E1-沐曦部署、量化、微调 InternVL3
输入“/”快速插入内容
🍠E1-沐曦部署、量化、微调 InternVL3
E1-沐曦部署、量化、微调 InternVL3
飞书用户1672
飞书用户3425飞书用户7972飞书用户4653飞书用户21925月15日修改
文档尚处于草稿阶段，完善中
MXC500是沐曦显卡，可以比较丝滑的用于大模型训练、量化、推理等场景。
浦语书生实战营的五一沐曦专场里已经有了InternLM3-8B-Instruct模型的量化、部署、微调的实测指南：沐曦玩转 LMDeploy、XTuner 和 InternLM3，现在打算在MXC500显卡上对InternVL3-8B模型进行部署、量化和微调。
对于算力平台的操作，在沐曦玩转 LMDeploy、XTuner 和 InternLM3文章里已经介绍。本文直接进入InternVL3-8B模型的部署、量化和微调。
这次的部署、量化和微调全都选择Dlinfer下属的Xtuner镜像：
如果使用Dlinfer下属的Dlinfer镜像，会因为triton版本太老，而出现<triton> check failed报错。
1.
MXC500上使用transformers部署InternVL3-8B模型
这里部署就不使用lmdeploy了，因为会遇到硬件限制Hardware limit。直接参考InternVL3-8B指南里给出的transformers部署的指南(https://huggingface.co/OpenGVLab/InternVL3-8B#inference-with-transformers)，用transformers部署InternVL3-8B模型。
1.1
cli-demo命令行部署
大致修改一下模型路径path = '/root/public-model/models/OpenGVLab/InternVL3-8B'和图片路径pixel_values = load_image('/root/cli-demo/tiger.jpeg', max_num=12).to(torch.bfloat16).cuda()
指南页面的demo代码里第8行修改为加入from transformers import AutoModel, AutoTokenizer, AutoConfig，
第119行修改为use_flash_attn=False,
第87行修改为config = AutoConfig.from_pretrained('/root/public-model/models/OpenGVLab/InternVL3-8B', trust_remote_code=True)
即为：
代码块
import math
import numpy as np
import torch
import torchvision.transforms as T
from decord import VideoReader, cpu
from PIL import Image
from torchvision.transforms.functional import InterpolationMode
from transformers import AutoModel, AutoTokenizer, AutoConfig
IMAGENET_MEAN = (0.485, 0.456, 0.406)
IMAGENET_STD = (0.229, 0.224, 0.225)
def build_transform(input_size):
MEAN, STD = IMAGENET_MEAN, IMAGENET_STD
transform = T.Compose([
T.Lambda(lambda img: img.convert('RGB') if img.mode != 'RGB' else img),
T.Resize((input_size, input_size), interpolation=InterpolationMode.BICUBIC),
T.ToTensor(),
T.Normalize(mean=MEAN, std=STD)
])
return transform
def find_closest_aspect_ratio(aspect_ratio, target_ratios, width, height, image_size):
best_ratio_diff = float('inf')
best_ratio = (1, 1)
area = width * height
for ratio in target_ratios:
target_aspect_ratio = ratio[0] / ratio[1]
ratio_diff = abs(aspect_ratio - target_aspect_ratio)
if ratio_diff < best_ratio_diff:
best_ratio_diff = ratio_diff
best_ratio = ratio
elif ratio_diff == best_ratio_diff:
if area > 0.5 * image_size * image_size * ratio[0] * ratio[1]:
best_ratio = ratio
return best_ratio
def dynamic_preprocess(image, min_num=1, max_num=12, image_size=448, use_thumbnail=False):
orig_width, orig_height = image.size
aspect_ratio = orig_width / orig_height
# calculate the existing image aspect ratio
target_ratios = set(
(i, j) for n in range(min_num, max_num + 1) for i in range(1, n + 1) for j in range(1, n + 1) if
i * j <= max_num and i * j >= min_num)
target_ratios = sorted(target_ratios, key=lambda x: x[0] * x[1])
# find the closest aspect ratio to the target
target_aspect_ratio = find_closest_aspect_ratio(
aspect_ratio, target_ratios, orig_width, orig_height, image_size)
# calculate the target width and height
target_width = image_size * target_aspect_ratio[0]
target_height = image_size * target_aspect_ratio[1]
blocks = target_aspect_ratio[0] * target_aspect_ratio[1]
# resize the image
resized_img = image.resize((target_width, target_height))
processed_images = []
for i in range(blocks):
box = (
(i % (target_width // image_size)) * image_size,
(i // (target_width // image_size)) * image_size,
((i % (target_width // image_size)) + 1) * image_size,
((i // (target_width // image_size)) + 1) * image_size
)
# split the image
split_img = resized_img.crop(box)
processed_images.append(split_img)
assert len(processed_images) == blocks
if use_thumbnail and len(processed_images) != 1:
thumbnail_img = image.resize((image_size, image_size))
processed_images.append(thumbnail_img)
return processed_images
def load_image(image_file, input_size=448, max_num=12):
image = Image.open(image_file).convert('RGB')
transform = build_transform(input_size=input_size)
images = dynamic_preprocess(image, image_size=input_size, use_thumbnail=True, max_num=max_num)
pixel_values = [transform(image) for image in images]
pixel_values = torch.stack(pixel_values)
return pixel_values
def split_model(model_name):
device_map = {}
world_size = torch.cuda.device_count()
config = AutoConfig.from_pretrained('/root/public-model/models/OpenGVLab/InternVL3-8B', trust_remote_code=True)
num_layers = config.llm_config.num_hidden_layers
# Since the first GPU will be used for ViT, treat it as half a GPU.
num_layers_per_gpu = math.ceil(num_layers / (world_size - 0.5))
num_layers_per_gpu = [num_layers_per_gpu] * world_size
num_layers_per_gpu[0] = math.ceil(num_layers_per_gpu[0] * 0.5)
layer_cnt = 0
for i, num_layer in enumerate(num_layers_per_gpu):
for j in range(num_layer):
device_map[f'language_model.model.layers.{layer_cnt}'] = i
layer_cnt += 1
device_map['vision_model'] = 0
device_map['mlp1'] = 0
device_map['language_model.model.tok_embeddings'] = 0
device_map['language_model.model.embed_tokens'] = 0
device_map['language_model.output'] = 0
device_map['language_model.model.norm'] = 0
device_map['language_model.model.rotary_emb'] = 0
device_map['language_model.lm_head'] = 0
device_map[f'language_model.model.layers.{num_layers - 1}'] = 0
return device_map
# If you set `load_in_8bit=True`, you will need two 80GB GPUs.
# If you set `load_in_8bit=False`, you will need at least three 80GB GPUs.
path = '/root/public-model/models/OpenGVLab/InternVL3-8B'
device_map = split_model('InternVL3-8B')
model = AutoModel.from_pretrained(
path,
torch_dtype=torch.bfloat16,
load_in_8bit=False,
low_cpu_mem_usage=True,
use_flash_attn=False,
trust_remote_code=True,
device_map=device_map).eval()
tokenizer = AutoTokenizer.from_pretrained(path, trust_remote_code=True, use_fast=False)
# set the max number of tiles in `max_num`
pixel_values = load_image('/root/demo/tiger.jpeg', max_num=12).to(torch.bfloat16).cuda()
generation_config = dict(max_new_tokens=1024, do_sample=True)
# single-image single-round conversation (单图单轮对话)
question = '<image>\n简述图片内容'
response = model.chat(tokenizer, pixel_values, question, generation_config)
print(f'User: {question}\nAssistant: {response}')
# single-image multi-round conversation (单图多轮对话)
question = '<image>\n详细叙述图片内容'
response, history = model.chat(tokenizer, pixel_values, question, generation_config, history=None, return_history=True)
print(f'User: {question}\nAssistant: {response}')
question = '请根据图片内容做一手古诗'
response, history = model.chat(tokenizer, pixel_values, question, generation_config, history=history, return_history=True)
print(f'User: {question}\nAssistant: {response}')