本地部署DeepSeek开源多模态大模型Janus-Pro-7B实操

本地部署DeepSeek开源多模态大模型Janus-Pro-7B实操

Janus-Pro-7B介绍

Janus-Pro-7B 是由 DeepSeek 开发的多模态 AI 模型，它在理解和生成方面取得了显著的进步。这意味着它不仅可以处理文本，还可以处理图像等其他模态的信息。
模型主要特点:Permalink
统一的架构: Janus-Pro 采用单一 transformer 架构来处理文本和图像信息，实现了真正的多模态理解和生成。
解耦的视觉编码: 为了更好地平衡理解和生成任务，Janus-Pro 将视觉编码解耦为独立的路径，提高了模型的灵活性和性能。
强大的性能: 在多个基准测试中，Janus-Pro 的性能超越了之前的统一模型，甚至可以与特定任务的模型相媲美。
开源: Janus-Pro-7B 是开源的，这意味着研究人员和开发者可以自由地访问和使用它，推动 AI 领域的创新。
具体来说，Janus-Pro-7B 有以下优势:

图像理解: 能够准确地识别和理解图像中的对象、场景和关系。
图像生成: 可以根据文本描述生成高质量的图像，甚至可以进行图像编辑和转换。
文本生成: 可以生成流畅、连贯的文本，例如故事、诗歌、代码等。
多模态推理: 可以结合文本和图像信息进行推理，例如根据图像内容回答问题，或者根据文本描述生成图像。
与其他模型的比较:
超越 DALL-E 3 和 Stable Diffusion: 在 GenEval 和 DPG-Bench 等基准测试中，Janus-Pro-7B 的性能优于 OpenAI 的 DALL-E 3 和 Stability AI 的 Stable Diffusion。
基于 DeepSeek-LLM: Janus-Pro 建立在 DeepSeek-LLM-1.5b-base/DeepSeek-LLM-7b-base 的基础上，并对其进行了多模态扩展。
应用场景:
Janus-Pro-7B 具有广泛的应用场景，例如：

内容创作: 可以帮助用户生成高质量的图像、文本和其他多媒体内容。
教育: 可以用于创建交互式学习体验，例如根据文本描述生成图像，或者根据图像内容回答问题。
客户服务: 可以用于构建更智能的聊天机器人，能够理解和回应用户的多模态查询。
辅助设计: 可以帮助设计师生成创意概念，并将其转化为可视化原型

1 启动Anaconda环境

2 进入命令环境

conda create -n myenv python=3.10 -ygit clone https://github.com/deepseek-ai/Janus.gitcd Januspip install -e .pip install webencodings beautifulsoup4 tinycss2pip install -e .[gradio]pip install 'pexpect>4.3'python demo/app_januspro.py

3 遇到默认配置下C盘磁盘空间不足问题

(myenvp) C:\Users\Administrator>python demo/app_januspro.py
python: can't open file 'C:\\Users\\Administrator\\demo\\app_januspro.py': [Errno 2] No such file or directory(myenvp) C:\Users\Administrator>e:(myenvp) E:\>cd ai(myenvp) E:\AI>cd Janus(myenvp) E:\AI\Janus>dir驱动器 E 中的卷是 chia-12T-1卷的序列号是 0AF0-159BE:\AI\Janus 的目录2025/01/31  12:26    <DIR>          .
2025/01/30  00:53    <DIR>          ..
2025/01/30  00:53               115 .gitattributes
2025/01/30  00:53             7,301 .gitignore
2025/01/30  01:47    <DIR>          .gradio
2025/01/30  01:18    <DIR>          .locks
2025/01/31  12:26                 0 4.3'
2025/01/30  00:53    <DIR>          demo
2025/01/30  00:53             4,515 generation_inference.py
2025/01/30  00:53    <DIR>          images
2025/01/30  00:53             2,642 inference.py
2025/01/30  00:53             5,188 interactivechat.py
2025/01/30  01:04    <DIR>          janus
2025/01/31  12:25    <DIR>          janus.egg-info
2025/01/30  00:53         2,846,268 janus_pro_tech_report.pdf
2025/01/30  00:53             1,065 LICENSE-CODE
2025/01/30  00:53            13,718 LICENSE-MODEL
2025/01/30  00:53             3,069 Makefile
2025/01/30  01:47    <DIR>          models--deepseek-ai--Janus-Pro-7B
2025/01/30  00:53             1,111 pyproject.toml
2025/01/30  00:53            26,742 README.md
2025/01/30  00:53               278 requirements.txt
2025/01/30  01:18                 1 version.txt14 个文件      2,912,013 字节9 个目录 9,387,683,614,720 可用字节

3.1 设置HF_DATASETS_CACHE环境变量没解决问题

(myenvp) E:\AI\Janus>set HF_DATASETS_CACHE="E:\AI\Janus"(myenvp) E:\AI\Janus>python demo/app_januspro.py
Python version is above 3.10, patching the collections module.
D:\anaconda3\envs\myenvp\lib\site-packages\transformers\models\auto\image_processing_auto.py:590: FutureWarning: The image_processor_class argument is deprecated and will be removed in v4.42. Please use `slow_image_processor_class`, or `fast_image_processor_class` insteadwarnings.warn(
Downloading shards:   0%|                                                                        | 0/2 [00:00<?, ?it/s]D:\anaconda3\envs\myenvp\lib\site-packages\huggingface_hub\file_download.py:651: UserWarning: Not enough free disk space to download the file. The expected file size is: 9988.18 MB. The target location C:\Users\Administrator\.cache\huggingface\hub\models--deepseek-ai--Janus-Pro-7B\blobs only has 8154.37 MB free disk space.warnings.warn(
pytorch_model-00001-of-00002.bin:  37%|███████████████▉                           | 3.71G/9.99G [00:05<02:38, 39.5MB/s]
Downloading shards:   0%|                                                                        | 0/2 [00:06<?, ?it/s]
Traceback (most recent call last):File "E:\AI\Janus\demo\app_januspro.py", line 19, in <module>vl_gpt = AutoModelForCausalLM.from_pretrained(model_path,File "D:\anaconda3\envs\myenvp\lib\site-packages\transformers\models\auto\auto_factory.py", line 564, in from_pretrainedreturn model_class.from_pretrained(File "D:\anaconda3\envs\myenvp\lib\site-packages\transformers\modeling_utils.py", line 3944, in from_pretrainedresolved_archive_file, sharded_metadata = get_checkpoint_shard_files(File "D:\anaconda3\envs\myenvp\lib\site-packages\transformers\utils\hub.py", line 1098, in get_checkpoint_shard_filescached_filename = cached_file(File "D:\anaconda3\envs\myenvp\lib\site-packages\transformers\utils\hub.py", line 403, in cached_fileresolved_file = hf_hub_download(File "D:\anaconda3\envs\myenvp\lib\site-packages\huggingface_hub\utils\_validators.py", line 114, in _inner_fnreturn fn(*args, **kwargs)File "D:\anaconda3\envs\myenvp\lib\site-packages\huggingface_hub\file_download.py", line 860, in hf_hub_downloadreturn _hf_hub_download_to_cache_dir(File "D:\anaconda3\envs\myenvp\lib\site-packages\huggingface_hub\file_download.py", line 1009, in _hf_hub_download_to_cache_dir_download_to_tmp_and_move(File "D:\anaconda3\envs\myenvp\lib\site-packages\huggingface_hub\file_download.py", line 1543, in _download_to_tmp_and_movehttp_get(File "D:\anaconda3\envs\myenvp\lib\site-packages\huggingface_hub\file_download.py", line 452, in http_getfor chunk in r.iter_content(chunk_size=constants.DOWNLOAD_CHUNK_SIZE):File "D:\anaconda3\envs\myenvp\lib\site-packages\requests\models.py", line 820, in generateyield from self.raw.stream(chunk_size, decode_content=True)File "D:\anaconda3\envs\myenvp\lib\site-packages\urllib3\response.py", line 1066, in streamdata = self.read(amt=amt, decode_content=decode_content)File "D:\anaconda3\envs\myenvp\lib\site-packages\urllib3\response.py", line 955, in readdata = self._raw_read(amt)File "D:\anaconda3\envs\myenvp\lib\site-packages\urllib3\response.py", line 879, in _raw_readdata = self._fp_read(amt, read1=read1) if not fp_closed else b""File "D:\anaconda3\envs\myenvp\lib\site-packages\urllib3\response.py", line 862, in _fp_readreturn self._fp.read(amt) if amt is not None else self._fp.read()File "D:\anaconda3\envs\myenvp\lib\http\client.py", line 466, in reads = self.fp.read(amt)File "D:\anaconda3\envs\myenvp\lib\socket.py", line 717, in readintoreturn self._sock.recv_into(b)File "D:\anaconda3\envs\myenvp\lib\ssl.py", line 1307, in recv_intoreturn self.read(nbytes, buffer)File "D:\anaconda3\envs\myenvp\lib\ssl.py", line 1163, in readreturn self._sslobj.read(len, buffer)
KeyboardInterrupt
^C

3.2 设置环境变量HF_HOME解决问题

(myenvp) E:\AI\Janus>set HF_HOME=E:\AI\Janus(myenvp) E:\AI\Janus>python demo/app_januspro.py
Python version is above 3.10, patching the collections module.
D:\anaconda3\envs\myenvp\lib\site-packages\transformers\models\auto\image_processing_auto.py:590: FutureWarning: The image_processor_class argument is deprecated and will be removed in v4.42. Please use `slow_image_processor_class`, or `fast_image_processor_class` insteadwarnings.warn(
config.json: 100%|████████████████████████████████████████████████████████████████████████| 1.28k/1.28k [00:00<?, ?B/s]
pytorch_model.bin.index.json: 100%|███████████████████████████████████████████████| 89.0k/89.0k [00:00<00:00, 1.67MB/s]
model.safetensors.index.json: 100%|███████████████████████████████████████████████| 92.8k/92.8k [00:00<00:00, 2.99MB/s]
pytorch_model-00001-of-00002.bin:  15%|██████▌                                    | 1.53G/9.99G [00:37<03:26, 41.0MB/s]
Downloading shards:   0%|                                                                        | 0/2 [00:37<?, ?it/s]
Traceback (most recent call last):File "E:\AI\Janus\demo\app_januspro.py", line 19, in <module>vl_gpt = AutoModelForCausalLM.from_pretrained(model_path,File "D:\anaconda3\envs\myenvp\lib\site-packages\transformers\models\auto\auto_factory.py", line 564, in from_pretrainedreturn model_class.from_pretrained(File "D:\anaconda3\envs\myenvp\lib\site-packages\transformers\modeling_utils.py", line 3944, in from_pretrainedresolved_archive_file, sharded_metadata = get_checkpoint_shard_files(File "D:\anaconda3\envs\myenvp\lib\site-packages\transformers\utils\hub.py", line 1098, in get_checkpoint_shard_filescached_filename = cached_file(File "D:\anaconda3\envs\myenvp\lib\site-packages\transformers\utils\hub.py", line 403, in cached_fileresolved_file = hf_hub_download(File "D:\anaconda3\envs\myenvp\lib\site-packages\huggingface_hub\utils\_validators.py", line 114, in _inner_fnreturn fn(*args, **kwargs)File "D:\anaconda3\envs\myenvp\lib\site-packages\huggingface_hub\file_download.py", line 860, in hf_hub_downloadreturn _hf_hub_download_to_cache_dir(File "D:\anaconda3\envs\myenvp\lib\site-packages\huggingface_hub\file_download.py", line 1009, in _hf_hub_download_to_cache_dir_download_to_tmp_and_move(File "D:\anaconda3\envs\myenvp\lib\site-packages\huggingface_hub\file_download.py", line 1543, in _download_to_tmp_and_movehttp_get(File "D:\anaconda3\envs\myenvp\lib\site-packages\huggingface_hub\file_download.py", line 452, in http_getfor chunk in r.iter_content(chunk_size=constants.DOWNLOAD_CHUNK_SIZE):File "D:\anaconda3\envs\myenvp\lib\site-packages\requests\models.py", line 820, in generateyield from self.raw.stream(chunk_size, decode_content=True)File "D:\anaconda3\envs\myenvp\lib\site-packages\urllib3\response.py", line 1066, in streamdata = self.read(amt=amt, decode_content=decode_content)File "D:\anaconda3\envs\myenvp\lib\site-packages\urllib3\response.py", line 955, in readdata = self._raw_read(amt)File "D:\anaconda3\envs\myenvp\lib\site-packages\urllib3\response.py", line 879, in _raw_readdata = self._fp_read(amt, read1=read1) if not fp_closed else b""File "D:\anaconda3\envs\myenvp\lib\site-packages\urllib3\response.py", line 862, in _fp_readreturn self._fp.read(amt) if amt is not None else self._fp.read()File "D:\anaconda3\envs\myenvp\lib\http\client.py", line 466, in reads = self.fp.read(amt)File "D:\anaconda3\envs\myenvp\lib\socket.py", line 717, in readintoreturn self._sock.recv_into(b)File "D:\anaconda3\envs\myenvp\lib\ssl.py", line 1307, in recv_intoreturn self.read(nbytes, buffer)File "D:\anaconda3\envs\myenvp\lib\ssl.py", line 1163, in readreturn self._sslobj.read(len, buffer)
KeyboardInterrupt
^C

3.3 如果没下载好模型文件忽略这步

如果之前已经下载好模型文件，将models–deepseek-ai–Janus-Pro-7B目录拷贝到E:\AI\Janus\hub

(myenvp) E:\AI\Janus>python demo/app_januspro.py
Python version is above 3.10, patching the collections module.
D:\anaconda3\envs\myenvp\lib\site-packages\transformers\models\auto\image_processing_auto.py:590: FutureWarning: The image_processor_class argument is deprecated and will be removed in v4.42. Please use `slow_image_processor_class`, or `fast_image_processor_class` insteadwarnings.warn(
Loading checkpoint shards: 100%|█████████████████████████████████████████████████████████| 2/2 [00:44<00:00, 22.13s/it]
Using a slow image processor as `use_fast` is unset and a slow processor was saved with this model. `use_fast=True` will be the default behavior in v4.48, even if the model was saved with a slow processor. This will result in minor differences in outputs. You'll still be able to use a slow processor with `use_fast=False`.
You are using the default legacy behaviour of the <class 'transformers.models.llama.tokenization_llama_fast.LlamaTokenizerFast'>. This is expected, and simply means that the `legacy` (previous) behavior will be used so nothing changes for you. If you want to use the new behaviour, set `legacy=False`. This should only be set if you understand what it means, and thoroughly read the reason why this was added as explained in https://github.com/huggingface/transformers/pull/24565 - if you loaded a llama tokenizer from a GGUF file you can ignore this message.
Some kwargs in processor config are unused and will not have any effect: ignore_id, num_image_tokens, add_special_token, mask_prompt, image_tag, sft_format.
Running on local URL:  http://127.0.0.1:7860
IMPORTANT: You are using gradio version 3.48.0, however version 4.44.1 is available, please upgrade.
--------
Running on public URL: https://cf6180260c7448cc2b.gradio.liveThis share link expires in 72 hours. For free permanent hosting and GPU upgrades, run `gradio deploy` from Terminal to deploy to Spaces (https://huggingface.co/spaces)
Keyboard interruption in main thread... closing server.
Traceback (most recent call last):File "D:\anaconda3\envs\myenvp\lib\site-packages\gradio\blocks.py", line 2361, in block_threadtime.sleep(0.1)
KeyboardInterruptDuring handling of the above exception, another exception occurred:Traceback (most recent call last):File "E:\AI\Janus\demo\app_januspro.py", line 244, in <module>demo.launch(share=True)File "D:\anaconda3\envs\myenvp\lib\site-packages\gradio\blocks.py", line 2266, in launchself.block_thread()File "D:\anaconda3\envs\myenvp\lib\site-packages\gradio\blocks.py", line 2365, in block_threadself.server.close()File "D:\anaconda3\envs\myenvp\lib\site-packages\gradio\networking.py", line 75, in closeself.thread.join()File "D:\anaconda3\envs\myenvp\lib\threading.py", line 1096, in joinself._wait_for_tstate_lock()File "D:\anaconda3\envs\myenvp\lib\threading.py", line 1116, in _wait_for_tstate_lockif lock.acquire(block, timeout):
KeyboardInterrupt
Killing tunnel 127.0.0.1:7860 <> https://cf6180260c7448cc2b.gradio.live
^C

4 强制使用显卡

(myenvp) E:\AI\Janus>python demo/app_januspro.py --device cuda
Python version is above 3.10, patching the collections module.
D:\anaconda3\envs\myenvp\lib\site-packages\transformers\models\auto\image_processing_auto.py:590: FutureWarning: The image_processor_class argument is deprecated and will be removed in v4.42. Please use `slow_image_processor_class`, or `fast_image_processor_class` insteadwarnings.warn(
Loading checkpoint shards: 100%|█████████████████████████████████████████████████████████| 2/2 [00:06<00:00,  3.29s/it]
Using a slow image processor as `use_fast` is unset and a slow processor was saved with this model. `use_fast=True` will be the default behavior in v4.48, even if the model was saved with a slow processor. This will result in minor differences in outputs. You'll still be able to use a slow processor with `use_fast=False`.
You are using the default legacy behaviour of the <class 'transformers.models.llama.tokenization_llama_fast.LlamaTokenizerFast'>. This is expected, and simply means that the `legacy` (previous) behavior will be used so nothing changes for you. If you want to use the new behaviour, set `legacy=False`. This should only be set if you understand what it means, and thoroughly read the reason why this was added as explained in https://github.com/huggingface/transformers/pull/24565 - if you loaded a llama tokenizer from a GGUF file you can ignore this message.
Some kwargs in processor config are unused and will not have any effect: num_image_tokens, image_tag, ignore_id, mask_prompt, sft_format, add_special_token.
Running on local URL:  http://127.0.0.1:7860
IMPORTANT: You are using gradio version 3.48.0, however version 4.44.1 is available, please upgrade.
--------
Running on public URL: https://342ecb20d5120e7d8c.gradio.liveThis share link expires in 72 hours. For free permanent hosting and GPU upgrades, run `gradio deploy` from Terminal to deploy to Spaces (https://huggingface.co/spaces)
Keyboard interruption in main thread... closing server.
Traceback (most recent call last):File "D:\anaconda3\envs\myenvp\lib\site-packages\gradio\blocks.py", line 2361, in block_threadtime.sleep(0.1)
KeyboardInterruptDuring handling of the above exception, another exception occurred:Traceback (most recent call last):File "E:\AI\Janus\demo\app_januspro.py", line 244, in <module>demo.launch(share=True)File "D:\anaconda3\envs\myenvp\lib\site-packages\gradio\blocks.py", line 2266, in launchself.block_thread()File "D:\anaconda3\envs\myenvp\lib\site-packages\gradio\blocks.py", line 2365, in block_threadself.server.close()File "D:\anaconda3\envs\myenvp\lib\site-packages\gradio\networking.py", line 75, in closeself.thread.join()File "D:\anaconda3\envs\myenvp\lib\threading.py", line 1096, in joinself._wait_for_tstate_lock()File "D:\anaconda3\envs\myenvp\lib\threading.py", line 1116, in _wait_for_tstate_lockif lock.acquire(block, timeout):
KeyboardInterrupt
Killing tunnel 127.0.0.1:7860 <> https://342ecb20d5120e7d8c.gradio.live
^C

5 部分部署过程

(myenvp) E:\AI\Janus>pip install -e .
Obtaining file:///E:/AI/JanusInstalling build dependencies ... doneChecking if build backend supports build_editable ... doneGetting requirements to build editable ... donePreparing editable metadata (pyproject.toml) ... done
Requirement already satisfied: torch>=2.0.1 in d:\anaconda3\envs\myenvp\lib\site-packages (from janus==1.0.0) (2.5.1+cu121)
Requirement already satisfied: transformers>=4.38.2 in d:\anaconda3\envs\myenvp\lib\site-packages (from janus==1.0.0) (4.48.1)
Requirement already satisfied: timm>=0.9.16 in d:\anaconda3\envs\myenvp\lib\site-packages (from janus==1.0.0) (1.0.14)
Requirement already satisfied: accelerate in d:\anaconda3\envs\myenvp\lib\site-packages (from janus==1.0.0) (1.3.0)
Requirement already satisfied: sentencepiece in d:\anaconda3\envs\myenvp\lib\site-packages (from janus==1.0.0) (0.1.96)
Requirement already satisfied: attrdict in d:\anaconda3\envs\myenvp\lib\site-packages (from janus==1.0.0) (2.0.1)
Requirement already satisfied: einops in d:\anaconda3\envs\myenvp\lib\site-packages (from janus==1.0.0) (0.8.0)
Requirement already satisfied: torchvision in d:\anaconda3\envs\myenvp\lib\site-packages (from timm>=0.9.16->janus==1.0.0) (0.20.1+cu121)
Requirement already satisfied: pyyaml in d:\anaconda3\envs\myenvp\lib\site-packages (from timm>=0.9.16->janus==1.0.0) (6.0.2)
Requirement already satisfied: huggingface_hub in d:\anaconda3\envs\myenvp\lib\site-packages (from timm>=0.9.16->janus==1.0.0) (0.28.0)
Requirement already satisfied: safetensors in d:\anaconda3\envs\myenvp\lib\site-packages (from timm>=0.9.16->janus==1.0.0) (0.5.2)
Requirement already satisfied: filelock in d:\anaconda3\envs\myenvp\lib\site-packages (from torch>=2.0.1->janus==1.0.0) (3.17.0)
Requirement already satisfied: typing-extensions>=4.8.0 in d:\anaconda3\envs\myenvp\lib\site-packages (from torch>=2.0.1->janus==1.0.0) (4.12.2)
Requirement already satisfied: networkx in d:\anaconda3\envs\myenvp\lib\site-packages (from torch>=2.0.1->janus==1.0.0) (3.4.2)
Requirement already satisfied: jinja2 in d:\anaconda3\envs\myenvp\lib\site-packages (from torch>=2.0.1->janus==1.0.0) (3.1.5)
Requirement already satisfied: fsspec in d:\anaconda3\envs\myenvp\lib\site-packages (from torch>=2.0.1->janus==1.0.0) (2024.12.0)
Requirement already satisfied: sympy==1.13.1 in d:\anaconda3\envs\myenvp\lib\site-packages (from torch>=2.0.1->janus==1.0.0) (1.13.1)
Requirement already satisfied: mpmath<1.4,>=1.1.0 in d:\anaconda3\envs\myenvp\lib\site-packages (from sympy==1.13.1->torch>=2.0.1->janus==1.0.0) (1.3.0)
Requirement already satisfied: numpy>=1.17 in d:\anaconda3\envs\myenvp\lib\site-packages (from transformers>=4.38.2->janus==1.0.0) (1.26.4)
Requirement already satisfied: packaging>=20.0 in d:\anaconda3\envs\myenvp\lib\site-packages (from transformers>=4.38.2->janus==1.0.0) (24.2)
Requirement already satisfied: regex!=2019.12.17 in d:\anaconda3\envs\myenvp\lib\site-packages (from transformers>=4.38.2->janus==1.0.0) (2024.11.6)
Requirement already satisfied: requests in d:\anaconda3\envs\myenvp\lib\site-packages (from transformers>=4.38.2->janus==1.0.0) (2.32.3)
Requirement already satisfied: tokenizers<0.22,>=0.21 in d:\anaconda3\envs\myenvp\lib\site-packages (from transformers>=4.38.2->janus==1.0.0) (0.21.0)
Requirement already satisfied: tqdm>=4.27 in d:\anaconda3\envs\myenvp\lib\site-packages (from transformers>=4.38.2->janus==1.0.0) (4.64.0)
Requirement already satisfied: psutil in d:\anaconda3\envs\myenvp\lib\site-packages (from accelerate->janus==1.0.0) (6.1.1)
Requirement already satisfied: six in d:\anaconda3\envs\myenvp\lib\site-packages (from attrdict->janus==1.0.0) (1.17.0)
Requirement already satisfied: colorama in d:\anaconda3\envs\myenvp\lib\site-packages (from tqdm>=4.27->transformers>=4.38.2->janus==1.0.0) (0.4.5)
Requirement already satisfied: MarkupSafe>=2.0 in d:\anaconda3\envs\myenvp\lib\site-packages (from jinja2->torch>=2.0.1->janus==1.0.0) (2.1.5)
Requirement already satisfied: charset-normalizer<4,>=2 in d:\anaconda3\envs\myenvp\lib\site-packages (from requests->transformers>=4.38.2->janus==1.0.0) (3.4.1)
Requirement already satisfied: idna<4,>=2.5 in d:\anaconda3\envs\myenvp\lib\site-packages (from requests->transformers>=4.38.2->janus==1.0.0) (3.10)
Requirement already satisfied: urllib3<3,>=1.21.1 in d:\anaconda3\envs\myenvp\lib\site-packages (from requests->transformers>=4.38.2->janus==1.0.0) (2.3.0)
Requirement already satisfied: certifi>=2017.4.17 in d:\anaconda3\envs\myenvp\lib\site-packages (from requests->transformers>=4.38.2->janus==1.0.0) (2024.12.14)
Requirement already satisfied: pillow!=8.3.*,>=5.3.0 in d:\anaconda3\envs\myenvp\lib\site-packages (from torchvision->timm>=0.9.16->janus==1.0.0) (10.4.0)
Building wheels for collected packages: janusBuilding editable for janus (pyproject.toml) ... doneCreated wheel for janus: filename=janus-1.0.0-0.editable-py3-none-any.whl size=16196 sha256=cdb0ebb0c36046bf768a84cbf9208824eadb31fadea888f3b6ff102de576f743Stored in directory: C:\Users\Administrator\AppData\Local\Temp\pip-ephem-wheel-cache-dhnej7iy\wheels\e4\87\ba\dd6e5c70086c786d25bcd3e6bddaeb7c46f5ae69dc25ea8be0
Successfully built janus
Installing collected packages: janusAttempting uninstall: janusFound existing installation: janus 1.0.0Uninstalling janus-1.0.0:Successfully uninstalled janus-1.0.0
Successfully installed janus-1.0.0(myenvp) E:\AI\Janus>pip install webencodings beautifulsoup4 tinycss2
Requirement already satisfied: webencodings in d:\anaconda3\envs\myenvp\lib\site-packages (0.5.1)
Requirement already satisfied: beautifulsoup4 in d:\anaconda3\envs\myenvp\lib\site-packages (4.12.3)
Requirement already satisfied: tinycss2 in d:\anaconda3\envs\myenvp\lib\site-packages (1.4.0)
Requirement already satisfied: soupsieve>1.2 in d:\anaconda3\envs\myenvp\lib\site-packages (from beautifulsoup4) (2.6)(myenvp) E:\AI\Janus>pip install 'pexpect>4.3'
ERROR: Invalid requirement: "'pexpect": Expected package name at the start of dependency specifier'pexpect^(myenvp) E:\AI\Janus>pip install 'pexpect>4.3'
ERROR: Invalid requirement: "'pexpect": Expected package name at the start of dependency specifier'pexpect^(myenvp) E:\AI\Janus>pip install "pexpect>4.3"
Requirement already satisfied: pexpect>4.3 in d:\anaconda3\envs\myenvp\lib\site-packages (4.9.0)
Requirement already satisfied: ptyprocess>=0.5 in d:\anaconda3\envs\myenvp\lib\site-packages (from pexpect>4.3) (0.7.0)(myenvp) E:\AI\Janus>python demo/app_januspro.py
Python version is above 3.10, patching the collections module.
D:\anaconda3\envs\myenvp\lib\site-packages\transformers\models\auto\image_processing_auto.py:590: FutureWarning: The image_processor_class argument is deprecated and will be removed in v4.42. Please use `slow_image_processor_class`, or `fast_image_processor_class` insteadwarnings.warn(
Loading checkpoint shards: 100%|█████████████████████████████████████████████████████████| 2/2 [00:06<00:00,  3.25s/it]
Using a slow image processor as `use_fast` is unset and a slow processor was saved with this model. `use_fast=True` will be the default behavior in v4.48, even if the model was saved with a slow processor. This will result in minor differences in outputs. You'll still be able to use a slow processor with `use_fast=False`.
You are using the default legacy behaviour of the <class 'transformers.models.llama.tokenization_llama_fast.LlamaTokenizerFast'>. This is expected, and simply means that the `legacy` (previous) behavior will be used so nothing changes for you. If you want to use the new behaviour, set `legacy=False`. This should only be set if you understand what it means, and thoroughly read the reason why this was added as explained in https://github.com/huggingface/transformers/pull/24565 - if you loaded a llama tokenizer from a GGUF file you can ignore this message.
Some kwargs in processor config are unused and will not have any effect: ignore_id, sft_format, image_tag, num_image_tokens, mask_prompt, add_special_token.
Running on local URL:  http://127.0.0.1:7860
IMPORTANT: You are using gradio version 3.48.0, however version 4.44.1 is available, please upgrade.
--------
Running on public URL: https://b0590adff3d54b2255.gradio.liveThis share link expires in 72 hours. For free permanent hosting and GPU upgrades, run `gradio deploy` from Terminal to deploy to Spaces (https://huggingface.co/spaces)
Keyboard interruption in main thread... closing server.
Killing tunnel 127.0.0.1:7860 <> https://b0590adff3d54b2255.gradio.live(myenvp) E:\AI\Janus>python demo/app_januspro.py --device cuda
Python version is above 3.10, patching the collections module.
D:\anaconda3\envs\myenvp\lib\site-packages\transformers\models\auto\image_processing_auto.py:590: FutureWarning: The image_processor_class argument is deprecated and will be removed in v4.42. Please use `slow_image_processor_class`, or `fast_image_processor_class` insteadwarnings.warn(
Loading checkpoint shards: 100%|█████████████████████████████████████████████████████████| 2/2 [00:06<00:00,  3.05s/it]
Using a slow image processor as `use_fast` is unset and a slow processor was saved with this model. `use_fast=True` will be the default behavior in v4.48, even if the model was saved with a slow processor. This will result in minor differences in outputs. You'll still be able to use a slow processor with `use_fast=False`.
You are using the default legacy behaviour of the <class 'transformers.models.llama.tokenization_llama_fast.LlamaTokenizerFast'>. This is expected, and simply means that the `legacy` (previous) behavior will be used so nothing changes for you. If you want to use the new behaviour, set `legacy=False`. This should only be set if you understand what it means, and thoroughly read the reason why this was added as explained in https://github.com/huggingface/transformers/pull/24565 - if you loaded a llama tokenizer from a GGUF file you can ignore this message.
Some kwargs in processor config are unused and will not have any effect: image_tag, sft_format, ignore_id, add_special_token, num_image_tokens, mask_prompt.
Running on local URL:  http://127.0.0.1:7860
IMPORTANT: You are using gradio version 3.48.0, however version 4.44.1 is available, please upgrade.
--------
Running on public URL: https://72d4294c2d37f91dc8.gradio.liveThis share link expires in 72 hours. For free permanent hosting and GPU upgrades, run `gradio deploy` from Terminal to deploy to Spaces (https://huggingface.co/spaces)

6 使用效果

6.1 识别图片

6.2 文生图

6.2.1 浣熊师父身穿滴水服装，扮演街头歹徒。

Master shifu racoon wearing drip attire as a street gangster.

6.2.2 美丽女孩的脸

The face of a beautiful girl

6.2.3 丛林中的宇航员，冷色调，柔和的色彩，细节丰富，8k

Astronaut in a jungle, cold color palette, muted colors, detailed, 8k

6.2.4 反光面上的一杯红酒。

A glass of red wine on a reflective surface.

6.2.5 一只可爱又迷人的小狐狸，有着大大的棕色眼睛，背景中秋叶迷人，永恒、蓬松、闪亮的鬃毛、花瓣、童话般的氛围，虚幻引擎 5 和 Octane 渲染器，细节丰富，具有照片级真实感，具有电影感，色彩自然。

A cute and adorable baby fox with big brown eyes, autumn leaves in the background enchanting,immortal,fluffy, shiny mane,Petals,fairyism,unreal engine 5 and Octane Render,highly detailed, photorealistic, cinematic, natural colors.

6.2.6 这幅画中的眼睛设计精巧，背景为圆形，饰有华丽的漩涡图案，既有现实主义的色彩，也有超现实主义的色彩。画中焦点是一只鲜艳的蓝色虹膜，周围环绕着从瞳孔向外辐射的细纹，营造出深度和强度。睫毛又长又黑，在周围的皮肤上投下微妙的阴影，皮肤看起来很光滑，但略带纹理，仿佛随着时间的流逝而老化或风化。眼睛上方有一个类似古典建筑的石头结构，为构图增添了神秘感和永恒的优雅。这一建筑元素与周围的有机曲线形成鲜明而和谐的对比。眼睛下方是另一个让人联想到巴洛克艺术的装饰图案，进一步增强了每个精心制作的细节所蕴含的整体永恒感。总体而言，氛围散发着一种神秘的气氛，与暗示永恒的元素无缝交织在一起，通过现实纹理和超现实艺术的并置实现。每一个组成部分——从吸引眼球的复杂设计到上方古老的石块——都以独特的方式创造出充满神秘魅力的视觉盛宴。

The image features an intricately designed eye set against a circular backdrop adorned with ornate swirl patterns that evoke both realism and surrealism. At the center of attention is a strikingly vivid blue iris surrounded by delicate veins radiating outward from the pupil to create depth and intensity. The eyelashes are long and dark, casting subtle shadows on the skin around them which appears smooth yet slightly textured as if aged or weathered over time.

Above the eye, there’s a stone-like structure resembling part of classical architecture, adding layers of mystery and timeless elegance to the composition. This architectural element contrasts sharply but harmoniously with the organic curves surrounding it. Below the eye lies another decorative motif reminiscent of baroque artistry, further enhancing the overall sense of eternity encapsulated within each meticulously crafted detail.

Overall, the atmosphere exudes a mysterious aura intertwined seamlessly with elements suggesting timelessness, achieved through the juxtaposition of realistic textures and surreal artistic flourishes. Each component—from the intricate designs framing the eye to the ancient-looking stone piece above—contributes uniquely towards creating a visually captivating tableau imbued with enigmatic allure.