| import torch |
| from transformers import AutoTokenizer, AutoModelForCausalLM |
|
|
| from model.modelclass import Model |
| class Kangaroo(Model): |
| def __init__(self): |
| Kangaroo_Init() |
|
|
| def Run(self, file, inp): |
| return Kangaroo_Run(file, inp) |
| |
| def name(self): |
| return "Kangaroo" |
| |
| tokenizer, model, terminators = None, None, None |
|
|
| def Kangaroo_Init(): |
| global tokenizer, model, terminators |
| tokenizer = AutoTokenizer.from_pretrained("KangarooGroup/kangaroo") |
| model = AutoModelForCausalLM.from_pretrained( |
| "KangarooGroup/kangaroo", |
| torch_dtype=torch.bfloat16, |
| trust_remote_code=True, |
| ) |
| model = model.to("cuda") |
| terminators = [tokenizer.eos_token_id, tokenizer.convert_tokens_to_ids("<|eot_id|>")] |
|
|
| def Kangaroo_Run(file, inp): |
| out, history = model.chat(video_path=file, |
| query=inp, |
| tokenizer=tokenizer, |
| max_new_tokens=512, |
| eos_token_id=terminators, |
| do_sample=True, |
| temperature=0.6, |
| top_p=0.9,) |
| print('Assitant: \n', out) |
| return out |
