alpha: 32 lora_dropout: 0.05 bias: "none" task_type: "CAUSAL_LM"