1from optimum.neuron import NeuronModelForCausalLM
2from transformers import AutoTokenizer
3tokenizer = AutoTokenizer.from_pretrained("mistralai/Mistral-7B-Instruct-v0.3")
4tokenizer.pad_token_id = tokenizer.eos_token_id if tokenizer.pad_token_id is None else tokenizer.pad_token_id
5model = NeuronModelForCausalLM.from_pretrained("yahavb/Mistral-7B-Instruct-v0-3-neuron", use_cache=True)
6inputs = tokenizer("how is the weather in paris", return_tensors="pt")
7outputs = model.generate(**inputs, max_new_tokens=50, do_sample=True, use_cache=True, temperature=0.7, top_k=50, top_p=0.9)
8outputs = outputs[0, inputs.input_ids.size(-1):]
9response = tokenizer.decode(outputs, skip_special_tokens=True)
10print(response)