unsloth/Qwen2.5-1.5B (1582479 - 2025-04-28 04:13:37 UTC)

arxiv :
- [object Object]
  The herein LoRa adapter can for instance be used as follows :

python

from transformers import AutoModelForCausalLM, AutoTokenizer
from torch import device, cuda

repo_id = "retrain-pipelines/function_caller_lora"
revision = "<model_revision_commit_hash>"
model = AutoModelForCausalLM.from_pretrained(
    repo_id, revision=revision, torch_dtype="auto", device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(
    repo_id, revision=revision, torch_dtype="auto", device_map="auto")

device = device("cuda" if cuda.is_available() else "cpu")
def generate_tool_calls_list(query, max_new_tokens=400) -> str:
    formatted_query = tokenizer.chat_template.format(query, "")
    inputs = tokenizer(formatted_query, return_tensors="pt").input_ids.to(device)
    outputs = model.generate(inputs, max_new_tokens=max_new_tokens, do_sample=False)
    generated_text = tokenizer.batch_decode(outputs, skip_special_tokens=True)[0]
    return generated_text[len(formatted_query):].strip()

generate_tool_calls_list("Is 49 a perfect square ?")