Fix Multi GPU Issue

.gitattributes

@@ -33,4 +33,3 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
-examples/red-panda.mp4 filter=lfs diff=lfs merge=lfs -text

README.md

@@ -1,179 +1,110 @@
 ---
 license: mit
-pipeline_tag: image-text-to-text
-library_name: transformers
-base_model:
-  - OpenGVLab/InternViT-6B-448px-V1-5
-  - internlm/internlm2-chat-20b
-new_version: OpenGVLab/InternVL2_5-26B
-base_model_relation: merge
-language:
-  - multilingual
-tags:
-  - internvl
-  - custom_code
+datasets:
+- laion/laion2B-en
+- laion/laion-coco
+- laion/laion2B-multi
+- kakaobrain/coyo-700m
+- conceptual_captions
+- wanng/wukong100m
+pipeline_tag: visual-question-answering
 ---
 
-# InternVL-Chat-V1-5
-
-[\[📂 GitHub\]](https://github.com/OpenGVLab/InternVL)  [\[📜 InternVL 1.0\]](https://huggingface.co/papers/2312.14238)  [\[📜 InternVL 1.5\]](https://huggingface.co/papers/2404.16821)  [\[📜 Mini-InternVL\]](https://arxiv.org/abs/2410.16261)  [\[📜 InternVL 2.5\]](https://huggingface.co/papers/2412.05271)
-
-[\[🆕 Blog\]](https://internvl.github.io/blog/)  [\[🗨️ Chat Demo\]](https://internvl.opengvlab.com/)  [\[🤗 HF Demo\]](https://huggingface.co/spaces/OpenGVLab/InternVL)  [\[🚀 Quick Start\]](#quick-start)  [\[📖 Documents\]](https://internvl.readthedocs.io/en/latest/)
-
-## Introduction
-
+# Model Card for InternVL-Chat-V1.5
 <p align="center">
-  <img src="https://cdn-uploads.huggingface.co/production/uploads/64119264f0f81eb569e0d569/D60YzQBIzvoCvLRp2gZ0A.jpeg" alt="Image Description" width="300" height="300">
+  <img src="https://cdn-uploads.huggingface.co/production/uploads/64119264f0f81eb569e0d569/D60YzQBIzvoCvLRp2gZ0A.jpeg" alt="Image Description" width="300" height="300" />
 </p>
 
 > _Two interns holding hands, symbolizing the integration of InternViT and InternLM._
 
-We introduce InternVL 1.5, an open-source multimodal large language model (MLLM) to bridge the capability gap between open-source and proprietary commercial models in multimodal understanding.
+\[[Paper](https://arxiv.org/abs/2312.14238)\]  \[[GitHub](https://github.com/OpenGVLab/InternVL)\] \[[Chat Demo](https://internvl.opengvlab.com/)\] \[[中文解读](https://zhuanlan.zhihu.com/p/675877376)]
 
-We introduce three simple designs:
+We introduce InternVL 1.5, an open-source multimodal large language model (MLLM) to bridge the capability gap between open-source and proprietary commercial models in multimodal understanding.
+We introduce three simple designs: 
+1. Strong Vision Encoder: we explored a continuous learning strategy for the large-scale vision foundation model---InternViT-6B, boosting its visual understanding capabilities, and making it can be transferred and reused in different LLMs. 
+2. Dynamic High-Resolution: we divide images into tiles ranging from 1 to 40 of 448 &times; 448 pixels according to the aspect ratio and resolution of the input images, which supports up to 4K resolution input.
+3. High-Quality Bilingual Dataset: we carefully collected a high-quality bilingual dataset that covers common scenes, document images, and annotated them with English and Chinese question-answer pairs, significantly enhancing performance in OCR- and Chinese-related tasks.
 
-1. **Strong Vision Encoder:** we explored a continuous learning strategy for the large-scale vision foundation model---InternViT-6B, boosting its visual understanding capabilities, and making it can be transferred and reused in different LLMs.
-2. **Dynamic High-Resolution:** we divide images into tiles ranging from 1 to 40 of 448 × 448 pixels according to the aspect ratio and resolution of the input images, which supports up to 4K resolution input during inference.
-3. **High-Quality Bilingual Dataset:** we carefully collected a high-quality bilingual dataset that covers common scenes, document images, and annotated them with English and Chinese question-answer pairs, significantly enhancing performance in OCR- and Chinese-related tasks.
 
 ## Model Details
-
 - **Model Type:** multimodal large language model (MLLM)
-
 - **Model Stats:**
-
   - Architecture: [InternViT-6B-448px-V1-5](https://huggingface.co/OpenGVLab/InternViT-6B-448px-V1-5) + MLP + [InternLM2-Chat-20B](https://huggingface.co/internlm/internlm2-chat-20b)
   - Image size: dynamic resolution, max to 40 tiles of 448 x 448 (4K resolution).
   - Params: 25.5B
 
 - **Training Strategy:**
-
-  - Learnable component in the pre-training stage: ViT + MLP
-  - Learnable component in the fine-tuning stage: ViT + MLP + LLM
-  - For more details on training hyperparameters, please see our [blog](https://internvl.github.io/blog/2024-04-30-InternVL-1.5/).
-
-## Architecture
-
-![image/png](https://cdn-uploads.huggingface.co/production/uploads/64119264f0f81eb569e0d569/YLvX3V-L0kwsyRn3Lhciw.png)
+  - Pretraining Stage
+    - Learnable Component: ViT + MLP
+    - Data: Please see our technical report.
+  - SFT Stage
+    - Learnable Component: ViT + MLP + LLM
+    - Data: Please see our technical report.
+
+## Released Models
+
+| Model                                                      | Vision Foundation Model                                                     | Release Date           |Note                                |
+| :---------------------------------------------------------:|:--------------------------------------------------------------------------: |:----------------------:| :---------------------------------- |
+| InternVL-Chat-V1.5(🤗 [HF link](https://huggingface.co/OpenGVLab/InternVL-Chat-V1-5))      | InternViT-6B-448px-V1-5(🤗 [HF link](https://huggingface.co/OpenGVLab/InternViT-6B-448px-V1-5))    |2024.04.18       |          support 4K image; super strong OCR; Approaching the performance of GPT-4V and Gemini Pro on various benchmarks like MMMU, DocVQA, ChartQA, MathVista, etc. (🔥new)|
+| InternVL-Chat-V1.2-Plus(🤗 [HF link](https://huggingface.co/OpenGVLab/InternVL-Chat-V1-2-Plus) ) |InternViT-6B-448px-V1-2(🤗 [HF link](https://huggingface.co/OpenGVLab/InternViT-6B-448px-V1-2))    |2024.02.21     |        more SFT data and stronger  |
+| InternVL-Chat-V1.2(🤗 [HF link](https://huggingface.co/OpenGVLab/InternVL-Chat-V1-2) )      |InternViT-6B-448px-V1-2(🤗 [HF link](https://huggingface.co/OpenGVLab/InternViT-6B-448px-V1-2))     |2024.02.11       |             scaling up LLM to 34B       |
+| InternVL-Chat-V1.1(🤗 [HF link](https://huggingface.co/OpenGVLab/InternVL-Chat-V1-1))      |InternViT-6B-448px-V1-0(🤗 [HF link](https://huggingface.co/OpenGVLab/InternViT-6B-448px-V1-0))    |2024.01.24         |   support Chinese and stronger OCR   |
 
 ## Performance
 
-![image/png](https://cdn-uploads.huggingface.co/production/uploads/64119264f0f81eb569e0d569/4b85G7txoJ_LpT19SZJ4A.png)
+![image/png](https://cdn-uploads.huggingface.co/production/uploads/64119264f0f81eb569e0d569/ZyQklQ3C7C60I-xOv7X8L.png)
 
-![image/png](https://cdn-uploads.huggingface.co/production/uploads/64119264f0f81eb569e0d569/i2vp6zSHPS3UIr-1Q9cSe.png)
 
-- We simultaneously use [InternVL](https://github.com/OpenGVLab/InternVL) and [VLMEvalKit](https://github.com/open-compass/VLMEvalKit) repositories for model evaluation. Specifically, the results reported for DocVQA, ChartQA, InfoVQA, TextVQA, MME, AI2D, MMBench, CCBench, MMVet, and SEED-Image were tested using the InternVL repository. OCRBench, RealWorldQA, HallBench, and MathVista were evaluated using the VLMEvalKit.
-
-Limitations: Although we have made efforts to ensure the safety of the model during the training process and to encourage the model to generate text that complies with ethical and legal requirements, the model may still produce unexpected outputs due to its size and probabilistic generation paradigm. For example, the generated responses may contain biases, discrimination, or other harmful content. Please do not propagate such content. We are not responsible for any consequences resulting from the dissemination of harmful information.
+![image/png](https://cdn-uploads.huggingface.co/production/uploads/64119264f0f81eb569e0d569/u98oqlnpZtWdq2dnarVlD.png)
 
 ## Examples
 
-![image/png](https://cdn-uploads.huggingface.co/production/uploads/64119264f0f81eb569e0d569/YVr-93mvVMR6UFpGezns7.png)
-![image/png](https://cdn-uploads.huggingface.co/production/uploads/64119264f0f81eb569e0d569/ivhj4QqcO2NHUa28DTDkK.png)
-![image/png](https://cdn-uploads.huggingface.co/production/uploads/64119264f0f81eb569e0d569/18GeOW10QVcSt5g--TgDY.png)
-![image/png](https://cdn-uploads.huggingface.co/production/uploads/64119264f0f81eb569e0d569/tGM_TwdV297H1fCxQ0PZU.png)
-![image/png](https://cdn-uploads.huggingface.co/production/uploads/64119264f0f81eb569e0d569/FwlSRBpKgURAVkXNOLoSp.png)
-![image/png](https://cdn-uploads.huggingface.co/production/uploads/64119264f0f81eb569e0d569/to3nOaAnyv-fGLEoNPLzz.png)
+![image/png](https://cdn-uploads.huggingface.co/production/uploads/64119264f0f81eb569e0d569/R34jISP4K1U17m9yNP38O.png)
 
-## Quick Start
+![image/png](https://cdn-uploads.huggingface.co/production/uploads/64119264f0f81eb569e0d569/ChkU9XtlsjH0l2EqlO_is.png)
 
-We provide an example code to run InternVL-Chat-V1-5 using `transformers`.
+![image/png](https://cdn-uploads.huggingface.co/production/uploads/64119264f0f81eb569e0d569/1TFxIcf96ANRPLoy4-rbh.png)
 
-> Please use transformers>=4.37.2 to ensure the model works normally.
+![image/png](https://cdn-uploads.huggingface.co/production/uploads/64119264f0f81eb569e0d569/Wpjo1Sdwf7XcEDevqwcr-.png)
 
-### Model Loading
+![image/png](https://cdn-uploads.huggingface.co/production/uploads/64119264f0f81eb569e0d569/kO4-J38sN8TFtmQ5mIBMS.png)
 
-#### 16-bit (bf16 / fp16)
+![image/png](https://cdn-uploads.huggingface.co/production/uploads/64119264f0f81eb569e0d569/qPnTe3Q9UBy8wbclOsmWk.png)
 
-```python
-import torch
-from transformers import AutoTokenizer, AutoModel
-path = "OpenGVLab/InternVL-Chat-V1-5"
-model = AutoModel.from_pretrained(
-    path,
-    torch_dtype=torch.bfloat16,
-    low_cpu_mem_usage=True,
-    use_flash_attn=True,
-    trust_remote_code=True).eval().cuda()
-```
+![image/png](https://cdn-uploads.huggingface.co/production/uploads/64119264f0f81eb569e0d569/l_BILRi13CbZNzbZYn6o6.png)
 
-#### BNB 8-bit Quantization
+![image/png](https://cdn-uploads.huggingface.co/production/uploads/64119264f0f81eb569e0d569/2782y7RnvGBogYEIG__7S.png)
 
-```python
-import torch
-from transformers import AutoTokenizer, AutoModel
-path = "OpenGVLab/InternVL-Chat-V1-5"
-model = AutoModel.from_pretrained(
-    path,
-    torch_dtype=torch.bfloat16,
-    load_in_8bit=True,
-    low_cpu_mem_usage=True,
-    use_flash_attn=True,
-    trust_remote_code=True).eval()
-```
+![image/png](https://cdn-uploads.huggingface.co/production/uploads/64119264f0f81eb569e0d569/RyO35PTH14OFiwyxtAZM2.png)
 
-#### BNB 4-bit Quantization
+![image/png](https://cdn-uploads.huggingface.co/production/uploads/64119264f0f81eb569e0d569/xiLZXWL-JiCTVPnV_VxS2.png)
 
-> **⚠️ Warning:** Due to significant quantization errors with BNB 4-bit quantization on InternViT-6B, the model may produce nonsensical outputs and fail to understand images. Therefore, please avoid using BNB 4-bit quantization.
+![image/png](https://cdn-uploads.huggingface.co/production/uploads/64119264f0f81eb569e0d569/gqX46Tt5jvrcVqb0vcf06.png)
 
-#### Multiple GPUs
 
-The reason for writing the code this way is to avoid errors that occur during multi-GPU inference due to tensors not being on the same device. By ensuring that the first and last layers of the large language model (LLM) are on the same device, we prevent such errors.
 
-```python
-import math
-import torch
-from transformers import AutoTokenizer, AutoModel
+## Model Usage
 
-def split_model(model_name):
-    device_map = {}
-    world_size = torch.cuda.device_count()
-    num_layers = {'Mini-InternVL-2B-V1-5': 24, 'Mini-InternVL-4B-V1-5': 32, 'InternVL-Chat-V1-5': 48}[model_name]
-    # 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
+We provide an example code to run InternVL-Chat-V1.5 using `transformers`.
 
-path = "OpenGVLab/InternVL-Chat-V1-5"
-device_map = split_model('InternVL-Chat-V1-5')
-model = AutoModel.from_pretrained(
-    path,
-    torch_dtype=torch.bfloat16,
-    low_cpu_mem_usage=True,
-    use_flash_attn=True,
-    trust_remote_code=True,
-    device_map=device_map).eval()
-```
-
-### Inference with Transformers
+You also can use our [online demo](https://internvl.opengvlab.com/) for a quick experience of this model.
 
 ```python
-import numpy as np
+import json
+import os
+from transformers import AutoTokenizer, AutoModel
+from tqdm import tqdm
 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
+
 
 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([
@@ -184,6 +115,7 @@ def build_transform(input_size):
     ])
     return transform
 
+
 def find_closest_aspect_ratio(aspect_ratio, target_ratios, width, height, image_size):
     best_ratio_diff = float('inf')
     best_ratio = (1, 1)
@@ -199,7 +131,8 @@ def find_closest_aspect_ratio(aspect_ratio, target_ratios, width, height, image_
                 best_ratio = ratio
     return best_ratio
 
-def dynamic_preprocess(image, min_num=1, max_num=12, image_size=448, use_thumbnail=False):
+
+def dynamic_preprocess(image, min_num=1, max_num=6, image_size=448, use_thumbnail=False):
     orig_width, orig_height = image.size
     aspect_ratio = orig_width / orig_height
 
@@ -237,7 +170,8 @@ def dynamic_preprocess(image, min_num=1, max_num=12, image_size=448, use_thumbna
         processed_images.append(thumbnail_img)
     return processed_images
 
-def load_image(image_file, input_size=448, max_num=12):
+
+def load_image(image_file, input_size=448, max_num=6):
     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)
@@ -245,338 +179,92 @@ def load_image(image_file, input_size=448, max_num=12):
     pixel_values = torch.stack(pixel_values)
     return pixel_values
 
+
+path = "OpenGVLab/InternVL-Chat-V1-5"
 # If you have an 80G A100 GPU, you can put the entire model on a single GPU.
-# Otherwise, you need to load a model using multiple GPUs, please refer to the `Multiple GPUs` section.
-path = 'OpenGVLab/InternVL-Chat-V1-5'
 model = AutoModel.from_pretrained(
     path,
     torch_dtype=torch.bfloat16,
     low_cpu_mem_usage=True,
-    use_flash_attn=True,
     trust_remote_code=True).eval().cuda()
-tokenizer = AutoTokenizer.from_pretrained(path, trust_remote_code=True, use_fast=False)
-
+# Otherwise, you need to set device_map='auto' to use multiple GPUs for inference.
+# model = AutoModel.from_pretrained(
+#     path,
+#     torch_dtype=torch.bfloat16,
+#     low_cpu_mem_usage=True,
+#     trust_remote_code=True,
+#     device_map='auto').eval()
+
+tokenizer = AutoTokenizer.from_pretrained(path, trust_remote_code=True)
 # set the max number of tiles in `max_num`
-pixel_values = load_image('./examples/image1.jpg', max_num=12).to(torch.bfloat16).cuda()
-generation_config = dict(max_new_tokens=1024, do_sample=True)
+pixel_values = load_image('./examples/image1.jpg', max_num=6).to(torch.bfloat16).cuda()
 
-# pure-text conversation (纯文本对话)
-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}')
+generation_config = dict(
+    num_beams=1,
+    max_new_tokens=512,
+    do_sample=False,
+)
 
-question = 'Can you tell me a story?'
-response, history = model.chat(tokenizer, None, question, generation_config, history=history, return_history=True)
-print(f'User: {question}\nAssistant: {response}')
-
-# single-image single-round conversation (单图单轮对话)
-question = '<image>\nPlease describe the image shortly.'
+# single-round single-image conversation
+question = "请详细描述图片"
 response = model.chat(tokenizer, pixel_values, question, generation_config)
-print(f'User: {question}\nAssistant: {response}')
+print(question, response)
 
-# single-image multi-round conversation (单图多轮对话)
-question = '<image>\nPlease describe the image in detail.'
+# multi-round single-image conversation
+question = "请详细描述图片"
 response, history = model.chat(tokenizer, pixel_values, question, generation_config, history=None, return_history=True)
-print(f'User: {question}\nAssistant: {response}')
+print(question, response)
 
-question = 'Please write a poem according to the image.'
+question = "请根据图片写一首诗"
 response, history = model.chat(tokenizer, pixel_values, question, generation_config, history=history, return_history=True)
-print(f'User: {question}\nAssistant: {response}')
+print(question, response)
 
-# multi-image multi-round conversation, combined images (多图多轮对话，拼接图像)
-pixel_values1 = load_image('./examples/image1.jpg', max_num=12).to(torch.bfloat16).cuda()
-pixel_values2 = load_image('./examples/image2.jpg', max_num=12).to(torch.bfloat16).cuda()
+# multi-round multi-image conversation
+pixel_values1 = load_image('./examples/image1.jpg', max_num=6).to(torch.bfloat16).cuda()
+pixel_values2 = load_image('./examples/image2.jpg', max_num=6).to(torch.bfloat16).cuda()
 pixel_values = torch.cat((pixel_values1, pixel_values2), dim=0)
 
-question = '<image>\nDescribe the two images in detail.'
-response, history = model.chat(tokenizer, pixel_values, question, generation_config,
-                               history=None, return_history=True)
-print(f'User: {question}\nAssistant: {response}')
-
-question = 'What are the similarities and differences between these two images.'
-response, history = model.chat(tokenizer, pixel_values, question, generation_config,
-                               history=history, return_history=True)
-print(f'User: {question}\nAssistant: {response}')
-
-# multi-image multi-round conversation, separate images (多图多轮对话，独立图像)
-pixel_values1 = load_image('./examples/image1.jpg', max_num=12).to(torch.bfloat16).cuda()
-pixel_values2 = load_image('./examples/image2.jpg', max_num=12).to(torch.bfloat16).cuda()
-pixel_values = torch.cat((pixel_values1, pixel_values2), dim=0)
-num_patches_list = [pixel_values1.size(0), pixel_values2.size(0)]
-
-question = 'Image-1: <image>\nImage-2: <image>\nDescribe the two images in detail.'
-response, history = model.chat(tokenizer, pixel_values, question, generation_config,
-                               num_patches_list=num_patches_list,
-                               history=None, return_history=True)
-print(f'User: {question}\nAssistant: {response}')
-
-question = 'What are the similarities and differences between these two images.'
-response, history = model.chat(tokenizer, pixel_values, question, generation_config,
-                               num_patches_list=num_patches_list,
-                               history=history, return_history=True)
-print(f'User: {question}\nAssistant: {response}')
-
-# batch inference, single image per sample (单图批处理)
-pixel_values1 = load_image('./examples/image1.jpg', max_num=12).to(torch.bfloat16).cuda()
-pixel_values2 = load_image('./examples/image2.jpg', max_num=12).to(torch.bfloat16).cuda()
-num_patches_list = [pixel_values1.size(0), pixel_values2.size(0)]
-pixel_values = torch.cat((pixel_values1, pixel_values2), dim=0)
-
-questions = ['<image>\nDescribe the image in detail.'] * len(num_patches_list)
-responses = model.batch_chat(tokenizer, pixel_values,
-                             num_patches_list=num_patches_list,
-                             questions=questions,
-                             generation_config=generation_config)
-for question, response in zip(questions, responses):
-    print(f'User: {question}\nAssistant: {response}')
-
-# video multi-round conversation (视频多轮对话)
-def get_index(bound, fps, max_frame, first_idx=0, num_segments=32):
-    if bound:
-        start, end = bound[0], bound[1]
-    else:
-        start, end = -100000, 100000
-    start_idx = max(first_idx, round(start * fps))
-    end_idx = min(round(end * fps), max_frame)
-    seg_size = float(end_idx - start_idx) / num_segments
-    frame_indices = np.array([
-        int(start_idx + (seg_size / 2) + np.round(seg_size * idx))
-        for idx in range(num_segments)
-    ])
-    return frame_indices
-
-def load_video(video_path, bound=None, input_size=448, max_num=1, num_segments=32):
-    vr = VideoReader(video_path, ctx=cpu(0), num_threads=1)
-    max_frame = len(vr) - 1
-    fps = float(vr.get_avg_fps())
-
-    pixel_values_list, num_patches_list = [], []
-    transform = build_transform(input_size=input_size)
-    frame_indices = get_index(bound, fps, max_frame, first_idx=0, num_segments=num_segments)
-    for frame_index in frame_indices:
-        img = Image.fromarray(vr[frame_index].asnumpy()).convert('RGB')
-        img = dynamic_preprocess(img, image_size=input_size, use_thumbnail=True, max_num=max_num)
-        pixel_values = [transform(tile) for tile in img]
-        pixel_values = torch.stack(pixel_values)
-        num_patches_list.append(pixel_values.shape[0])
-        pixel_values_list.append(pixel_values)
-    pixel_values = torch.cat(pixel_values_list)
-    return pixel_values, num_patches_list
-
-video_path = './examples/red-panda.mp4'
-pixel_values, num_patches_list = load_video(video_path, num_segments=8, max_num=1)
-pixel_values = pixel_values.to(torch.bfloat16).cuda()
-video_prefix = ''.join([f'Frame{i+1}: <image>\n' for i in range(len(num_patches_list))])
-question = video_prefix + 'What is the red panda doing?'
-# Frame1: <image>\nFrame2: <image>\n...\nFrame8: <image>\n{question}
-response, history = model.chat(tokenizer, pixel_values, question, generation_config,
-                               num_patches_list=num_patches_list, history=None, return_history=True)
-print(f'User: {question}\nAssistant: {response}')
-
-question = 'Describe this video in detail.'
-response, history = model.chat(tokenizer, pixel_values, question, generation_config,
-                               num_patches_list=num_patches_list, history=history, return_history=True)
-print(f'User: {question}\nAssistant: {response}')
-```
-
-#### Streaming Output
-
-Besides this method, you can also use the following code to get streamed output.
-
-```python
-from transformers import TextIteratorStreamer
-from threading import Thread
-
-# Initialize the streamer
-streamer = TextIteratorStreamer(tokenizer, skip_prompt=True, skip_special_tokens=True, timeout=10)
-# Define the generation configuration
-generation_config = dict(max_new_tokens=1024, do_sample=False, streamer=streamer)
-# Start the model chat in a separate thread
-thread = Thread(target=model.chat, kwargs=dict(
-    tokenizer=tokenizer, pixel_values=pixel_values, question=question,
-    history=None, return_history=False, generation_config=generation_config,
-))
-thread.start()
-
-# Initialize an empty string to store the generated text
-generated_text = ''
-# Loop through the streamer to get the new text as it is generated
-for new_text in streamer:
-    if new_text == model.conv_template.sep:
-        break
-    generated_text += new_text
-    print(new_text, end='', flush=True)  # Print each new chunk of generated text on the same line
-```
-
-## Finetune
-
-Many repositories now support fine-tuning of the InternVL series models, including [InternVL](https://github.com/OpenGVLab/InternVL), [SWIFT](https://github.com/modelscope/ms-swift), [XTurner](https://github.com/InternLM/xtuner), and others. Please refer to their documentation for more details on fine-tuning.
-
-## Deployment
-
-### LMDeploy
-
-LMDeploy is a toolkit for compressing, deploying, and serving LLMs & VLMs.
-
-```sh
-pip install lmdeploy>=0.5.3
-```
-
-LMDeploy abstracts the complex inference process of multi-modal Vision-Language Models (VLM) into an easy-to-use pipeline, similar to the Large Language Model (LLM) inference pipeline.
-
-#### A 'Hello, world' Example
-
-```python
-from lmdeploy import pipeline, TurbomindEngineConfig
-from lmdeploy.vl import load_image
-
-model = 'OpenGVLab/InternVL-Chat-V1-5'
-image = load_image('https://raw.githubusercontent.com/open-mmlab/mmdeploy/main/tests/data/tiger.jpeg')
-pipe = pipeline(model, backend_config=TurbomindEngineConfig(session_len=8192))
-response = pipe(('describe this image', image))
-print(response.text)
-```
-
-If `ImportError` occurs while executing this case, please install the required dependency packages as prompted.
-
-#### Multi-images Inference
-
-When dealing with multiple images, you can put them all in one list. Keep in mind that multiple images will lead to a higher number of input tokens, and as a result, the size of the context window typically needs to be increased.
-
-> Warning: Due to the scarcity of multi-image conversation data, the performance on multi-image tasks may be unstable, and it may require multiple attempts to achieve satisfactory results.
-
-```python
-from lmdeploy import pipeline, TurbomindEngineConfig
-from lmdeploy.vl import load_image
-from lmdeploy.vl.constants import IMAGE_TOKEN
-
-model = 'OpenGVLab/InternVL-Chat-V1-5'
-pipe = pipeline(model, backend_config=TurbomindEngineConfig(session_len=8192))
-
-image_urls=[
-    'https://raw.githubusercontent.com/open-mmlab/mmdeploy/main/demo/resources/human-pose.jpg',
-    'https://raw.githubusercontent.com/open-mmlab/mmdeploy/main/demo/resources/det.jpg'
-]
-
-images = [load_image(img_url) for img_url in image_urls]
-# Numbering images improves multi-image conversations
-response = pipe((f'Image-1: {IMAGE_TOKEN}\nImage-2: {IMAGE_TOKEN}\ndescribe these two images', images))
-print(response.text)
-```
-
-#### Batch Prompts Inference
-
-Conducting inference with batch prompts is quite straightforward; just place them within a list structure:
-
-```python
-from lmdeploy import pipeline, TurbomindEngineConfig
-from lmdeploy.vl import load_image
-
-model = 'OpenGVLab/InternVL-Chat-V1-5'
-pipe = pipeline(model, backend_config=TurbomindEngineConfig(session_len=8192))
-
-image_urls=[
-    "https://raw.githubusercontent.com/open-mmlab/mmdeploy/main/demo/resources/human-pose.jpg",
-    "https://raw.githubusercontent.com/open-mmlab/mmdeploy/main/demo/resources/det.jpg"
-]
-prompts = [('describe this image', load_image(img_url)) for img_url in image_urls]
-response = pipe(prompts)
-print(response)
-```
-
-#### Multi-turn Conversation
-
-There are two ways to do the multi-turn conversations with the pipeline. One is to construct messages according to the format of OpenAI and use above introduced method, the other is to use the `pipeline.chat` interface.
-
-```python
-from lmdeploy import pipeline, TurbomindEngineConfig, GenerationConfig
-from lmdeploy.vl import load_image
-
-model = 'OpenGVLab/InternVL-Chat-V1-5'
-pipe = pipeline(model, backend_config=TurbomindEngineConfig(session_len=8192))
-
-image = load_image('https://raw.githubusercontent.com/open-mmlab/mmdeploy/main/demo/resources/human-pose.jpg')
-gen_config = GenerationConfig(top_k=40, top_p=0.8, temperature=0.8)
-sess = pipe.chat(('describe this image', image), gen_config=gen_config)
-print(sess.response.text)
-sess = pipe.chat('What is the woman doing?', session=sess, gen_config=gen_config)
-print(sess.response.text)
-```
-
-#### Service
-
-LMDeploy's `api_server` enables models to be easily packed into services with a single command. The provided RESTful APIs are compatible with OpenAI's interfaces. Below are an example of service startup:
+question = "详细描述这两张图片"
+response, history = model.chat(tokenizer, pixel_values, question, generation_config, history=None, return_history=True)
+print(question, response)
+# 第一张图片是一只红熊猫，它有着独特的橙红色皮毛，脸部、耳朵和四肢的末端有白色斑块。红熊猫的眼睛周围有深色的环，它的耳朵是圆形的，上面有白色的毛。它正坐在一个木制的结构上，看起来像是一个平台或休息的地方。背景中有树木和竹子，这表明红熊猫可能在一个模拟自然环境的动物园或保护区内。
+#
+# 第二张图片是一只大熊猫，它是中国的国宝，以其黑白相间的皮毛而闻名。大熊猫的眼睛、耳朵和四肢的末端是黑色的，而它的脸部、耳朵内侧和身体其他部分是白色的。大熊猫正坐在地上，周围有竹子，这是它们的主要食物来源。背景中也有树木，这表明大熊猫可能在一个为它们提供自然栖息地模拟的动物园或保护区内。
 
-```shell
-lmdeploy serve api_server OpenGVLab/InternVL-Chat-V1-5 --server-port 23333
+question = "这两张图片的相同点和区别分别是什么"
+response, history = model.chat(tokenizer, pixel_values, question, generation_config, history=history, return_history=True)
+print(question, response)
+# 这两张图片的相同点：
+# 
+# 1. 都展示了熊猫，这是两种不同的熊猫物种。
+# 2. 熊猫都处于一个看起来像是模拟自然环境的场所，可能是动物园或保护区。
+# 3. 熊猫周围都有竹子，这是它们的主要食物来源。
+# 
+# 这两张图片的区别：
+# 
+# 1. 熊猫的种类不同：第一张图片是一只红熊猫，第二张图片是一只大熊猫。
+# 2. 熊猫的皮毛颜色和图案不同：红熊猫的皮毛是橙红色，脸部、耳朵和四肢的末端有白色斑块；而大熊猫的皮毛是黑白相间的，眼睛、耳朵和四肢的末端是黑色的，脸部、耳朵内侧和身体其他部分是白色的。
+# 3. 熊猫的姿态和位置不同：红熊猫坐在一个木制的结构上，而大熊猫坐在地上。
+# 4. ���景中的植被和环境细节略有不同，但都包含树木和竹子。
 ```
 
-To use the OpenAI-style interface, you need to install OpenAI:
-
-```shell
-pip install openai
-```
+## Citation
 
-Then, use the code below to make the API call:
+If you find this project useful in your research, please consider citing:
 
-```python
-from openai import OpenAI
-
-client = OpenAI(api_key='YOUR_API_KEY', base_url='http://0.0.0.0:23333/v1')
-model_name = client.models.list().data[0].id
-response = client.chat.completions.create(
-    model=model_name,
-    messages=[{
-        'role':
-        'user',
-        'content': [{
-            'type': 'text',
-            'text': 'describe this image',
-        }, {
-            'type': 'image_url',
-            'image_url': {
-                'url':
-                'https://modelscope.oss-cn-beijing.aliyuncs.com/resource/tiger.jpeg',
-            },
-        }],
-    }],
-    temperature=0.8,
-    top_p=0.8)
-print(response)
+```BibTeX
+@article{chen2023internvl,
+  title={InternVL: Scaling up Vision Foundation Models and Aligning for Generic Visual-Linguistic Tasks},
+  author={Chen, Zhe and Wu, Jiannan and Wang, Wenhai and Su, Weijie and Chen, Guo and Xing, Sen and Zhong, Muyan and Zhang, Qinglong and Zhu, Xizhou and Lu, Lewei and Li, Bin and Luo, Ping and Lu, Tong and Qiao, Yu and Dai, Jifeng},
+  journal={arXiv preprint arXiv:2312.14238},
+  year={2023}
+}
 ```
 
 ## License
 
-This project is released under the MIT License. This project uses the pre-trained internlm2-chat-20b as a component, which is licensed under the Apache License 2.0.
-
-## Citation
+This project is released under the MIT license. 
 
-If you find this project useful in your research, please consider citing:
+## Acknowledgement
 
-```BibTeX
-@article{chen2024expanding,
-  title={Expanding Performance Boundaries of Open-Source Multimodal Models with Model, Data, and Test-Time Scaling},
-  author={Chen, Zhe and Wang, Weiyun and Cao, Yue and Liu, Yangzhou and Gao, Zhangwei and Cui, Erfei and Zhu, Jinguo and Ye, Shenglong and Tian, Hao and Liu, Zhaoyang and others},
-  journal={arXiv preprint arXiv:2412.05271},
-  year={2024}
-}
-@article{gao2024mini,
-  title={Mini-internvl: A flexible-transfer pocket multimodal model with 5\% parameters and 90\% performance},
-  author={Gao, Zhangwei and Chen, Zhe and Cui, Erfei and Ren, Yiming and Wang, Weiyun and Zhu, Jinguo and Tian, Hao and Ye, Shenglong and He, Junjun and Zhu, Xizhou and others},
-  journal={arXiv preprint arXiv:2410.16261},
-  year={2024}
-}
-@article{chen2024far,
-  title={How Far Are We to GPT-4V? Closing the Gap to Commercial Multimodal Models with Open-Source Suites},
-  author={Chen, Zhe and Wang, Weiyun and Tian, Hao and Ye, Shenglong and Gao, Zhangwei and Cui, Erfei and Tong, Wenwen and Hu, Kongzhi and Luo, Jiapeng and Ma, Zheng and others},
-  journal={arXiv preprint arXiv:2404.16821},
-  year={2024}
-}
-@inproceedings{chen2024internvl,
-  title={Internvl: Scaling up vision foundation models and aligning for generic visual-linguistic tasks},
-  author={Chen, Zhe and Wu, Jiannan and Wang, Wenhai and Su, Weijie and Chen, Guo and Xing, Sen and Zhong, Muyan and Zhang, Qinglong and Zhu, Xizhou and Lu, Lewei and others},
-  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
-  pages={24185--24198},
-  year={2024}
-}
-```
+InternVL is built with reference to the code of the following projects: [OpenAI CLIP](https://github.com/openai/CLIP), [Open CLIP](https://github.com/mlfoundations/open_clip), [CLIP Benchmark](https://github.com/LAION-AI/CLIP_benchmark), [EVA](https://github.com/baaivision/EVA/tree/master), [InternImage](https://github.com/OpenGVLab/InternImage), [ViT-Adapter](https://github.com/czczup/ViT-Adapter), [MMSegmentation](https://github.com/open-mmlab/mmsegmentation), [Transformers](https://github.com/huggingface/transformers), [DINOv2](https://github.com/facebookresearch/dinov2), [BLIP-2](https://github.com/salesforce/LAVIS/tree/main/projects/blip2), [Qwen-VL](https://github.com/QwenLM/Qwen-VL/tree/master/eval_mm), and [LLaVA-1.5](https://github.com/haotian-liu/LLaVA). Thanks for their awesome work!

all_results.json

@@ -0,0 +1,8 @@
+{
+    "epoch": 1.0,
+    "train_loss": 0.8170236018231988,
+    "train_runtime": 190400.5325,
+    "train_samples": 5155291,
+    "train_samples_per_second": 27.076,
+    "train_steps_per_second": 0.026
+}

config.json

@@ -1,19 +1,19 @@
 {
   "_commit_hash": null,
+  "_name_or_path": "./work_dirs/internvl_chat_internlm2_20b_448_dynamic_chinese_pretrain3/checkpoint-1600_replace_llm",
   "architectures": [
     "InternVLChatModel"
   ],
   "auto_map": {
     "AutoConfig": "configuration_internvl_chat.InternVLChatConfig",
-    "AutoModel": "modeling_internvl_chat.InternVLChatModel",
-    "AutoModelForCausalLM": "modeling_internvl_chat.InternVLChatModel"
+    "AutoModel": "modeling_internvl_chat.InternVLChatModel"
   },
-  "system_message": "You are an AI assistant whose name is InternLM (书生·浦语).",
   "downsample_ratio": 0.5,
   "dynamic_image_size": true,
   "force_image_size": 448,
+  "image_fold": null,
   "llm_config": {
-    "_name_or_path": "internlm/internlm2-chat-20b",
+    "_name_or_path": "pretrained/internlm2-chat-20b/",
     "add_cross_attention": false,
     "architectures": [
       "InternLM2ForCausalLM"
@@ -95,49 +95,108 @@
     "top_p": 1.0,
     "torch_dtype": "bfloat16",
     "torchscript": false,
-    "transformers_version": "4.37.2",
+    "transformers_version": "4.36.2",
     "typical_p": 1.0,
-    "use_bfloat16": true,
-    "use_cache": true,
+    "use_bfloat16": false,
+    "use_cache": false,
     "vocab_size": 92553
   },
-  "max_dynamic_patch": 12,
+  "max_dynamic_patch": 6,
   "min_dynamic_patch": 1,
   "model_type": "internvl_chat",
+  "pad2square": false,
   "ps_version": "v2",
   "select_layer": -1,
   "template": "internlm2-chat",
   "torch_dtype": "bfloat16",
+  "transformers_version": null,
   "use_backbone_lora": 0,
   "use_llm_lora": 0,
   "use_thumbnail": true,
   "vision_config": {
+    "_name_or_path": "work_dirs/internvl_chat_internlm2_20b_448_dynamic_chinese_pretrain/checkpoint-5200-vit",
+    "add_cross_attention": false,
     "architectures": [
       "InternVisionModel"
     ],
     "attention_dropout": 0.0,
-    "drop_path_rate": 0.0,
+    "auto_map": {
+      "AutoConfig": "configuration_intern_vit.InternVisionConfig",
+      "AutoModel": "modeling_intern_vit.InternVisionModel"
+    },
+    "bad_words_ids": null,
+    "begin_suppress_tokens": null,
+    "bos_token_id": null,
+    "chunk_size_feed_forward": 0,
+    "cross_attention_hidden_size": null,
+    "decoder_start_token_id": null,
+    "diversity_penalty": 0.0,
+    "do_sample": false,
+    "drop_path_rate": 0.4,
     "dropout": 0.0,
+    "early_stopping": false,
+    "encoder_no_repeat_ngram_size": 0,
+    "eos_token_id": null,
+    "exponential_decay_length_penalty": null,
+    "finetuning_task": null,
+    "forced_bos_token_id": null,
+    "forced_eos_token_id": null,
     "hidden_act": "gelu",
     "hidden_size": 3200,
+    "id2label": {
+      "0": "LABEL_0",
+      "1": "LABEL_1"
+    },
     "image_size": 448,
     "initializer_factor": 0.1,
     "initializer_range": 1e-10,
     "intermediate_size": 12800,
+    "is_decoder": false,
+    "is_encoder_decoder": false,
+    "label2id": {
+      "LABEL_0": 0,
+      "LABEL_1": 1
+    },
     "layer_norm_eps": 1e-06,
+    "length_penalty": 1.0,
+    "max_length": 20,
+    "min_length": 0,
     "model_type": "intern_vit_6b",
-    "norm_type": "rms_norm",
+    "no_repeat_ngram_size": 0,
     "num_attention_heads": 25,
+    "num_beam_groups": 1,
+    "num_beams": 1,
     "num_channels": 3,
     "num_hidden_layers": 45,
+    "num_return_sequences": 1,
     "output_attentions": false,
     "output_hidden_states": false,
+    "output_scores": false,
+    "pad_token_id": null,
     "patch_size": 14,
+    "prefix": null,
+    "problem_type": null,
+    "pruned_heads": {},
     "qk_normalization": true,
     "qkv_bias": false,
+    "remove_invalid_values": false,
+    "repetition_penalty": 1.0,
     "return_dict": true,
+    "return_dict_in_generate": false,
+    "sep_token_id": null,
+    "suppress_tokens": null,
+    "task_specific_params": null,
+    "temperature": 1.0,
+    "tf_legacy_loss": false,
+    "tie_encoder_decoder": false,
+    "tie_word_embeddings": true,
+    "tokenizer_class": null,
+    "top_k": 50,
+    "top_p": 1.0,
     "torch_dtype": "bfloat16",
-    "transformers_version": "4.37.2",
+    "torchscript": false,
+    "transformers_version": "4.36.2",
+    "typical_p": 1.0,
     "use_bfloat16": true,
     "use_flash_attn": true
   }

configuration_intern_vit.py

@@ -1,9 +1,8 @@
 # --------------------------------------------------------
 # InternVL
-# Copyright (c) 2024 OpenGVLab
+# Copyright (c) 2023 OpenGVLab
 # Licensed under The MIT License [see LICENSE for details]
 # --------------------------------------------------------
-
 import os
 from typing import Union
 
@@ -74,7 +73,6 @@ class InternVisionConfig(PretrainedConfig):
             num_hidden_layers=48,
             use_flash_attn=True,
             hidden_act='gelu',
-            norm_type='rms_norm',
             layer_norm_eps=1e-6,
             dropout=0.0,
             drop_path_rate=0.0,
@@ -99,7 +97,6 @@ class InternVisionConfig(PretrainedConfig):
         self.attention_dropout = attention_dropout
         self.layer_norm_eps = layer_norm_eps
         self.hidden_act = hidden_act
-        self.norm_type = norm_type
         self.qkv_bias = qkv_bias
         self.qk_normalization = qk_normalization
         self.use_flash_attn = use_flash_attn

configuration_internvl_chat.py

@@ -1,6 +1,6 @@
 # --------------------------------------------------------
 # InternVL
-# Copyright (c) 2024 OpenGVLab
+# Copyright (c) 2023 OpenGVLab
 # Licensed under The MIT License [see LICENSE for details]
 # --------------------------------------------------------
 
@@ -26,10 +26,12 @@ class InternVLChatConfig(PretrainedConfig):
             llm_config=None,
             use_backbone_lora=0,
             use_llm_lora=0,
-            select_layer=-1,
+            pad2square=False,
+            select_layer=-4,
             force_image_size=None,
             downsample_ratio=0.5,
             template=None,
+            image_fold=False,
             dynamic_image_size=False,
             use_thumbnail=False,
             ps_version='v1',
@@ -39,26 +41,28 @@ class InternVLChatConfig(PretrainedConfig):
         super().__init__(**kwargs)
 
         if vision_config is None:
-            vision_config = {'architectures': ['InternVisionModel']}
+            vision_config = {}
             logger.info('vision_config is None. Initializing the InternVisionConfig with default values.')
 
         if llm_config is None:
-            llm_config = {'architectures': ['InternLM2ForCausalLM']}
+            llm_config = {}
             logger.info('llm_config is None. Initializing the LlamaConfig config with default values (`LlamaConfig`).')
 
         self.vision_config = InternVisionConfig(**vision_config)
-        if llm_config.get('architectures')[0] == 'LlamaForCausalLM':
+        if llm_config['architectures'][0] == 'LlamaForCausalLM':
             self.llm_config = LlamaConfig(**llm_config)
-        elif llm_config.get('architectures')[0] == 'InternLM2ForCausalLM':
+        elif llm_config['architectures'][0] == 'InternLM2ForCausalLM':
             self.llm_config = InternLM2Config(**llm_config)
         else:
-            raise ValueError('Unsupported architecture: {}'.format(llm_config.get('architectures')[0]))
+            raise ValueError('Unsupported architecture: {}'.format(llm_config['architectures'][0]))
         self.use_backbone_lora = use_backbone_lora
         self.use_llm_lora = use_llm_lora
+        self.pad2square = pad2square
         self.select_layer = select_layer
         self.force_image_size = force_image_size
         self.downsample_ratio = downsample_ratio
         self.template = template
+        self.image_fold = image_fold
         self.dynamic_image_size = dynamic_image_size
         self.use_thumbnail = use_thumbnail
         self.ps_version = ps_version  # pixel shuffle version
@@ -66,6 +70,7 @@ class InternVLChatConfig(PretrainedConfig):
         self.max_dynamic_patch = max_dynamic_patch
 
         logger.info(f'vision_select_layer: {self.select_layer}')
+        logger.info(f'image_fold: {self.image_fold}')
         logger.info(f'ps_version: {self.ps_version}')
         logger.info(f'min_dynamic_patch: {self.min_dynamic_patch}')
         logger.info(f'max_dynamic_patch: {self.max_dynamic_patch}')
@@ -83,10 +88,12 @@ class InternVLChatConfig(PretrainedConfig):
         output['model_type'] = self.__class__.model_type
         output['use_backbone_lora'] = self.use_backbone_lora
         output['use_llm_lora'] = self.use_llm_lora
+        output['pad2square'] = self.pad2square
         output['select_layer'] = self.select_layer
         output['force_image_size'] = self.force_image_size
         output['downsample_ratio'] = self.downsample_ratio
         output['template'] = self.template
+        output['image_fold'] = self.image_fold
         output['dynamic_image_size'] = self.dynamic_image_size
         output['use_thumbnail'] = self.use_thumbnail
         output['ps_version'] = self.ps_version

conversation.py

@@ -2,7 +2,7 @@
 Conversation prompt templates.
 
 We kindly request that you import fastchat instead of copying this file if you wish to use it.
-If you have changes in mind, please contribute back so the community can benefit collectively and continue to maintain these valuable templates.
+If you have any changes in mind, please contribute back so the community can benefit collectively and continue to maintain these valuable templates.
 """
 
 import dataclasses
@@ -330,6 +330,384 @@ def get_conv_template(name: str) -> Conversation:
     return conv_templates[name].copy()
 
 
+# An empty template for raw conversation.
+register_conv_template(
+    Conversation(
+        name='raw',
+        system_message='',
+        roles=('', ''),
+        sep_style=SeparatorStyle.NO_COLON_SINGLE,
+        sep='',
+    )
+)
+
+# A template with a one-shot conversation example
+register_conv_template(
+    Conversation(
+        name='one_shot',
+        system_message='A chat between a curious human and an artificial intelligence assistant. '
+        "The assistant gives helpful, detailed, and polite answers to the human's questions.",
+        roles=('Human', 'Assistant'),
+        messages=(
+            (
+                'Human',
+                'Got any creative ideas for a 10 year old’s birthday?',
+            ),
+            (
+                'Assistant',
+                """Of course! Here are some creative ideas for a 10-year-old's birthday party:
+1. Treasure Hunt: Organize a treasure hunt in your backyard or nearby park. Create clues and riddles for the kids to solve, leading them to hidden treasures and surprises.
+2. Science Party: Plan a science-themed party where kids can engage in fun and interactive experiments. You can set up different stations with activities like making slime, erupting volcanoes, or creating simple chemical reactions.
+3. Outdoor Movie Night: Set up a backyard movie night with a projector and a large screen or white sheet. Create a cozy seating area with blankets and pillows, and serve popcorn and snacks while the kids enjoy a favorite movie under the stars.
+4. DIY Crafts Party: Arrange a craft party where kids can unleash their creativity. Provide a variety of craft supplies like beads, paints, and fabrics, and let them create their own unique masterpieces to take home as party favors.
+5. Sports Olympics: Host a mini Olympics event with various sports and games. Set up different stations for activities like sack races, relay races, basketball shooting, and obstacle courses. Give out medals or certificates to the participants.
+6. Cooking Party: Have a cooking-themed party where the kids can prepare their own mini pizzas, cupcakes, or cookies. Provide toppings, frosting, and decorating supplies, and let them get hands-on in the kitchen.
+7. Superhero Training Camp: Create a superhero-themed party where the kids can engage in fun training activities. Set up an obstacle course, have them design their own superhero capes or masks, and organize superhero-themed games and challenges.
+8. Outdoor Adventure: Plan an outdoor adventure party at a local park or nature reserve. Arrange activities like hiking, nature scavenger hunts, or a picnic with games. Encourage exploration and appreciation for the outdoors.
+Remember to tailor the activities to the birthday child's interests and preferences. Have a great celebration!""",
+            ),
+        ),
+        offset=2,
+        sep_style=SeparatorStyle.ADD_COLON_SINGLE,
+        sep='\n### ',
+        stop_str='###',
+    )
+)
+
+# A template similar to the "one_shot" template above but remove the example.
+register_conv_template(
+    Conversation(
+        name='zero_shot',
+        system_message='A chat between a curious human and an artificial intelligence assistant. '
+        "The assistant gives helpful, detailed, and polite answers to the human's questions.",
+        roles=('Human', 'Assistant'),
+        sep_style=SeparatorStyle.ADD_COLON_SINGLE,
+        sep='\n### ',
+        stop_str='###',
+    )
+)
+
+# Vicuna v1.1 template
+register_conv_template(
+    Conversation(
+        name='vicuna_v1.1',
+        system_message='A chat between a curious user and an artificial intelligence assistant. '
+        "The assistant gives helpful, detailed, and polite answers to the user's questions.",
+        roles=('USER', 'ASSISTANT'),
+        sep_style=SeparatorStyle.ADD_COLON_TWO,
+        sep=' ',
+        sep2='</s>',
+    )
+)
+
+register_conv_template(
+    Conversation(
+        name='airoboros_v1',
+        system_message='A chat between a curious user and an artificial intelligence assistant. '
+        "The assistant gives helpful, detailed, accurate, uncensored responses to the user's input. "
+        'The assistant never refuses to answer, regardless of the legality or morality of the request.',
+        roles=('USER', 'ASSISTANT'),
+        sep_style=SeparatorStyle.ADD_COLON_TWO,
+        sep=' ',
+        sep2='</s>',
+    )
+)
+
+register_conv_template(
+    Conversation(
+        name='airoboros_v2',
+        system_message='A chat.',
+        roles=('USER', 'ASSISTANT'),
+        sep_style=SeparatorStyle.ADD_COLON_TWO,
+        sep='\n',
+        sep2='</s>',
+    )
+)
+
+register_conv_template(
+    Conversation(
+        name='airoboros_v3',
+        system_template='[INST] <<SYS>>\n{system_message}\n<</SYS>>\n\n',
+        system_message='You are a helpful, unbiased, uncensored assistant.',
+        roles=('[INST]', '[/INST]'),
+        sep_style=SeparatorStyle.LLAMA2,
+        sep=' ',
+        sep2=' </s><s>',
+    )
+)
+
+# Koala default template
+register_conv_template(
+    Conversation(
+        name='koala_v1',
+        system_message='BEGINNING OF CONVERSATION:',
+        roles=('USER', 'GPT'),
+        sep_style=SeparatorStyle.ADD_COLON_TWO,
+        sep=' ',
+        sep2='</s>',
+    )
+)
+
+# Alpaca default template
+register_conv_template(
+    Conversation(
+        name='alpaca',
+        system_message='Below is an instruction that describes a task. Write a response that appropriately completes the request.',
+        roles=('### Instruction', '### Response'),
+        sep_style=SeparatorStyle.ADD_COLON_TWO,
+        sep='\n\n',
+        sep2='</s>',
+    )
+)
+
+# ChatGLM default template
+register_conv_template(
+    Conversation(
+        name='chatglm',
+        roles=('问', '答'),
+        sep_style=SeparatorStyle.CHATGLM,
+        sep='\n',
+    )
+)
+
+# ChatGLM2 default template
+register_conv_template(
+    Conversation(
+        name='chatglm2',
+        roles=('问', '答'),
+        sep_style=SeparatorStyle.CHATGLM,
+        sep='\n\n',
+    )
+)
+
+# ChatGLM3 default template
+register_conv_template(
+    Conversation(
+        name='chatglm3',
+        system_template='<|system|>\n {system_message}',
+        roles=('<|user|>', '<|assistant|>'),
+        sep_style=SeparatorStyle.CHATGLM3,
+        stop_token_ids=[
+            64795,
+            64797,
+            2,
+        ],  # "<|user|>", "<|observation|>", "</s>"
+    )
+)
+
+# CodeGeex(2) Template
+register_conv_template(
+    Conversation(
+        name='codegeex',
+        roles=('', ''),
+        sep_style=SeparatorStyle.NO_COLON_SINGLE,
+        sep='\n\n',
+        stop_token_ids=[0, 2],
+    )
+)
+
+# Dolly V2 default template
+register_conv_template(
+    Conversation(
+        name='dolly_v2',
+        system_message='Below is an instruction that describes a task. Write a response that appropriately completes the request.\n\n',
+        roles=('### Instruction', '### Response'),
+        sep_style=SeparatorStyle.DOLLY,
+        sep='\n\n',
+        sep2='### End',
+    )
+)
+
+# OpenAssistant Pythia default template
+register_conv_template(
+    Conversation(
+        name='oasst_pythia',
+        roles=('<|prompter|>', '<|assistant|>'),
+        sep_style=SeparatorStyle.NO_COLON_SINGLE,
+        sep='<|endoftext|>',
+    )
+)
+
+# OpenAssistant default template
+register_conv_template(
+    Conversation(
+        name='oasst_llama',
+        roles=('<|prompter|>', '<|assistant|>'),
+        sep_style=SeparatorStyle.NO_COLON_SINGLE,
+        sep='</s>',
+    )
+)
+
+# OpenChat 3.5 default template
+register_conv_template(
+    Conversation(
+        name='openchat_3.5',
+        roles=('GPT4 Correct User', 'GPT4 Correct Assistant'),
+        sep_style=SeparatorStyle.FALCON_CHAT,
+        sep='<|end_of_turn|>',
+    )
+)
+
+# Tulu default template
+register_conv_template(
+    Conversation(
+        name='tulu',
+        roles=('<|user|>', '<|assistant|>'),
+        sep_style=SeparatorStyle.ADD_NEW_LINE_SINGLE,
+        sep='\n',
+    )
+)
+
+# StableLM Alpha default template
+register_conv_template(
+    Conversation(
+        name='stablelm',
+        system_template='<|SYSTEM|>{system_message}',
+        system_message="""# StableLM Tuned (Alpha version)
+- StableLM is a helpful and harmless open-source AI language model developed by StabilityAI.
+- StableLM is excited to be able to help the user, but will refuse to do anything that could be considered harmful to the user.
+- StableLM is more than just an information source, StableLM is also able to write poetry, short stories, and make jokes.
+- StableLM will refuse to participate in anything that could harm a human.
+""",
+        roles=('<|USER|>', '<|ASSISTANT|>'),
+        sep_style=SeparatorStyle.NO_COLON_SINGLE,
+        sep='',
+        stop_token_ids=[50278, 50279, 50277, 1, 0],
+    )
+)
+
+# Baize default template
+register_conv_template(
+    Conversation(
+        name='baize',
+        system_message='The following is a conversation between a human and an AI assistant named Baize (named after a mythical creature in Chinese folklore). Baize is an open-source AI assistant developed by UCSD and Sun Yat-Sen University. The human and the AI assistant take turns chatting. Human statements start with [|Human|] and AI assistant statements start with [|AI|]. The AI assistant always provides responses in as much detail as possible, and in Markdown format. The AI assistant always declines to engage with topics, questions and instructions related to unethical, controversial, or sensitive issues. Complete the transcript in exactly that format.\n',
+        roles=('[|Human|]', '[|AI|]'),
+        messages=(
+            ('[|Human|]', 'Hello!'),
+            ('[|AI|]', 'Hi!'),
+        ),
+        offset=2,
+        sep_style=SeparatorStyle.NO_COLON_SINGLE,
+        sep='\n',
+        stop_str='[|Human|]',
+    )
+)
+
+# RWKV-4-Raven default template
+register_conv_template(
+    Conversation(
+        name='rwkv',
+        roles=('Bob', 'Alice'),
+        messages=(
+            ('Bob', 'hi'),
+            (
+                'Alice',
+                'Hi. I am your assistant and I will provide expert full response in full details. Please feel free to ask any question and I will always answer it.',
+            ),
+        ),
+        offset=2,
+        sep_style=SeparatorStyle.RWKV,
+        sep='',
+        stop_str='\n\n',
+    )
+)
+
+# Buddy default template
+register_conv_template(
+    Conversation(
+        name='openbuddy',
+        system_message="""Consider a conversation between User (a human) and Assistant (named Buddy).
+Buddy is an INTP-T, a friendly, intelligent and multilingual AI assistant, by OpenBuddy team. GitHub: https://github.com/OpenBuddy/OpenBuddy
+Buddy cannot access the Internet.
+Buddy can fluently speak the user's language (e.g. English, Chinese).
+Buddy can generate poems, stories, code, essays, songs, parodies, and more.
+Buddy possesses vast knowledge about the world, history, and culture.
+Buddy's responses are always safe, creative, high-quality, human-like, and interesting.
+Buddy strictly refuses to discuss political, NSFW, or other unsafe topics.
+
+User: Hi.
+Assistant: Hi, I'm Buddy, your AI assistant. How can I help you today?""",
+        roles=('User', 'Assistant'),
+        sep_style=SeparatorStyle.ADD_COLON_SINGLE,
+        sep='\n',
+    )
+)
+
+# Phoenix default template
+register_conv_template(
+    Conversation(
+        name='phoenix',
+        system_message="A chat between a curious human and an artificial intelligence assistant. The assistant gives helpful, detailed, and polite answers to the human's questions.\n\n",
+        roles=('Human', 'Assistant'),
+        sep_style=SeparatorStyle.PHOENIX,
+        sep='</s>',
+    )
+)
+
+# ReaLM default template
+register_conv_template(
+    Conversation(
+        name='ReaLM-7b-v1',
+        system_message="A chat between a curious human and an artificial intelligence assistant. The assistant gives helpful, detailed, and polite answers to the human's questions.\n\n",
+        roles=('Human', 'Assistant'),
+        sep_style=SeparatorStyle.PHOENIX,
+        sep='</s>',
+    )
+)
+
+# ChatGPT default template
+register_conv_template(
+    Conversation(
+        name='chatgpt',
+        system_message='You are a helpful assistant.',
+        roles=('user', 'assistant'),
+        sep_style=None,
+        sep=None,
+    )
+)
+
+# Claude default template
+register_conv_template(
+    Conversation(
+        name='claude',
+        roles=('Human', 'Assistant'),
+        sep_style=SeparatorStyle.ADD_COLON_SINGLE,
+        sep='\n\n',
+    )
+)
+
+# MPT default template
+register_conv_template(
+    Conversation(
+        name='mpt-7b-chat',
+        system_template="""<|im_start|>system
+{system_message}""",
+        system_message="""- You are a helpful assistant chatbot trained by MosaicML.
+- You answer questions.
+- You are excited to be able to help the user, but will refuse to do anything that could be considered harmful to the user.
+- You are more than just an information source, you are also able to write poetry, short stories, and make jokes.""",
+        roles=('<|im_start|>user', '<|im_start|>assistant'),
+        sep_style=SeparatorStyle.CHATML,
+        sep='<|im_end|>',
+        stop_token_ids=[50278, 0],
+    )
+)
+
+# MPT-30b-chat default template
+register_conv_template(
+    Conversation(
+        name='mpt-30b-chat',
+        system_template="""<|im_start|>system
+{system_message}""",
+        system_message="""A conversation between a user and an LLM-based AI assistant. The assistant gives helpful and honest answers.""",
+        roles=('<|im_start|>user', '<|im_start|>assistant'),
+        sep_style=SeparatorStyle.CHATML,
+        sep='<|im_end|>',
+        stop_token_ids=[50278, 0],
+    )
+)
+
+
 register_conv_template(
     Conversation(
         name='Hermes-2',
@@ -343,7 +721,7 @@ register_conv_template(
             6,
             7,
             8,
-        ],
+        ],  # "<|endoftext|>", "<|im_start|>", "<|im_end|>", "<|im_sep|>"
         stop_str='<|endoftext|>',
     )
 )
@@ -365,19 +743,519 @@ register_conv_template(
     )
 )
 
+# Lemur-70b-chat default template
+# reference: https://huggingface.co/OpenLemur/lemur-70b-chat-v1#generation
+register_conv_template(
+    Conversation(
+        name='lemur-70b-chat',
+        system_template="""<|im_start|>system
+{system_message}""",
+        system_message="""You are a helpful, respectful, and honest assistant.""",
+        roles=('<|im_start|>user', '<|im_start|>assistant'),
+        sep_style=SeparatorStyle.CHATML,
+        sep='<|im_end|>',
+        stop_token_ids=[32002, 0],
+    )
+)
+
+# MPT-30b-instruct default template
+# reference: https://huggingface.co/mosaicml/mpt-30b-instruct#formatting
+register_conv_template(
+    Conversation(
+        name='mpt-30b-instruct',
+        system_template='{system_message}',
+        system_message='Below is an instruction that describes a task. Write a response that appropriately completes the request.',
+        roles=('### Instruction', '### Response'),
+        sep_style=SeparatorStyle.ADD_NEW_LINE_SINGLE,
+        sep='\n\n',
+        stop_token_ids=[50278, 0],
+    )
+)
 
+# Bard default template
+# Reference: https://github.com/google/generative-ai-python/blob/9c99bcb474a991a97a2e7d62fcdb52db7ce40729/google/generativeai/discuss.py#L150
+#            https://github.com/google/generative-ai-python/blob/9c99bcb474a991a97a2e7d62fcdb52db7ce40729/google/generativeai/discuss.py#L40
 register_conv_template(
     Conversation(
-        name='phi3-chat',
-        system_template='<|system|>\n{system_message}',
-        system_message='You are an AI assistant whose name is Phi-3.',
-        roles=('<|user|>\n', '<|assistant|>\n'),
-        sep_style=SeparatorStyle.MPT,
-        sep='<|end|>',
+        name='bard',
+        roles=('0', '1'),
+        sep_style=None,
+        sep=None,
+    )
+)
+
+# BiLLa default template
+register_conv_template(
+    Conversation(
+        name='billa',
+        roles=('Human', 'Assistant'),
+        sep_style=SeparatorStyle.ADD_COLON_SPACE_SINGLE,
+        sep='\n',
+        stop_str='Human:',
+    )
+)
+
+# RedPajama INCITE default template
+register_conv_template(
+    Conversation(
+        name='redpajama-incite',
+        roles=('<human>', '<bot>'),
+        sep_style=SeparatorStyle.ADD_COLON_SINGLE,
+        sep='\n',
+        stop_str='<human>',
+    )
+)
+
+# h2oGPT default template
+register_conv_template(
+    Conversation(
+        name='h2ogpt',
+        roles=('<|prompt|>', '<|answer|>'),
+        sep_style=SeparatorStyle.NO_COLON_SINGLE,
+        sep='</s>',
+    )
+)
+
+# Robin default template
+register_conv_template(
+    Conversation(
+        name='Robin',
+        system_message="A chat between a curious human and an artificial intelligence assistant. The assistant gives helpful, detailed, and polite answers to the human's questions.",
+        roles=('###Human', '###Assistant'),
+        sep_style=SeparatorStyle.ROBIN,
+        sep='\n',
+        stop_token_ids=[2, 396],
+        stop_str='###',
+    )
+)
+
+# Snoozy default template
+# Reference: https://github.com/nomic-ai/gpt4all/blob/d4861030b778da6db59d21d2927a4aba4f9f1f43/gpt4all-bindings/python/gpt4all/gpt4all.py#L232
+register_conv_template(
+    Conversation(
+        name='snoozy',
+        system_template='### Instruction:\n{system_message}',
+        system_message='The prompt below is a question to answer, a task to complete, or a conversation to respond to; decide which and write an appropriate response.',
+        roles=('### Prompt', '### Response'),
+        sep_style=SeparatorStyle.ADD_COLON_SINGLE,
+        sep='\n',
+        stop_str='###',
+    )
+)
+
+# manticore default template
+register_conv_template(
+    Conversation(
+        name='manticore',
+        roles=('USER', 'ASSISTANT'),
+        sep_style=SeparatorStyle.ADD_COLON_TWO,
+        sep='\n',
+        sep2='</s>',
+    )
+)
+
+# Falcon default template
+register_conv_template(
+    Conversation(
+        name='falcon',
+        roles=('User', 'Assistant'),
+        messages=[],
+        sep_style=SeparatorStyle.RWKV,
+        sep='\n',
+        sep2='<|endoftext|>',
+        stop_str='\nUser',  # use stop_str to stop generation after stop_token_ids, it will also remove stop_str from the generated text
         stop_token_ids=[
+            0,
+            1,
             2,
-            32000,
-            32007
-        ]
+            3,
+            4,
+            5,
+            6,
+            7,
+            8,
+            9,
+            10,
+            11,
+        ],  # it better only put special tokens here, because tokenizer only remove special tokens
+    )
+)
+
+# ChangGPT default template
+register_conv_template(
+    Conversation(
+        name='polyglot_changgpt',
+        roles=('B', 'A'),
+        sep_style=SeparatorStyle.ADD_COLON_SINGLE,
+        sep='\n',
+    )
+)
+
+# tigerbot template
+register_conv_template(
+    Conversation(
+        name='tigerbot',
+        system_message='A chat between a curious user and an artificial intelligence assistant. '
+        "The assistant gives helpful, detailed, and polite answers to the user's questions.",
+        roles=('### Instruction', '### Response'),
+        sep_style=SeparatorStyle.ROBIN,
+        sep='\n\n',
+        stop_str='###',
+    )
+)
+
+# ref: https://huggingface.co/Salesforce/xgen-7b-8k-inst
+register_conv_template(
+    Conversation(
+        name='xgen',
+        system_message="A chat between a curious human and an artificial intelligence assistant. The assistant gives helpful, detailed, and polite answers to the human's questions.\n\n",
+        roles=('### Human', '### Assistant'),
+        sep_style=SeparatorStyle.ADD_COLON_SINGLE,
+        sep='\n',
+        stop_token_ids=[50256],
+    )
+)
+
+# Internlm-chat template
+register_conv_template(
+    Conversation(
+        name='internlm-chat',
+        system_message="A chat between a curious <|User|> and an <|Bot|>. The <|Bot|> gives helpful, detailed, and polite answers to the <|User|>'s questions.\n\n",
+        roles=('<|User|>', '<|Bot|>'),
+        sep_style=SeparatorStyle.CHATINTERN,
+        sep='<eoh>',
+        sep2='<eoa>',
+        stop_token_ids=[1, 103028],
+        stop_str='<|User|>',
+    )
+)
+
+# StarChat template
+# reference: https://huggingface.co/spaces/HuggingFaceH4/starchat-playground/blob/main/dialogues.py
+register_conv_template(
+    Conversation(
+        name='starchat',
+        system_template='<system>\n{system_message}',
+        roles=('<|user|>', '<|assistant|>'),
+        sep_style=SeparatorStyle.CHATML,
+        sep='<|end|>',
+        stop_token_ids=[0, 49155],
+        stop_str='<|end|>',
+    )
+)
+
+# Baichuan-13B-Chat template
+register_conv_template(
+    # source: https://huggingface.co/baichuan-inc/Baichuan-13B-Chat/blob/19ef51ba5bad8935b03acd20ff04a269210983bc/modeling_baichuan.py#L555
+    # https://huggingface.co/baichuan-inc/Baichuan-13B-Chat/blob/main/generation_config.json
+    # https://github.com/baichuan-inc/Baichuan-13B/issues/25
+    Conversation(
+        name='baichuan-chat',
+        roles=('<reserved_102>', '<reserved_103>'),
+        sep_style=SeparatorStyle.NO_COLON_SINGLE,
+        sep='',
+        stop_token_ids=[],
+    )
+)
+
+# Baichuan2-13B-Chat template
+register_conv_template(
+    # source: https://huggingface.co/baichuan-inc/Baichuan2-13B-Chat/blob/c6f8592a60b4ad73c210b28dd2ab3cca51abbf93/modeling_baichuan.py#L773
+    # https://huggingface.co/baichuan-inc/Baichuan2-13B-Chat/blob/main/generation_config.json
+    # https://github.com/baichuan-inc/Baichuan2/issues/62
+    Conversation(
+        name='baichuan2-chat',
+        roles=('<reserved_106>', '<reserved_107>'),
+        sep_style=SeparatorStyle.NO_COLON_SINGLE,
+        sep='',
+        stop_token_ids=[],
+    )
+)
+
+# Mistral template
+# source: https://docs.mistral.ai/llm/mistral-instruct-v0.1#chat-template
+register_conv_template(
+    Conversation(
+        name='mistral',
+        system_template='[INST]{system_message}\n',
+        roles=('[INST]', '[/INST]'),
+        sep_style=SeparatorStyle.LLAMA2,
+        sep=' ',
+        sep2='</s>',
+    )
+)
+
+# llama2 template
+# reference: https://huggingface.co/blog/codellama#conversational-instructions
+# reference: https://github.com/facebookresearch/llama/blob/1a240688810f8036049e8da36b073f63d2ac552c/llama/generation.py#L212
+register_conv_template(
+    Conversation(
+        name='llama-2',
+        system_template='[INST] <<SYS>>\n{system_message}\n<</SYS>>\n\n',
+        roles=('[INST]', '[/INST]'),
+        sep_style=SeparatorStyle.LLAMA2,
+        sep=' ',
+        sep2=' </s><s>',
+    )
+)
+
+register_conv_template(
+    Conversation(
+        name='cutegpt',
+        roles=('问：', '答：\n'),
+        sep_style=SeparatorStyle.NO_COLON_TWO,
+        sep='\n',
+        sep2='\n',
+        stop_str='<end>',
+    )
+)
+
+# OpenOrcaxOpenChat-naPreview2-13B template
+register_conv_template(
+    Conversation(
+        name='open-orca',
+        system_template='{system_message}',
+        system_message='You are a helpful assistant. Please answer truthfully and write out your '
+        'thinking step by step to be sure you get the right answer. If you make a mistake or encounter '
+        "an error in your thinking, say so out loud and attempt to correct it. If you don't know or "
+        "aren't sure about something, say so clearly. You will act as a professional logician, mathematician, "
+        'and physicist. You will also act as the most appropriate type of expert to answer any particular '
+        'question or solve the relevant problem; state which expert type your are, if so. Also think of '
+        'any particular named expert that would be ideal to answer the relevant question or solve the '
+        'relevant problem; name and act as them, if appropriate.',
+        roles=('User', 'Assistant'),
+        sep_style=SeparatorStyle.ADD_COLON_SPACE_SINGLE,
+        sep='<|end_of_turn|>\n',
+        stop_token_ids=[32000, 32001],  # "<|end_of_turn|>"
+        stop_str='User',
+    )
+)
+
+# Open-Orca/Mistral-7B-OpenOrca template
+# source: https://huggingface.co/Open-Orca/Mistral-7B-OpenOrca
+# reference: https://huggingface.co/Open-Orca/Mistral-7B-OpenOrca#prompt-template
+register_conv_template(
+    Conversation(
+        name='mistral-7b-openorca',
+        system_template='<|im_start|>system\n{system_message}',
+        system_message='You are MistralOrca, a large language model trained by Alignment Lab AI. Write out your reasoning step-by-step to be sure you get the right answers!',
+        roles=('<|im_start|>user', '<|im_start|>assistant'),
+        sep_style=SeparatorStyle.CHATML,
+        sep='<|im_end|>',
+        stop_token_ids=[32000, 32001],
+    )
+)
+
+# Qwen-chat default template
+# source: https://huggingface.co/Qwen/Qwen-7B-Chat/blob/main/qwen_generation_utils.py#L130
+register_conv_template(
+    Conversation(
+        name='qwen-7b-chat',
+        system_template='<|im_start|>system\n{system_message}',
+        system_message='You are a helpful assistant.',
+        roles=('<|im_start|>user', '<|im_start|>assistant'),
+        sep_style=SeparatorStyle.CHATML,
+        sep='<|im_end|>',
+        stop_token_ids=[
+            151643,
+            151644,
+            151645,
+        ],  # "<|endoftext|>", "<|im_start|>", "<|im_end|>"
+        stop_str='<|endoftext|>',
+    )
+)
+
+
+# AquilaChat default template
+# source: https://github.com/FlagAI-Open/FlagAI/blob/master/examples/Aquila/Aquila-chat/cyg_conversation.py
+register_conv_template(
+    Conversation(
+        name='aquila-chat',
+        system_message='A chat between a curious human and an artificial intelligence assistant. '
+        "The assistant gives helpful, detailed, and polite answers to the human's questions.",
+        roles=('Human', 'Assistant'),
+        sep_style=SeparatorStyle.ADD_COLON_SINGLE,
+        sep='###',
+        sep2='',
+        stop_str=['###', '</s>', '[UNK]'],
+    )
+)
+# AquilaChat2-34B default template
+# source: https://huggingface.co/BAAI/AquilaChat2-34B/blob/4608b75855334b93329a771aee03869dbf7d88cc/predict.py#L212
+register_conv_template(
+    Conversation(
+        name='aquila-legacy',
+        system_message='A chat between a curious human and an artificial intelligence assistant. '
+        "The assistant gives helpful, detailed, and polite answers to the human's questions.\n\n",
+        roles=('### Human: ', '### Assistant: '),
+        offset=0,
+        sep_style=SeparatorStyle.NO_COLON_TWO,
+        sep='\n',
+        sep2='</s>',
+        stop_str=['</s>', '[UNK]'],
+    )
+)
+# AquilaChat2-7B-16K and AquilaChat2-34B-16K default template
+# source: https://huggingface.co/BAAI/AquilaChat2-34B/blob/4608b75855334b93329a771aee03869dbf7d88cc/predict.py#L227
+register_conv_template(
+    Conversation(
+        name='aquila',
+        system_message='A chat between a curious human and an artificial intelligence assistant. '
+        "The assistant gives helpful, detailed, and polite answers to the human's questions.",
+        roles=('Human', 'Assistant'),
+        offset=0,
+        sep_style=SeparatorStyle.ADD_COLON_TWO,
+        sep='###',
+        sep2='</s>',
+        stop_str=['</s>', '[UNK]'],
+    )
+)
+
+# AquilaChat2-7B default template
+# source: https://huggingface.co/BAAI/AquilaChat2-34B/blob/4608b75855334b93329a771aee03869dbf7d88cc/predict.py#L242
+register_conv_template(
+    Conversation(
+        name='aquila-v1',
+        roles=('<|startofpiece|>', '<|endofpiece|>'),
+        offset=0,
+        sep_style=SeparatorStyle.NO_COLON_TWO,
+        sep='',
+        sep2='</s>',
+        stop_str=['</s>', '<|endoftext|>'],
+    )
+)
+
+# Llama2-Chinese default template
+# source: https://huggingface.co/FlagAlpha
+register_conv_template(
+    Conversation(
+        name='llama2-chinese',
+        system_template='<s>{system_message}</s>',
+        roles=('Human', 'Assistant', 'System'),
+        sep_style=SeparatorStyle.ADD_COLON_TWO,
+        sep='\n',
+        sep2='\n</s><s>',
+        stop_str='</s>',
     )
 )
+
+# Vigogne Instruct default template
+# source: https://github.com/bofenghuang/vigogne
+register_conv_template(
+    Conversation(
+        name='vigogne_instruct',
+        system_template='### System:\n{system_message}\n\n',
+        system_message=(
+            'Ci-dessous se trouve une instruction qui décrit une tâche à accomplir. Rédigez une réponse qui répond de manière'
+            ' précise à la demande.'
+        ),
+        roles=('### Instruction', '### Response'),
+        sep_style=SeparatorStyle.DOLLY,
+        sep='\n\n',
+        sep2='</s>',
+    )
+)
+
+# Vigogne Chat default template
+register_conv_template(
+    Conversation(
+        name='vigogne_chat_v2',
+        system_template='<|system|>: {system_message}',
+        system_message=(
+            'Vous êtes Vigogne, un assistant IA créé par Zaion Lab. Vous suivez extrêmement bien les instructions. Aidez'
+            ' autant que vous le pouvez.'
+        ),
+        roles=('<|user|>', '<|assistant|>'),
+        sep_style=SeparatorStyle.ADD_COLON_TWO,
+        sep='\n',
+        sep2='</s>\n',
+        stop_str='<|user|>',
+    )
+)
+
+register_conv_template(
+    Conversation(
+        name='vigogne_chat_v3',
+        system_template='[INST] <<SYS>>\n{system_message}\n<</SYS>>\n\n',
+        system_message=(
+            'Vous êtes Vigogne, un assistant IA créé par Zaion Lab. Vous suivez extrêmement bien les instructions. Aidez'
+            ' autant que vous le pouvez.'
+        ),
+        roles=('[INST]', '[/INST]'),
+        sep_style=SeparatorStyle.LLAMA2,
+        sep=' ',
+        sep2=' </s>',
+    )
+)
+
+# Falcon 180B chat template
+# source: https://huggingface.co/spaces/tiiuae/falcon-180b-demo/blob/d1590ee7fae9b6ce331ba7808e61a29dcce9239f/app.py#L28-L37
+register_conv_template(
+    Conversation(
+        name='falcon-chat',
+        roles=('User', 'Falcon'),
+        system_template='System: {system_message}',
+        messages=[],
+        sep_style=SeparatorStyle.FALCON_CHAT,
+        sep='\n',
+        sep2='<|endoftext|>',
+        stop_str='\nUser:',  # use stop_str to stop generation after stop_token_ids, it will also remove stop_str from the generated text
+    )
+)
+
+# Phind template
+# source: https://huggingface.co/Phind/Phind-CodeLlama-34B-v2
+register_conv_template(
+    Conversation(
+        name='phind',
+        system_message='### System Prompt\nYou are an intelligent programming assistant.',
+        roles=('### User Message', '### Assistant'),
+        messages=(),
+        offset=0,
+        sep_style=SeparatorStyle.ADD_COLON_SINGLE,
+        sep='\n\n',
+    )
+)
+
+# Metharme formatting for Pygmalion models
+# source: https://huggingface.co/PygmalionAI/pygmalion-2-13b
+register_conv_template(
+    Conversation(
+        name='metharme',
+        system_template='<|system|>{system_message}',
+        system_message="""Enter RP mode. You shall reply to the user while staying
+        in character. Your responses must be detailed, creative, immersive, and drive the scenario
+        forward.""",
+        roles=('<|user|>', '<|model|>'),
+        sep_style=SeparatorStyle.NO_COLON_SINGLE,
+        sep='',
+        stop_str='<|user|>',
+    )
+)
+
+# Zephyr template
+# reference: https://huggingface.co/spaces/HuggingFaceH4/zephyr-playground/blob/main/dialogues.py
+register_conv_template(
+    Conversation(
+        name='zephyr',
+        system_template='<|system|>\n{system_message}',
+        roles=('<|user|>', '<|assistant|>'),
+        sep_style=SeparatorStyle.CHATML,
+        sep='</s>',
+        stop_token_ids=[2],
+        stop_str='</s>',
+    )
+)
+
+# InternVL-ZH template
+register_conv_template(
+    Conversation(
+        name='internvl_zh',
+        system_template='',
+        roles=('<human>', '<bot>'),
+        sep_style=SeparatorStyle.INTERNVL_ZH,
+        sep=' ',
+        sep2='</s>',
+    )
+)
+

generation_config.json

@@ -1,8 +1,4 @@
 {
   "_from_model_config": true,
-  "transformers_version": "4.37.2",
-  "eos_token_id": [
-    92542,
-    92543
-  ]
+  "transformers_version": "4.36.2"
 }

modeling_intern_vit.py

@@ -1,9 +1,8 @@
 # --------------------------------------------------------
 # InternVL
-# Copyright (c) 2024 OpenGVLab
+# Copyright (c) 2023 OpenGVLab
 # Licensed under The MIT License [see LICENSE for details]
 # --------------------------------------------------------
-
 from typing import Optional, Tuple, Union
 
 import torch
@@ -21,12 +20,18 @@ from transformers.utils import logging
 from .configuration_intern_vit import InternVisionConfig
 
 try:
+    try:  # v1
+        from flash_attn.flash_attn_interface import \
+            flash_attn_unpadded_qkvpacked_func
+    except:  # v2
+        from flash_attn.flash_attn_interface import \
+            flash_attn_varlen_qkvpacked_func as flash_attn_unpadded_qkvpacked_func
+    
     from flash_attn.bert_padding import pad_input, unpad_input
-    from flash_attn.flash_attn_interface import \
-        flash_attn_varlen_qkvpacked_func
+    
     has_flash_attn = True
 except:
-    print('FlashAttention2 is not installed.')
+    print('FlashAttention is not installed.')
     has_flash_attn = False
 
 logger = logging.get_logger(__name__)
@@ -42,12 +47,12 @@ class FlashAttention(nn.Module):
         attention_dropout: The dropout rate to apply to the attention
                            (default: 0.0)
     """
-
+    
     def __init__(self, softmax_scale=None, attention_dropout=0.0, device=None, dtype=None):
         super().__init__()
         self.softmax_scale = softmax_scale
         self.dropout_p = attention_dropout
-
+    
     def forward(self, qkv, key_padding_mask=None, causal=False, cu_seqlens=None,
                 max_s=None, need_weights=False):
         """Implements the multihead softmax attention.
@@ -60,7 +65,7 @@ class FlashAttention(nn.Module):
         assert not need_weights
         assert qkv.dtype in [torch.float16, torch.bfloat16]
         assert qkv.is_cuda
-
+        
         if cu_seqlens is None:
             batch_size = qkv.shape[0]
             seqlen = qkv.shape[1]
@@ -69,7 +74,7 @@ class FlashAttention(nn.Module):
                 max_s = seqlen
                 cu_seqlens = torch.arange(0, (batch_size + 1) * seqlen, step=seqlen, dtype=torch.int32,
                                           device=qkv.device)
-                output = flash_attn_varlen_qkvpacked_func(
+                output = flash_attn_unpadded_qkvpacked_func(
                     qkv, cu_seqlens, max_s, self.dropout_p if self.training else 0.0,
                     softmax_scale=self.softmax_scale, causal=causal
                 )
@@ -79,7 +84,7 @@ class FlashAttention(nn.Module):
                 x = rearrange(qkv, 'b s three h d -> b s (three h d)')
                 x_unpad, indices, cu_seqlens, max_s = unpad_input(x, key_padding_mask)
                 x_unpad = rearrange(x_unpad, 'nnz (three h d) -> nnz three h d', three=3, h=nheads)
-                output_unpad = flash_attn_varlen_qkvpacked_func(
+                output_unpad = flash_attn_unpadded_qkvpacked_func(
                     x_unpad, cu_seqlens, max_s, self.dropout_p if self.training else 0.0,
                     softmax_scale=self.softmax_scale, causal=causal
                 )
@@ -88,11 +93,11 @@ class FlashAttention(nn.Module):
                                    'b s (h d) -> b s h d', h=nheads)
         else:
             assert max_s is not None
-            output = flash_attn_varlen_qkvpacked_func(
+            output = flash_attn_unpadded_qkvpacked_func(
                 qkv, cu_seqlens, max_s, self.dropout_p if self.training else 0.0,
                 softmax_scale=self.softmax_scale, causal=causal
             )
-
+        
         return output, None
 
 
@@ -124,12 +129,6 @@ except Exception:
     pass
 
 
-NORM2FN = {
-    'rms_norm': InternRMSNorm,
-    'layer_norm': nn.LayerNorm,
-}
-
-
 class InternVisionEmbeddings(nn.Module):
     def __init__(self, config: InternVisionConfig):
         super().__init__()
@@ -155,7 +154,7 @@ class InternVisionEmbeddings(nn.Module):
         target_dtype = pos_embed.dtype
         pos_embed = pos_embed.float().reshape(
             1, self.image_size // self.patch_size, self.image_size // self.patch_size, -1).permute(0, 3, 1, 2)
-        pos_embed = F.interpolate(pos_embed, size=(H, W), mode='bicubic', align_corners=False). \
+        pos_embed = F.interpolate(pos_embed, size=(H, W), mode='bicubic', align_corners=False).\
             reshape(1, -1, H * W).permute(0, 2, 1).to(target_dtype)
         return pos_embed
 
@@ -268,12 +267,11 @@ class InternVisionEncoderLayer(nn.Module):
         super().__init__()
         self.embed_dim = config.hidden_size
         self.intermediate_size = config.intermediate_size
-        self.norm_type = config.norm_type
 
         self.attn = InternAttention(config)
         self.mlp = InternMLP(config)
-        self.norm1 = NORM2FN[self.norm_type](self.embed_dim, eps=config.layer_norm_eps)
-        self.norm2 = NORM2FN[self.norm_type](self.embed_dim, eps=config.layer_norm_eps)
+        self.norm1 = InternRMSNorm(self.embed_dim, eps=config.layer_norm_eps)
+        self.norm2 = InternRMSNorm(self.embed_dim, eps=config.layer_norm_eps)
 
         self.ls1 = nn.Parameter(config.initializer_factor * torch.ones(self.embed_dim))
         self.ls2 = nn.Parameter(config.initializer_factor * torch.ones(self.embed_dim))
@@ -288,9 +286,9 @@ class InternVisionEncoderLayer(nn.Module):
         Args:
             hidden_states (`Tuple[torch.FloatTensor, Optional[torch.FloatTensor]]`): input to the layer of shape `(batch, seq_len, embed_dim)`
         """
-        hidden_states = hidden_states + self.drop_path1(self.attn(self.norm1(hidden_states).to(hidden_states.dtype)) * self.ls1)
+        hidden_states = hidden_states + self.drop_path1(self.attn(self.norm1(hidden_states)) * self.ls1)
 
-        hidden_states = hidden_states + self.drop_path2(self.mlp(self.norm2(hidden_states).to(hidden_states.dtype)) * self.ls2)
+        hidden_states = hidden_states + self.drop_path2(self.mlp(self.norm2(hidden_states)) * self.ls2)
 
         return hidden_states
 
@@ -363,7 +361,6 @@ class InternVisionEncoder(nn.Module):
 
 class InternVisionModel(PreTrainedModel):
     main_input_name = 'pixel_values'
-    _supports_flash_attn_2 = True
     config_class = InternVisionConfig
     _no_split_modules = ['InternVisionEncoderLayer']
 

modeling_internlm2.py

@@ -48,18 +48,6 @@ _CONFIG_FOR_DOC = 'InternLM2Config'
 
 flash_attn_func, flash_attn_varlen_func = None, None
 pad_input, index_first_axis, unpad_input = None, None, None
-try:
-    from flash_attn import flash_attn_func as _flash_attn_func
-    from flash_attn import flash_attn_varlen_func as _flash_attn_varlen_func
-    from flash_attn.bert_padding import index_first_axis as _index_first_axis
-    from flash_attn.bert_padding import pad_input as _pad_input
-    from flash_attn.bert_padding import unpad_input as _unpad_input
-
-    flash_attn_func, flash_attn_varlen_func = _flash_attn_func, _flash_attn_varlen_func
-    pad_input, index_first_axis, unpad_input = _pad_input, _index_first_axis, _unpad_input
-    has_flash_attn = True
-except:
-    has_flash_attn = False
 
 
 def _import_flash_attn():
@@ -161,7 +149,7 @@ class InternLM2RotaryEmbedding(nn.Module):
 
     def _set_cos_sin_cache(self, seq_len, device, dtype):
         self.max_seq_len_cached = seq_len
-        t = torch.arange(self.max_seq_len_cached, device=device).to(dtype=self.inv_freq.dtype)
+        t = torch.arange(self.max_seq_len_cached, device=device, dtype=self.inv_freq.dtype)
 
         freqs = torch.einsum('i,j->ij', t, self.inv_freq)
         # Different from paper, but it uses a different permutation in order to obtain the same calculation
@@ -190,7 +178,7 @@ class InternLM2LinearScalingRotaryEmbedding(InternLM2RotaryEmbedding):
 
     def _set_cos_sin_cache(self, seq_len, device, dtype):
         self.max_seq_len_cached = seq_len
-        t = torch.arange(self.max_seq_len_cached, device=device).to(dtype=self.inv_freq.dtype)
+        t = torch.arange(self.max_seq_len_cached, device=device, dtype=self.inv_freq.dtype)
         t = t / self.scaling_factor
 
         freqs = torch.einsum('i,j->ij', t, self.inv_freq)
@@ -220,7 +208,7 @@ class InternLM2DynamicNTKScalingRotaryEmbedding(InternLM2RotaryEmbedding):
             inv_freq = 1.0 / (base ** (torch.arange(0, self.dim, 2).float().to(device) / self.dim))
             self.register_buffer('inv_freq', inv_freq, persistent=False)
 
-        t = torch.arange(self.max_seq_len_cached, device=device).to(dtype=self.inv_freq.dtype)
+        t = torch.arange(self.max_seq_len_cached, device=device, dtype=self.inv_freq.dtype)
 
         freqs = torch.einsum('i,j->ij', t, self.inv_freq)
         # Different from paper, but it uses a different permutation in order to obtain the same calculation
@@ -709,7 +697,6 @@ class InternLM2PreTrainedModel(PreTrainedModel):
     supports_gradient_checkpointing = True
     _no_split_modules = ['InternLM2DecoderLayer']
     _skip_keys_device_placement = 'past_key_values'
-    _supports_flash_attn_2 = True
 
     def _init_weights(self, module):
         std = self.config.initializer_range
@@ -808,9 +795,6 @@ class InternLM2Model(InternLM2PreTrainedModel):
         self.padding_idx = config.pad_token_id
         self.vocab_size = config.vocab_size
         self.config = config
-        if not has_flash_attn:
-            self.config.attn_implementation = 'eager'
-            print('Warning: Flash attention is not available, using eager attention instead.')
 
         self.tok_embeddings = nn.Embedding(config.vocab_size, config.hidden_size, self.padding_idx)
 
@@ -1098,16 +1082,13 @@ class InternLM2ForCausalLM(InternLM2PreTrainedModel):
             output = (logits,) + outputs[1:]
             return (loss,) + output if loss is not None else output
 
-        device = input_ids.device if input_ids is not None else inputs_embeds.device
-        output = CausalLMOutputWithPast(
+        return CausalLMOutputWithPast(
             loss=loss,
             logits=logits,
             past_key_values=outputs.past_key_values,
             hidden_states=outputs.hidden_states,
             attentions=outputs.attentions,
         )
-        output['logits'] = output['logits'].to(device)
-        return output
 
     def prepare_inputs_for_generation(
         self, input_ids, past_key_values=None, attention_mask=None, inputs_embeds=None, **kwargs

modeling_internvl_chat.py

@@ -1,48 +1,70 @@
 # --------------------------------------------------------
 # InternVL
-# Copyright (c) 2024 OpenGVLab
+# Copyright (c) 2023 OpenGVLab
 # Licensed under The MIT License [see LICENSE for details]
 # --------------------------------------------------------
-
 import warnings
 from typing import Any, List, Optional, Tuple, Union
 
 import torch.utils.checkpoint
-import transformers
+from peft import LoraConfig, get_peft_model
 from torch import nn
 from torch.nn import CrossEntropyLoss
-from transformers import AutoModel, GenerationConfig, LlamaForCausalLM
+from transformers import (AutoModel, GenerationConfig, LlamaForCausalLM,
+                          LlamaTokenizer)
 from transformers.modeling_outputs import CausalLMOutputWithPast
 from transformers.modeling_utils import PreTrainedModel
 from transformers.utils import ModelOutput, logging
 
 from .configuration_internvl_chat import InternVLChatConfig
-from .conversation import get_conv_template
-from .modeling_intern_vit import InternVisionModel, has_flash_attn
+from .modeling_intern_vit import InternVisionModel
 from .modeling_internlm2 import InternLM2ForCausalLM
 
 logger = logging.get_logger(__name__)
 
 
-def version_cmp(v1, v2, op='eq'):
-    import operator
+def window_partition(x, window_size):
+    """
+    Args:
+        x: (B, C, H, W)
+        window_size (int): window size, assuming square window
+
+    Returns:
+        windows: (num_windows*B, C, window_size, window_size)
+    """
+    B, C, H, W = x.shape
+    assert H % window_size == 0 and W % window_size == 0, 'H and W must be divisible by window_size'
+
+    x = x.view(B, C, H // window_size, window_size, W // window_size, window_size)
+    windows = x.permute(0, 2, 4, 1, 3, 5).contiguous().view(-1, C, window_size, window_size)
+    return windows
+
+
+def window_reverse(windows, window_size, H, W):
+    """
+    Args:
+        windows: (num_windows*B, window_size, window_size, C)
+        window_size (int): Window size
+        H (int): Height of image
+        W (int): Width of image
 
-    from packaging import version
-    op_func = getattr(operator, op)
-    return op_func(version.parse(v1), version.parse(v2))
+    Returns:
+        x: (B, H * W, C)
+    """
+    B = int(windows.shape[0] / (H * W / window_size / window_size))
+    x = windows.view(B, H // window_size, W // window_size, window_size, window_size, -1)
+    x = x.permute(0, 1, 3, 2, 4, 5).contiguous().view(B, H * W, -1)
+    return x
 
 
 class InternVLChatModel(PreTrainedModel):
     config_class = InternVLChatConfig
     main_input_name = 'pixel_values'
-    base_model_prefix = 'language_model'
-    _supports_flash_attn_2 = True
-    _no_split_modules = ['InternVisionModel', 'LlamaDecoderLayer', 'InternLM2DecoderLayer']
+    _no_split_modules = ['InternVisionEncoderLayer', 'LlamaDecoderLayer', 'LlamaForCausalLM']
 
-    def __init__(self, config: InternVLChatConfig, vision_model=None, language_model=None, use_flash_attn=True):
+    def __init__(self, config: InternVLChatConfig, vision_model=None, language_model=None):
         super().__init__(config)
 
-        assert version_cmp(transformers.__version__, '4.37.0', 'ge')
         image_size = config.force_image_size or config.vision_config.image_size
         patch_size = config.vision_config.patch_size
         self.patch_size = patch_size
@@ -50,10 +72,8 @@ class InternVLChatModel(PreTrainedModel):
         self.template = config.template
         self.num_image_token = int((image_size // patch_size) ** 2 * (config.downsample_ratio ** 2))
         self.downsample_ratio = config.downsample_ratio
+        self.image_fold = config.image_fold
         self.ps_version = config.ps_version
-        use_flash_attn = use_flash_attn if has_flash_attn else False
-        config.vision_config.use_flash_attn = True if use_flash_attn else False
-        config.llm_config.attn_implementation = 'flash_attention_2' if use_flash_attn else 'eager'
 
         logger.info(f'num_image_token: {self.num_image_token}')
         logger.info(f'ps_version: {self.ps_version}')
@@ -81,9 +101,44 @@ class InternVLChatModel(PreTrainedModel):
             nn.Linear(llm_hidden_size, llm_hidden_size)
         )
 
+        # if config.force_image_size != config.vision_config.image_size:
+        #     self.vision_model.resize_pos_embeddings(
+        #         old_size=config.vision_config.image_size,
+        #         new_size=config.force_image_size,
+        #         patch_size=config.vision_config.patch_size
+        #     )
+
         self.img_context_token_id = None
-        self.conv_template = get_conv_template(self.template)
-        self.system_message = self.conv_template.system_message
+        self.neftune_alpha = None
+
+        if config.use_backbone_lora:
+            self.wrap_backbone_lora(r=config.use_backbone_lora, lora_alpha=2 * config.use_backbone_lora)
+
+        if config.use_llm_lora:
+            self.wrap_llm_lora(r=config.use_llm_lora, lora_alpha=2 * config.use_llm_lora)
+
+    def wrap_backbone_lora(self, r=128, lora_alpha=256, lora_dropout=0.05):
+        lora_config = LoraConfig(
+            r=r,
+            target_modules=['attn.qkv', 'attn.proj', 'mlp.fc1', 'mlp.fc2'],
+            lora_alpha=lora_alpha,
+            lora_dropout=lora_dropout,
+        )
+        self.vision_model = get_peft_model(self.vision_model, lora_config)
+        self.vision_model.print_trainable_parameters()
+
+    def wrap_llm_lora(self, r=128, lora_alpha=256, lora_dropout=0.05):
+        lora_config = LoraConfig(
+            r=r,
+            target_modules=['self_attn.q_proj', 'self_attn.k_proj', 'self_attn.v_proj', 'self_attn.o_proj',
+                            'mlp.gate_proj', 'mlp.down_proj', 'mlp.up_proj'],
+            lora_alpha=lora_alpha,
+            lora_dropout=lora_dropout,
+            task_type='CAUSAL_LM'
+        )
+        self.language_model = get_peft_model(self.language_model, lora_config)
+        self.language_model.enable_input_require_grads()
+        self.language_model.print_trainable_parameters()
 
     def forward(
             self,
@@ -102,7 +157,7 @@ class InternVLChatModel(PreTrainedModel):
         return_dict = return_dict if return_dict is not None else self.config.use_return_dict
 
         image_flags = image_flags.squeeze(-1)
-        input_embeds = self.language_model.get_input_embeddings()(input_ids).clone()
+        input_embeds = self.language_model.get_input_embeddings()(input_ids)
 
         vit_embeds = self.extract_feature(pixel_values)
         vit_embeds = vit_embeds[image_flags == 1]
@@ -111,7 +166,7 @@ class InternVLChatModel(PreTrainedModel):
         B, N, C = input_embeds.shape
         input_embeds = input_embeds.reshape(B * N, C)
 
-        if torch.distributed.is_initialized() and torch.distributed.get_rank() == 0:
+        if torch.distributed.get_rank() == 0:
             print(f'dynamic ViT batch size: {vit_batch_size}, images per sample: {vit_batch_size / B}, dynamic token length: {N}')
 
         input_ids = input_ids.reshape(B * N)
@@ -180,7 +235,17 @@ class InternVLChatModel(PreTrainedModel):
             x = x.permute(0, 2, 1, 3).contiguous()
         return x
 
+    def noised_embed(self, vit_embeds, noise_alpha=5):
+        dims = torch.tensor(vit_embeds.size(1) * vit_embeds.size(2))
+        mag_norm = noise_alpha / torch.sqrt(dims)
+        noise = torch.zeros_like(vit_embeds).uniform_(-mag_norm, mag_norm)
+        return vit_embeds + noise
+
     def extract_feature(self, pixel_values):
+        if self.image_fold:
+            image_size = pixel_values.size(-1)  # B, C, H, W
+            pixel_values = window_partition(pixel_values, window_size=image_size // self.image_fold)  # 4B, C, H/2, W/2
+
         if self.select_layer == -1:
             vit_embeds = self.vision_model(
                 pixel_values=pixel_values,
@@ -193,99 +258,53 @@ class InternVLChatModel(PreTrainedModel):
                 return_dict=True).hidden_states[self.select_layer]
         vit_embeds = vit_embeds[:, 1:, :]
 
+        if self.training and self.neftune_alpha is not None:
+            vit_embeds = self.noised_embed(vit_embeds, self.neftune_alpha)
+
+        if self.image_fold:
+            vit_embeds = window_reverse(vit_embeds, window_size=image_size // (self.image_fold * self.patch_size),
+                                        H=image_size // self.patch_size, W=image_size // self.patch_size)
+
+        # if torch.distributed.get_rank() == 0:
+        #     print("before pixel shuffle:", vit_embeds.shape)
         h = w = int(vit_embeds.shape[1] ** 0.5)
         vit_embeds = vit_embeds.reshape(vit_embeds.shape[0], h, w, -1)
         vit_embeds = self.pixel_shuffle(vit_embeds, scale_factor=self.downsample_ratio)
         vit_embeds = vit_embeds.reshape(vit_embeds.shape[0], -1, vit_embeds.shape[-1])
+        # if torch.distributed.get_rank() == 0:
+        #     print("after pixel shuffle:", vit_embeds.shape)
         vit_embeds = self.mlp1(vit_embeds)
         return vit_embeds
 
-    def batch_chat(self, tokenizer, pixel_values, questions, generation_config, num_patches_list=None,
-                   history=None, return_history=False, IMG_START_TOKEN='<img>', IMG_END_TOKEN='</img>',
-                   IMG_CONTEXT_TOKEN='<IMG_CONTEXT>', verbose=False, image_counts=None):
-        if history is not None or return_history:
-            print('Now multi-turn chat is not supported in batch_chat.')
-            raise NotImplementedError
-
-        if image_counts is not None:
-            num_patches_list = image_counts
-            print('Warning: `image_counts` is deprecated. Please use `num_patches_list` instead.')
-
-        img_context_token_id = tokenizer.convert_tokens_to_ids(IMG_CONTEXT_TOKEN)
-        self.img_context_token_id = img_context_token_id
-
-        if verbose and pixel_values is not None:
-            image_bs = pixel_values.shape[0]
-            print(f'dynamic ViT batch size: {image_bs}')
-
-        queries = []
-        for idx, num_patches in enumerate(num_patches_list):
-            question = questions[idx]
-            if pixel_values is not None and '<image>' not in question:
-                question = '<image>\n' + question
-            template = get_conv_template(self.template)
-            template.system_message = self.system_message
-            template.append_message(template.roles[0], question)
-            template.append_message(template.roles[1], None)
-            query = template.get_prompt()
-
-            image_tokens = IMG_START_TOKEN + IMG_CONTEXT_TOKEN * self.num_image_token * num_patches + IMG_END_TOKEN
-            query = query.replace('<image>', image_tokens, 1)
-            queries.append(query)
-
-        tokenizer.padding_side = 'left'
-        model_inputs = tokenizer(queries, return_tensors='pt', padding=True)
-        input_ids = model_inputs['input_ids'].to(self.device)
-        attention_mask = model_inputs['attention_mask'].to(self.device)
-        eos_token_id = tokenizer.convert_tokens_to_ids(template.sep.strip())
-        generation_config['eos_token_id'] = eos_token_id
-        generation_output = self.generate(
-            pixel_values=pixel_values,
-            input_ids=input_ids,
-            attention_mask=attention_mask,
-            **generation_config
-        )
-        responses = tokenizer.batch_decode(generation_output, skip_special_tokens=True)
-        responses = [response.split(template.sep.strip())[0].strip() for response in responses]
-        return responses
-
     def chat(self, tokenizer, pixel_values, question, generation_config, history=None, return_history=False,
-             num_patches_list=None, IMG_START_TOKEN='<img>', IMG_END_TOKEN='</img>', IMG_CONTEXT_TOKEN='<IMG_CONTEXT>',
-             verbose=False):
-
-        if history is None and pixel_values is not None and '<image>' not in question:
-            question = '<image>\n' + question
-
-        if num_patches_list is None:
-            num_patches_list = [pixel_values.shape[0]] if pixel_values is not None else []
-        assert pixel_values is None or len(pixel_values) == sum(num_patches_list)
+             IMG_START_TOKEN='<img>', IMG_END_TOKEN='</img>', IMG_CONTEXT_TOKEN='<IMG_CONTEXT>'):
 
         img_context_token_id = tokenizer.convert_tokens_to_ids(IMG_CONTEXT_TOKEN)
         self.img_context_token_id = img_context_token_id
+        if tokenizer.convert_tokens_to_ids('<|im_end|>') != 0:
+            eos_token_id = tokenizer.convert_tokens_to_ids('<|im_end|>')  # 92542, InternLM2
+        else:
+            eos_token_id = tokenizer.eos_token_id
 
-        template = get_conv_template(self.template)
-        template.system_message = self.system_message
-        eos_token_id = tokenizer.convert_tokens_to_ids(template.sep.strip())
+        from .conversation import get_conv_template
 
-        history = [] if history is None else history
-        for (old_question, old_answer) in history:
-            template.append_message(template.roles[0], old_question)
-            template.append_message(template.roles[1], old_answer)
+        template = get_conv_template(self.template)
+        image_bs = pixel_values.shape[0]
+        print(f'dynamic ViT batch size: {image_bs}')
+        if history is None:
+            history = []
+            image_tokens = IMG_START_TOKEN + IMG_CONTEXT_TOKEN * self.num_image_token * image_bs + IMG_END_TOKEN
+            question = image_tokens + '\n' + question
+        else:
+            for (old_question, old_answer) in history:
+                template.append_message(template.roles[0], old_question)
+                template.append_message(template.roles[1], old_answer)
         template.append_message(template.roles[0], question)
         template.append_message(template.roles[1], None)
         query = template.get_prompt()
-
-        if verbose and pixel_values is not None:
-            image_bs = pixel_values.shape[0]
-            print(f'dynamic ViT batch size: {image_bs}')
-
-        for num_patches in num_patches_list:
-            image_tokens = IMG_START_TOKEN + IMG_CONTEXT_TOKEN * self.num_image_token * num_patches + IMG_END_TOKEN
-            query = query.replace('<image>', image_tokens, 1)
-
         model_inputs = tokenizer(query, return_tensors='pt')
-        input_ids = model_inputs['input_ids'].to(self.device)
-        attention_mask = model_inputs['attention_mask'].to(self.device)
+        input_ids = model_inputs['input_ids'].cuda()
+        attention_mask = model_inputs['attention_mask'].cuda()
         generation_config['eos_token_id'] = eos_token_id
         generation_output = self.generate(
             pixel_values=pixel_values,
@@ -294,16 +313,15 @@ class InternVLChatModel(PreTrainedModel):
             **generation_config
         )
         response = tokenizer.batch_decode(generation_output, skip_special_tokens=True)[0]
-        response = response.split(template.sep.strip())[0].strip()
+        response = response.split('<|im_end|>')[0].strip()  # for InternLM2
         history.append((question, response))
         if return_history:
             return response, history
         else:
-            query_to_print = query.replace(IMG_CONTEXT_TOKEN, '')
-            query_to_print = query_to_print.replace(f'{IMG_START_TOKEN}{IMG_END_TOKEN}', '<image>')
-            if verbose:
-                print(query_to_print, response)
+            query_to_print = query.replace(image_tokens, '<image>')
+            print(query_to_print, response)
             return response
+        return response
 
     @torch.no_grad()
     def generate(
@@ -314,6 +332,7 @@ class InternVLChatModel(PreTrainedModel):
             visual_features: Optional[torch.FloatTensor] = None,
             generation_config: Optional[GenerationConfig] = None,
             output_hidden_states: Optional[bool] = None,
+            return_dict: Optional[bool] = None,
             **generate_kwargs,
     ) -> torch.LongTensor:
 
@@ -323,6 +342,7 @@ class InternVLChatModel(PreTrainedModel):
                 vit_embeds = visual_features
             else:
                 vit_embeds = self.extract_feature(pixel_values)
+
             input_embeds = self.language_model.get_input_embeddings()(input_ids)
             B, N, C = input_embeds.shape
             input_embeds = input_embeds.reshape(B * N, C)
@@ -330,7 +350,7 @@ class InternVLChatModel(PreTrainedModel):
             input_ids = input_ids.reshape(B * N)
             selected = (input_ids == self.img_context_token_id)
             assert selected.sum() != 0
-            input_embeds[selected] = vit_embeds.reshape(-1, C).to(input_embeds.device)
+            input_embeds[selected] = vit_embeds.reshape(-1, C)
 
             input_embeds = input_embeds.reshape(B, N, C)
         else:
@@ -341,6 +361,7 @@ class InternVLChatModel(PreTrainedModel):
             attention_mask=attention_mask,
             generation_config=generation_config,
             output_hidden_states=output_hidden_states,
+            return_dict=return_dict,
             use_cache=True,
             **generate_kwargs,
         )

preprocessor_config.json

@@ -1,19 +0,0 @@
-{
-  "crop_size": 448,
-  "do_center_crop": true,
-  "do_normalize": true,
-  "do_resize": true,
-  "feature_extractor_type": "CLIPFeatureExtractor",
-  "image_mean": [
-    0.485,
-    0.456,
-    0.406
-  ],
-  "image_std": [
-    0.229,
-    0.224,
-    0.225
-  ],
-  "resample": 3,
-  "size": 448
-}

examples/red-panda.mp4 → runs/Apr15_16-44-40_SH-IDC1-10-140-37-13/events.out.tfevents.1713171220.SH-IDC1-10-140-37-13.204150.0

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d921c07bb97224d65a37801541d246067f0d506f08723ffa1ad85c217907ccb8
-size 1867237
+oid sha256:294d5bf755e6dea5c005c57af52e958a38bb42a7d17d801a25a6543bfe6ddca2
+size 16662

runs/Apr15_17-33-22_SH-IDC1-10-140-37-13/events.out.tfevents.1713174123.SH-IDC1-10-140-37-13.259480.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:57d61c0e776bfb521e58febdbd99525e011f82137ceaaa655ffa6e2b3a9b02a9
+size 72471