Description

RecAlpaca is a model which instruction tuned Alpaca-LoRA model with recommendation task dataset MovieLens 100k.
Usage

from transformers import GenerationConfig, LlamaForCausalLM, LlamaTokenizer
from peft import PeftModel

device = "cuda"
base_model_name = "decapoda-research/llama-7b-hf"
lora_weights = "AGI-Edgerunners/RecAlpaca-lora-7b-v1"

tokenizer = LlamaTokenizer.from_pretrained(base_model_name)
model = PeftModel.from_pretrained(LlamaForCausalLM.from_pretrained(base_model_name), lora_weights).to(device)

generation_config = GenerationConfig(
    temperature=0.1,
    top_p=0.73,
    top_k=40,
    num_beams=4
)
max_new_tokens = 128

instruction = "Based on the movies that I've watched before, could you suggest some similar movies for me to watch next? Please use the MovieLens 100K dataset to recommend movies that you think would appeal to my tastes."
inputs = "The Long Kiss Goodnight, French Kiss, The Maltese Falcon, Dazed and Confused, and Strange Days"

prompt = f"""Below is an instruction that describes a task, paired with an input that provides further context. Write a response that appropriately completes the request.  # noqa: E501

### Instruction:
{instruction}

### Input:
{inputs}

### Response:
"""

inputs = tokenizer(prompt, return_tensors="pt")
input_ids = inputs["input_ids"].to(device)
generation_output = model.generate(input_ids=input_ids,
                                   generation_config=generation_config,
                                   return_dict_in_generate=True,
                                   output_scores=True,
                                   max_new_tokens=max_new_tokens,
                                   )
print(tokenizer.decode(generation_output.sequences[0]))
More Detail

see our github repository: RecAlpaca