--- license: cc-by-nc-4.0 language: - en pipeline_tag: text-generation tags: - nvidia - NVLM - pytorch - multimodal - conversational --- ## Model Details Today (September 17th, 2024), we introduce [NVLM 1.0](https://arxiv.org/abs/2409.11402), a family of frontier-class multimodal large language models (LLMs) that achieve state-of-the-art results on vision-language tasks, rivaling the leading proprietary models (e.g., GPT-4o) and open-access models (e.g., Llama 3-V 405B and InternVL 2). Remarkably, NVLM 1.0 shows improved text-only performance over its LLM backbone after multimodal training. We are open-sourcing the model weights and code for the community. ## Other Resources [Inference Code (HF)](https://huggingface.co/nvidia/NVLM-1.0-D-72B/tree/main)   [Training Code (Coming soon)]()   [Website](https://nvlm-project.github.io/)   [Paper](https://arxiv.org/abs/2409.11402) ## Benchmark Results We train our model with legacy [Megatron-LM](https://github.com/NVIDIA/Megatron-LM/tree/main/megatron/legacy) and adapt the codebase to Huggingface for model hosting, reproducibility, and inference. We observe numerical differences between the Megatron and Huggingface codebases, which are within the expected range of variation. We provide the results from both the Huggingface codebase and the Megatron codebase for reproducibility and comparison with other models. Results (as of September 17th, 2024) in the multimodal benchmarks are as follows: | Benchmark | MMMU (val / test) | MathVista | OCRBench | AI2D | ChartQA | DocVQA | TextVQA | RealWorldQA | VQAv2 | |------------------------------|-------------------|-----------|----------|------|---------|--------|---------|-------------|-------| | NVLM-D 1.0 72B (Huggingface) | 58.7 / 54.9 | 65.2 | 852 | 94.2 | 86.0 | 92.6 | 82.6 | 69.5 | 85.4 | | NVLM-D 1.0 72B (Megatron) | 59.7 / 54.6 | 65.2 | 853 | 94.2 | 86.0 | 92.6 | 82.1 | 69.7 | 85.4 | | Llama 3.2 90B | 60.3 / - | 57.3 | - | 92.3 | 85.5 | 90.1 | - | - | 78.1 | | Llama 3-V 70B | 60.6 / - | - | - | 93.0 | 83.2 | 92.2 | 83.4 | - | 79.1 | | Llama 3-V 405B | 64.5 / - | - | - | 94.1 | 85.8 | 92.6 | 84.8 | - | 80.2 | | InternVL2-Llama3-76B | 55.2 / - | 65.5 | 839 | 94.8 | 88.4 | 94.1 | 84.4 | 72.2 | - | | GPT-4V | 56.8 / 55.7 | 49.9 | 645 | 78.2 | 78.5 | 88.4 | 78.0 | 61.4 | 77.2 | | GPT-4o | 69.1 / - | 63.8 | 736 | 94.2 | 85.7 | 92.8 | - | - | - | | Claude 3.5 Sonnet | 68.3 / - | 67.7 | 788 | 94.7 | 90.8 | 95.2 | - | - | - | | Gemini 1.5 Pro (Aug 2024) | 62.2 / - | 63.9 | 754 | 94.4 | 87.2 | 93.1 | 78.7 | 70.4 | 80.2 | ## How to use When converting Megatron checkpoint to Huggingface, we adapt [InternVL codebase](https://huggingface.co/OpenGVLab/InternVL2-Llama3-76B) to support model loading and multi-GPU inference in HF. For training, please refer to [Megatron-LM (Coming soon)](). ### Prepare the environment We provide a docker build file in the [Dockerfile](Dockerfile) for reproduction. The docker image is based on `nvcr.io/nvidia/pytorch:23.09-py3`. *Note: We observe that different transformer versions / CUDA versions / docker versions can lead to slight benchmark number differences. We recommend using the Dockerfile above for precise reproduction.* ### Model loading ```python import torch from transformers import AutoModel path = "nvidia/NVLM-D-72B" model = AutoModel.from_pretrained( path, torch_dtype=torch.bfloat16, low_cpu_mem_usage=True, use_flash_attn=False, trust_remote_code=True).eval() ``` ### Multiple GPUs The model can be loaded on multiple GPUs as follows: ```python import torch import math from transformers import AutoModel def split_model(): device_map = {} world_size = torch.cuda.device_count() num_layers = 80 # 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.lm_head'] = 0 device_map[f'language_model.model.layers.{num_layers - 1}'] = 0 return device_map path = "nvidia/NVLM-D-72B" device_map = split_model() model = AutoModel.from_pretrained( path, torch_dtype=torch.bfloat16, low_cpu_mem_usage=True, use_flash_attn=False, trust_remote_code=True, device_map=device_map).eval() ``` ### Inference ```python import torch from transformers import AutoTokenizer, AutoModel import math from PIL import Image import torchvision.transforms as T from torchvision.transforms.functional import InterpolationMode def split_model(): device_map = {} world_size = torch.cuda.device_count() num_layers = 80 # 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.lm_head'] = 0 device_map[f'language_model.model.layers.{num_layers - 1}'] = 0 return device_map 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 path = "nvidia/NVLM-D-72B" device_map = split_model() model = AutoModel.from_pretrained( path, torch_dtype=torch.bfloat16, low_cpu_mem_usage=True, use_flash_attn=False, trust_remote_code=True, device_map=device_map).eval() print(model) tokenizer = AutoTokenizer.from_pretrained(path, trust_remote_code=True, use_fast=False) generation_config = dict(max_new_tokens=1024, do_sample=False) # pure-text conversation question = 'Hello, who are you?' response, history = model.chat(tokenizer, None, question, generation_config, history=None, return_history=True) print(f'User: {question}\nAssistant: {response}') # single-image single-round conversation pixel_values = load_image('path/to/your/example/image.jpg', max_num=6).to( torch.bfloat16) question = '\nPlease describe the image shortly.' response = model.chat(tokenizer, pixel_values, question, generation_config) print(f'User: {question}\nAssistant: {response}') ``` ## Correspondence to Wenliang Dai* (wdai@nvidia.com), Nayeon Lee* (nayeonl@nvidia.com), Boxin Wang* (boxinw@nvidia.com), Zhuolin Yang* (zhuoliny@nvidia.com), Wei Ping* (wping@nvidia.com) *Equal contribution ## Citation 
@article{nvlm2024,
  title={NVLM: Open Frontier-Class Multimodal LLMs},
  author={Dai, Wenliang and Lee, Nayeon and Wang, Boxin and Yang, Zhuolin and Liu, Zihan and Barker, Jon and Rintamaki, Tuomas and Shoeybi, Mohammad and Catanzaro, Bryan and Ping, Wei},
  journal={arXiv preprint},
  year={2024}}
## License The use of this model is governed by the [cc-by-nc-4.0](https://spdx.org/licenses/CC-BY-NC-4.0)