git clone https://github.com/X-PLUG/mPLUG-Owl.git

cd mPLUG-Owl/mPLUG-Owl3pip install -r requirements.txt

pip install modelscope

#模型下载
from modelscope import snapshot_downloadmodel_dir = snapshot_download('iic/mplug-owl3-7b-240728', cache_dir='/root/autodl-tmp', revision='master')

import torchmodel_path = '/root/autodl-tmp/mPLUG/mPLUG-Owl3-7B-240728'config = mPLUGOwl3Config.from_pretrained(model_path)print(config)# model = mPLUGOwl3Model(config).cuda().half()model = mPLUGOwl3Model.from_pretrained(model_path, attn_implementation='sdpa', torch_dtype=torch.half)model.eval().cuda()

from PIL import Imagefrom transformers import AutoTokenizer, AutoProcessorfrom decord import VideoReader, cpu # pip install decordmodel_path = '/root/autodl-tmp/mPLUG/mPLUG-Owl3-7B-240728'tokenizer = AutoTokenizer.from_pretrained(model_path)processor = model.init_processor(tokenizer)image = Image.new('RGB', (500, 500), color='red')messages = [  {"role": "user", "content": """<|image|>Describe this image."""},{"role": "assistant", "content": ""}]inputs = processor(messages, images=image, videos=None)inputs.to('cuda')inputs.update({'tokenizer': tokenizer,'max_new_tokens':100,'decode_text':True,})g = model.generate(**inputs)print(g)

from PIL import Imagefrom transformers import AutoTokenizer, AutoProcessorfrom decord import VideoReader, cpu # pip install decord
model_path = '/root/autodl-tmp/mPLUG/mPLUG-Owl3-7B-240728'tokenizer = AutoTokenizer.from_pretrained(model_path)processor = model.init_processor(tokenizer)messages = [  {"role": "user", "content": """<|video|>Describe this video."""},{"role": "assistant", "content": ""}]
videos = ['/nas-mmu-data/examples/car_room.mp4']MAX_NUM_FRAMES=16
def encode_video(video_path):def uniform_sample(l, n):    gap = len(l) / nidxs = [int(i * gap + gap / 2) for i in range(n)]return [l[i] for i in idxs]  vr = VideoReader(video_path, ctx=cpu(0))sample_fps = round(vr.get_avg_fps() / 1)# FPSframe_idx = [i for i in range(0, len(vr), sample_fps)]if len(frame_idx) > MAX_NUM_FRAMES:frame_idx = uniform_sample(frame_idx, MAX_NUM_FRAMES)    frames = vr.get_batch(frame_idx).asnumpy()frames = [Image.fromarray(v.astype('uint8')) for v in frames]print('num frames:', len(frames))return framesvideo_frames = [encode_video(_) for _ in videos]inputs = processor(messages, images=None, videos=video_frames)
inputs.to('cuda')inputs.update({'tokenizer': tokenizer,'max_new_tokens':100,'decode_text':True,})g = model.generate(**inputs)print(g)