Skip to content

Commit d1d6167

Browse files
committed
Add media glossary, enhance mpv shortcuts, update resource links, improve network debugging commands, refine Shadowsocks setup, enhance Sing-Box documentation, expand HTTP/3 information, update rclone usage examples, enrich software resources, document Chrome versioning, enhance color resources, improve pnpm configuration details, expand web development tools, refine MikroORM documentation, clarify Node.js TypeScript execution, enhance JavaScript decorators references, update JavaScript versioning details, clarify bignumber.js usage, add Cascadia Code font reference, and document Storybook versioning.
1 parent acbf2d4 commit d1d6167

Some content is hidden

Large Commits have some content hidden by default. Use the searchbox below for content that may be hidden.

78 files changed

+1999
-638
lines changed

notes/ai/ai-awesome.md

Lines changed: 9 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -124,6 +124,9 @@ tags:
124124
- Agent/MCP
125125
- [microsoft/autogen](https://github.com/microsoft/autogen)
126126
- Multi Agent
127+
- [magnitudedev/magnitude](https://github.com/magnitudedev/magnitude)
128+
- Aapche-2.0, TS
129+
- browser automation framework
127130
- [google-gemini/gemini-fullstack-langgraph-quickstart](https://github.com/google-gemini/gemini-fullstack-langgraph-quickstart)
128131
- Apache-2.0, TS, Python, React
129132
- building Fullstack Agents using Gemini 2.5 and LangGraph
@@ -159,15 +162,14 @@ tags:
159162
- Platform
160163
- [langfuse/langfuse](https://github.com/langfuse/langfuse)
161164
- MIT+EE, TS
162-
- LLM Observability, metrics, evals, prompt management, playground, datasets.
165+
- LLM Observability, metrics, evals, prompt management, playground, datasets.
163166
- https://github.com/NielsRogge/Transformers-Tutorials
164167
- news/trending
165168
- https://huggingface.co/models
166169
- https://aihot.today/
167170
- usecase/showcase
168171
- [601 real-world gen AI use cases from the world's leading organizations](https://cloud.google.com/transform/101-real-world-generative-ai-use-cases-from-industry-leaders)
169172

170-
171173
## 应用 {#applications}
172174

173175
**产品/服务**
@@ -205,6 +207,11 @@ tags:
205207

206208
---
207209

210+
- [automattic/harper](https://github.com/automattic/harper)
211+
- Apache-2.0, Rust
212+
- privacy-first grammar checker
213+
- English grammar checker
214+
- alternative to Grammarly
208215
- [openrecall/openrecall](https://github.com/openrecall/openrecall)
209216
- AGPLv3, Python
210217
- Windows 11 Recall 替代

notes/ai/ai-glossary.md

Lines changed: 5 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -67,6 +67,11 @@ tags:
6767
- open-vocabulary detection
6868
- 开放词汇检测
6969
- 识别和处理未在训练数据中出现过的词汇或短语的能力
70+
- CAM (Class Activation Mapping)(Heatmap)
71+
- 类激活映射
72+
- 用于可视化卷积神经网络(CNN)在图像分类任务中关注的区域
73+
- 通过将特定类别的预测与输入图像的特征图进行关联,生成热力图,显示模型在做出决策时关注的图像区域
74+
- 模型可解释性
7075

7176
| en | cn |
7277
| ---------------- | -------- |

notes/ai/dev/comfyui.md

Lines changed: 81 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -18,6 +18,15 @@ title: ComfyUI
1818
- ComfyDeploy
1919
- 另存 (API 格式)
2020
- 调用 /prompt
21+
- 参考
22+
- https://docs.comfy.org/tutorials/
23+
24+
:::caution
25+
26+
- 只能单一显卡
27+
- https://github.com/comfyanonymous/ComfyUI/discussions/4139
28+
29+
:::
2130

2231
```bash
2332
git clone --depth 1 https://github.com/comfyanonymous/ComfyUI ComfyUI
@@ -36,6 +45,9 @@ uv pip install --pre torch torchvision torchaudio --extra-index-url https://down
3645
uv pip install torch torchvision torchaudio --extra-index-url https://download.pytorch.org/whl/cu121
3746
#uv pip install torch torchvision torchaudio --extra-index-url https://download.pytorch.org/whl/cu128
3847
uv pip install -r requirements.txt
48+
49+
# Nvidia APEX normalization not installed, using PyTorch LayerNorm
50+
uv pip install xformers
3951
```
4052

4153
```bash title="mps.py"
@@ -77,6 +89,33 @@ No module named pip
7789
7890
## Notes
7991
92+
好的,遵照您的要求,这里是精简后的版本,仅包含**目录****主要用途说明**两列。
93+
94+
### ComfyUI Models 目录结构详解
95+
96+
| dir | for |
97+
| :----------------- | :-------------------------------------------------------------------------------------------------------------------------------------------------------- |
98+
| `checkpoints` | 核心基础模型,也叫“大模型”。这是文生图的起点,决定了图像生成的基础风格和能力。例如 Stable Diffusion v1.5, SDXL, 以及社区训练的各种整合模型。 |
99+
| `loras` | LoRA 模型。这些是小型微调文件,用于向基础模型添加特定的角色、画风、概念或服装,灵活性极高。 |
100+
| `vae` | VAE 模型。用于图像的编码和解码。独立的 VAE 文件可以修正图像的色彩(如改善灰蒙蒙的问题)或修复手部等细节问题。SDXL 模型通常不需要额外 VAE。 |
101+
| `controlnet` | ControlNet 模型。用于精确控制图像的生成,例如通过姿势骨架、深度图、线稿、二维码等来引导构图和内容。 |
102+
| `upscale_models` | 图像放大模型。用于“图像放大 (模型)”节点,提升图片分辨率并优化细节。例如 ESRGAN, SwinIR, 4x-UltraSharp 等。 |
103+
| `embeddings` | 文本反演 (Textual Inversion) 嵌入,也叫 Embedding。这些是极小的文件,通过一个关键词触发特定的概念、角色或画风。常用于负面提示词(如 `bad-hands-5`)。 |
104+
| `clip` | CLIP 文本编码器模型。通常 ComfyUI 会自动从大模型中加载,但你也可以把独立的 CLIP 模型放在这里,供高级工作流使用。 |
105+
| `clip_vision` | CLIP Vision 模型。用于分析图像内容,是 IPAdapter、PhotoMaker 等“图像提示”功能的核心组件。 |
106+
| `style_models` | 风格模型。主要用于 T2I-Adapter,功能与 ControlNet 类似,但更侧重于风格的迁移。 |
107+
| `hypernetworks` | Hypernetwork 模型。一种比 LoRA 更早出现的微调技术,现在已不常用,但 ComfyUI 仍然支持加载。 |
108+
| `unet` | U-Net 模型。U-Net 是 Stable Diffusion 模型的核心降噪网络。普通用户几乎不会用到这个目录,主要用于模型开发和研究,将 U-Net 单独分离出来加载。 |
109+
| `text_encoders` | 文本编码器模型。与 `unet` 类似,用于模型研究,允许单独加载和替换文本编码器部分。 |
110+
| `photomaker` | PhotoMaker 模型。一种专门用于根据输入人脸照片生成统一角色的模型。 |
111+
| `sams` | SAM (Segment Anything Model) 模型。由 Meta 开发的图像分割模型,在 ComfyUI 中用于精确地创建和分离遮罩 (Mask)。 |
112+
| `gligen` | GLIGEN 模型。用于“限定区域生成”,允许你通过画框来指定某个物体在图像中的特定位置和大小。 |
113+
| `diffusers` | 用于存放 Hugging Face 的 Diffusers 格式模型。这种格式是一个包含多个子目录和文件的文件夹,而不是单个文件。ComfyUI 可以直接加载这种格式。 |
114+
| `configs` | 配置文件。存放一些旧的 `.ckpt` 模型所需要的 `.yaml` 配置文件,以帮助 ComfyUI 识别其模型架构(如 v1 或 v2)。现在的 `.safetensors` 模型通常不需要。 |
115+
| `vae_approx` | VAE 近似解码器模型。这些是极小的、速度极快的模型,用于在 KSampler 采样过程中生成快速预览图,而不是每次都调用完整的 VAE。 |
116+
| `onnx` | ONNX 模型。用于存放已转换为 ONNX (Open Neural Network Exchange) 格式的模型,通常用于在非 NVIDIA 硬件(如 AMD 显卡)上通过 DirectML 或 Olive 进行推理。 |
117+
| `diffusion_models` | 扩散模型组件。一个更通用的目录,类似于 `unet`,用于存放扩散模型的某些部分。主要供模型开发者使用。 |
118+
80119
**AI Art**
81120
82121
- Text2Img
@@ -119,6 +158,22 @@ No module named pip
119158
- SD 1.5
120159
- LAION 5B
121160
- SDXL
161+
- Upscaler
162+
- ESRGAN
163+
- SwinIR
164+
- 4x-UltraSharp
165+
- OmniSR
166+
- MoSR
167+
- DRCT
168+
- ADT
169+
- DAT
170+
- RealPLKSR
171+
- SPAN
172+
- RGT
173+
- HAT
174+
- SRFormer
175+
- SwiftESRGAN
176+
- SPSR
122177
- KSampler
123178
- 用于采样生成图像
124179
- sampler
@@ -142,6 +197,22 @@ No module named pip
142197
- 912x1216
143198
- 1008x1344
144199
- 9:16
200+
- 512x896
201+
- 576x1024
202+
- 768x1366
203+
- 1024x1820
204+
205+
输出可以包含日期
206+
207+
```
208+
%date:yyyy-MM-dd%/ComfyUI
209+
```
210+
211+
## API
212+
213+
```bash
214+
215+
```
145216
146217
## 参考 {#reference}
147218
@@ -226,3 +297,13 @@ CUDA kernel errors might be asynchronously reported at some other API call, so t
226297
For debugging consider passing CUDA_LAUNCH_BLOCKING=1
227298
Compile with `TORCH_USE_CUDA_DSA` to enable device-side assertions.
228299
```
300+
301+
## ImportError: cannot import name 'guidedFilter' from 'cv2.ximgproc'
302+
303+
```bash
304+
uv pip uninstall opencv-python opencv-python-headless opencv-contrib-python-headless opencv-contrib-python
305+
uv pip install opencv-python opencv-python-headless opencv-contrib-python-headless
306+
uv pip install opencv-contrib-python
307+
```
308+
309+
- https://github.com/chflame163/ComfyUI_LayerStyle/issues/5
File renamed without changes.

notes/ai/dev/localai.md

Lines changed: 89 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,89 @@
1+
---
2+
tags:
3+
- Engine
4+
---
5+
6+
# LocalAI
7+
8+
- [mudler/LocalAI](https://github.com/mudler/LocalAI)
9+
- MIT, Go
10+
- 支持后端 llama.cpp, vLLM, Diffusers, Transformers, whisper.cpp, [stable-diffusion.cpp](https://github.com/leejet/stable-diffusion.cpp)
11+
- 支持 文、图、音
12+
- 支持 文生文、文生图、文生音、语音转文本、向量嵌入
13+
- OpenAI 兼容 API
14+
- 参考
15+
- https://models.localai.io/
16+
- 兼容模型 https://localai.io/model-compatibility/index.html
17+
18+
```bash
19+
# CPU: latest
20+
# Nvidia: latest-gpu-nvidia-cuda-12, latest-gpu-nvidia-cuda-11, latest-nvidia-l4t-arm64
21+
# AMD GPU ROCm: latest-gpu-hipblas
22+
# 推荐 AIO/All in One - 例如 latest-aio-cpu, latest-aio-gpu-nvidia-cuda-12
23+
# 健康 http://localhost:8080/readyz
24+
# 参考 https://localai.io/basics/container/
25+
# 会直接下载一些模型 https://github.com/mudler/LocalAI/tree/master/aio/gpu-8g
26+
# qwen2.5-7b, whisper-1, DreamShaper, MiniCPM-V-2_6
27+
# 内置模型名 whisper-1, stablediffusion
28+
# 支持 https_proxy
29+
docker run --rm -ti --gpus all \
30+
-p 8080:8080 \
31+
-v $PWD/localai/models:/build/models:cached \
32+
--name localai localai/localai:latest-aio-gpu-nvidia-cuda-12
33+
34+
docker exec -it localai bash
35+
./local-ai
36+
```
37+
38+
```yaml
39+
version: '3.9'
40+
services:
41+
api:
42+
image: localai/localai:localai/localai:latest-aio-gpu-nvidia-cuda-12
43+
healthcheck:
44+
test: ['CMD', 'curl', '-f', 'http://localhost:8080/readyz']
45+
interval: 1m
46+
timeout: 20m
47+
retries: 5
48+
ports:
49+
- 8080:8080
50+
environment:
51+
- DEBUG=true
52+
volumes:
53+
- ./models:/build/models:cached
54+
deploy:
55+
resources:
56+
reservations:
57+
devices:
58+
- driver: nvidia
59+
count: 1
60+
capabilities: [gpu]
61+
```
62+
63+
| env | default | desc |
64+
| ----------- | ------------- | ------------------- |
65+
| MODELS_PATH | /build/models |
66+
| THREADS | nproc-1 |
67+
| PORT | 8080 |
68+
| API_KEY |
69+
| MODELS | | list of models YAML |
70+
| PROFILE | | cpu, gpu-8g |
71+
72+
- 容器镜像
73+
- quay.io/go-skynet/local-ai
74+
- docker.m.daocloud.io/localai/localai
75+
- MODELS_PATH
76+
- 容器 /build/models
77+
- /usr/share/local-ai/models
78+
79+
## RuntimeError: operator torchvision::nms does not exist
80+
81+
82+
- 确认版本一致
83+
- `torch.__version__` 和 `torchvision.__version__`
84+
85+
```bash
86+
python -c "import torch; print(torch.__version__)"
87+
python -c "import torchvision; print(torchvision.__version__)"
88+
```
89+

notes/ai/lm/ollama.md renamed to notes/ai/dev/ollama.md

Lines changed: 3 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -31,6 +31,7 @@ title: ollama
3131
- ~~Support tools in OpenAI-compatible API [#4386](https://github.com/ollama/ollama/issues/4386)~~
3232
- phi4-multimodel https://github.com/ollama/ollama/issues/9387
3333
- qwen3 rerank https://github.com/ollama/ollama/issues/10989
34+
- InternVL3 https://github.com/ollama/ollama/issues/10248
3435

3536
:::
3637

@@ -78,6 +79,8 @@ ollama pull llama3.2-vision:11b
7879

7980
- https://github.com/ollama/ollama/blob/main/envconfig/config.go
8081
- https://github.com/ollama/ollama/issues/2941
82+
- OLLAMA_KV_CACHE_TYPE 依赖 flash attention
83+
- https://github.com/ollama/ollama/blob/502028968ddca04bd19c0859a73fb4e0cbeac3e1/llm/server.go#L221-L223
8184

8285
```shell
8386
/set verbose
File renamed without changes.

notes/ai/diffusion/diffusion-awesome.md

Lines changed: 0 additions & 15 deletions
Original file line numberDiff line numberDiff line change
@@ -86,18 +86,3 @@ tags:
8686
- CompVis
8787
- 慕尼黑大学 (LMU Munich) 的计算机视觉与学习研究组 (Computer Vision & Learning Group)
8888
- LAION: 大型人工智能开放网络 (Large-scale Artificial Intelligence Open Network),提供了训练数据集。
89-
90-
| date | model | size | author | notes |
91-
| ------- | ------------------ | -------- | ----------------- | ----- |
92-
| 2024-10 | FLUX.1 | dev 12B | Black Forest Labs |
93-
| 2024-10 | SD 3.5 | 2.5B, 8B |
94-
| 2024-02 | SD 3.0 | 800M, 8B |
95-
| 2023-11 | SDXL Turbo | |
96-
| 2023-07 | SDXL 1.0 | 3.5B |
97-
| 2022-12 | SD v2.1 |
98-
| 2022-11 | SD v2.0 | |
99-
| 2022-10 | SD 1.5 | 983M | RunwayML |
100-
| 2022-08 | SD 1.1 1.2 1.3 1.4 | | CompVis |
101-
102-
- CFG - Classifier-Free Diffusion Guidance (2022)
103-
- [black-forest-labs/flux](https://github.com/black-forest-labs/flux)

notes/ai/ml/huggingface.md

Lines changed: 15 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -82,6 +82,21 @@ completion = client.chat.completions.create(
8282
)
8383
```
8484

85+
## modelscope
86+
87+
```bash
88+
# 国内镜像下载 HF Repo
89+
pip install modelscope
90+
```
91+
92+
- cache_dir= ~/.cache/modelscope/hub
93+
- allow_patterns
94+
- ignore_patterns
95+
- --include
96+
- --exclude
97+
- 模型缓存目录
98+
- cache_dir/MODEL_ID/THE_MODEL_FILES
99+
85100
# FAQ
86101

87102
## The model Qwen/Qwen2.5-VL-72B-Instruct is too large to be loaded automatically (146GB > 10GB).

notes/ai/ml/ml-awesome.md

Lines changed: 3 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -85,6 +85,9 @@ tags:
8585
- [Tencent/ncnn](https://github.com/Tencent/ncnn)
8686
- BSD-3, C++, C
8787
- neural network inference framework optimized for the mobile platform
88+
- [gpustack/gpustack](https://github.com/gpustack/gpustack)
89+
- Apache-2.0, Python
90+
- backend llama.cpp, stable-diffusion.cpp , vLLM, vox-box
8891
- ~~[johnolafenwa/deepstack](https://github.com/johnolafenwa/deepstack)~~
8992
- Apache-2.0, Go, Python
9093
- Cross Platform AI Engine for Edge Devices

0 commit comments

Comments
 (0)