[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M6G98D671CAP7DCT8RSP32PD":3,"skill-same:01M6G98D671CAP7DCT8RSP32PD":86},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":16,"categories":17,"keywords":23,"publishedAt":43,"linkOnly":44,"markdownContent":45,"markdownBody":46,"frontMatter":47,"refName":64,"commitSha":65,"skillDirectory":66,"forksCount":67,"repositoryHtmlUrl":68,"repositoryHomepageUrl":6,"ownerHtmlUrl":69,"ownerBlogUrl":6,"ownerDescription":70,"updatedAt":71,"install":72,"resources":85},"01M6G98D671CAP7DCT8RSP32PD","llava",null,"Large Language and Vision Assistant. Enables visual instruction tuning and image-based conversations. Combines CLIP vision encoder with Vicuna\u002FLLaMA language models. Supports multi-turn image chat, visual question answering, and instruction following. Use for vision-language chatbots or image understanding tasks. Best for conversational image analysis.","zapabob\u002FSkills","zapabob","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F130830376?v=4","01M6DBMYZAGRTFQDSQYW3W764P","峯岸 亮","01M6DBMZ3GKDSJ01NZ9XTZD7JD",2,"MIT","https:\u002F\u002Fgithub.com\u002Fzapabob\u002FSkills\u002Fblob\u002Fmain\u002Farchive\u002Fskill-md-reviewed\u002Fsha256\u002F74\u002F74d97b228ac84b180022684fd12cb7d8975ad2bd970a9b75271e28531d4b7787\u002FSKILL.md",[18],{"categoryNo":19,"categoryName":20,"taxonomyType":21,"taxonomyTypeDesc":22},"01JBSEEDCAT000000000000016","图像与多媒体","SCENARIO","使用场景",[5,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42],"question","conversations","image understanding","Vicuna","Image Chat","Visual Question Answering","Vision-Language","vision","clip","image","analysis","tuning","Transformers","multimodal","tasks","windows","linux","macos","assistant","2026-11-10 10:12:30",false,"---\r\nname: llava\r\ndescription: Large Language and Vision Assistant. Enables visual instruction tuning and image-based conversations. Combines CLIP vision encoder with Vicuna\u002FLLaMA language models. Supports multi-turn image chat, visual question answering, and instruction following. Use for vision-language chatbots or image understanding tasks. Best for conversational image analysis.\r\nversion: 1.0.0\r\nauthor: Orchestra Research\r\nlicense: MIT\r\ndependencies: [transformers, torch, pillow]\r\nplatforms: [linux, macos, windows]\r\nmetadata:\r\n  hermes:\r\n    tags: [LLaVA, Vision-Language, Multimodal, Visual Question Answering, Image Chat, CLIP, Vicuna, Conversational AI, Instruction Tuning, VQA]\r\n\r\n---\r\n\r\n# LLaVA - Large Language and Vision Assistant\r\n\r\nOpen-source vision-language model for conversational image understanding.\r\n\r\n## When to use LLaVA\r\n\r\n**Use when:**\r\n- Building vision-language chatbots\r\n- Visual question answering (VQA)\r\n- Image description and captioning\r\n- Multi-turn image conversations\r\n- Visual instruction following\r\n- Document understanding with images\r\n\r\n**Metrics**:\r\n- **23,000+ GitHub stars**\r\n- GPT-4V level capabilities (targeted)\r\n- Apache 2.0 License\r\n- Multiple model sizes (7B-34B params)\r\n\r\n**Use alternatives instead**:\r\n- **GPT-4V**: Highest quality, API-based\r\n- **CLIP**: Simple zero-shot classification\r\n- **BLIP-2**: Better for captioning only\r\n- **Flamingo**: Research, not open-source\r\n\r\n## Quick start\r\n\r\n### Installation\r\n\r\n```bash\r\n# Clone repository\r\ngit clone https:\u002F\u002Fgithub.com\u002Fhaotian-liu\u002FLLaVA\r\ncd LLaVA\r\n\r\n# Install\r\npip install -e .\r\n```\r\n\r\n### Basic usage\r\n\r\n```python\r\nfrom llava.model.builder import load_pretrained_model\r\nfrom llava.mm_utils import get_model_name_from_path, process_images, tokenizer_image_token\r\nfrom llava.constants import IMAGE_TOKEN_INDEX, DEFAULT_IMAGE_TOKEN\r\nfrom llava.conversation import conv_templates\r\nfrom PIL import Image\r\nimport torch\r\n\r\n# Load model\r\nmodel_path = \"liuhaotian\u002Fllava-v1.5-7b\"\r\ntokenizer, model, image_processor, context_len = load_pretrained_model(\r\n    model_path=model_path,\r\n    model_base=None,\r\n    model_name=get_model_name_from_path(model_path)\r\n)\r\n\r\n# Load image\r\nimage = Image.open(\"image.jpg\")\r\nimage_tensor = process_images([image], image_processor, model.config)\r\nimage_tensor = image_tensor.to(model.device, dtype=torch.float16)\r\n\r\n# Create conversation\r\nconv = conv_templates[\"llava_v1\"].copy()\r\nconv.append_message(conv.roles[0], DEFAULT_IMAGE_TOKEN + \"\\nWhat is in this image?\")\r\nconv.append_message(conv.roles[1], None)\r\nprompt = conv.get_prompt()\r\n\r\n# Generate response\r\ninput_ids = tokenizer_image_token(prompt, tokenizer, IMAGE_TOKEN_INDEX, return_tensors='pt').unsqueeze(0).to(model.device)\r\n\r\nwith torch.inference_mode():\r\n    output_ids = model.generate(\r\n        input_ids,\r\n        images=image_tensor,\r\n        do_sample=True,\r\n        temperature=0.2,\r\n        max_new_tokens=512\r\n    )\r\n\r\nresponse = tokenizer.decode(output_ids[0], skip_special_tokens=True).strip()\r\nprint(response)\r\n```\r\n\r\n## Available models\r\n\r\n| Model | Parameters | VRAM | Quality |\r\n|-------|------------|------|---------|\r\n| LLaVA-v1.5-7B | 7B | ~14 GB | Good |\r\n| LLaVA-v1.5-13B | 13B | ~28 GB | Better |\r\n| LLaVA-v1.6-34B | 34B | ~70 GB | Best |\r\n\r\n```python\r\n# Load different models\r\nmodel_7b = \"liuhaotian\u002Fllava-v1.5-7b\"\r\nmodel_13b = \"liuhaotian\u002Fllava-v1.5-13b\"\r\nmodel_34b = \"liuhaotian\u002Fllava-v1.6-34b\"\r\n\r\n# 4-bit quantization for lower VRAM\r\nload_4bit = True  # Reduces VRAM by ~4×\r\n```\r\n\r\n## CLI usage\r\n\r\n```bash\r\n# Single image query\r\npython -m llava.serve.cli \\\r\n    --model-path liuhaotian\u002Fllava-v1.5-7b \\\r\n    --image-file image.jpg \\\r\n    --query \"What is in this image?\"\r\n\r\n# Multi-turn conversation\r\npython -m llava.serve.cli \\\r\n    --model-path liuhaotian\u002Fllava-v1.5-7b \\\r\n    --image-file image.jpg\r\n# Then type questions interactively\r\n```\r\n\r\n## Web UI (Gradio)\r\n\r\n```bash\r\n# Launch Gradio interface\r\npython -m llava.serve.gradio_web_server \\\r\n    --model-path liuhaotian\u002Fllava-v1.5-7b \\\r\n    --load-4bit  # Optional: reduce VRAM\r\n\r\n# Access at http:\u002F\u002Flocalhost:7860\r\n```\r\n\r\n## Multi-turn conversations\r\n\r\n```python\r\n# Initialize conversation\r\nconv = conv_templates[\"llava_v1\"].copy()\r\n\r\n# Turn 1\r\nconv.append_message(conv.roles[0], DEFAULT_IMAGE_TOKEN + \"\\nWhat is in this image?\")\r\nconv.append_message(conv.roles[1], None)\r\nresponse1 = generate(conv, model, image)  # \"A dog playing in a park\"\r\n\r\n# Turn 2\r\nconv.messages[-1][1] = response1  # Add previous response\r\nconv.append_message(conv.roles[0], \"What breed is the dog?\")\r\nconv.append_message(conv.roles[1], None)\r\nresponse2 = generate(conv, model, image)  # \"Golden Retriever\"\r\n\r\n# Turn 3\r\nconv.messages[-1][1] = response2\r\nconv.append_message(conv.roles[0], \"What time of day is it?\")\r\nconv.append_message(conv.roles[1], None)\r\nresponse3 = generate(conv, model, image)\r\n```\r\n\r\n## Common tasks\r\n\r\n### Image captioning\r\n\r\n```python\r\nquestion = \"Describe this image in detail.\"\r\nresponse = ask(model, image, question)\r\n```\r\n\r\n### Visual question answering\r\n\r\n```python\r\nquestion = \"How many people are in the image?\"\r\nresponse = ask(model, image, question)\r\n```\r\n\r\n### Object detection (textual)\r\n\r\n```python\r\nquestion = \"List all the objects you can see in this image.\"\r\nresponse = ask(model, image, question)\r\n```\r\n\r\n### Scene understanding\r\n\r\n```python\r\nquestion = \"What is happening in this scene?\"\r\nresponse = ask(model, image, question)\r\n```\r\n\r\n### Document understanding\r\n\r\n```python\r\nquestion = \"What is the main topic of this document?\"\r\nresponse = ask(model, document_image, question)\r\n```\r\n\r\n## Training custom model\r\n\r\n```bash\r\n# Stage 1: Feature alignment (558K image-caption pairs)\r\nbash scripts\u002Fv1_5\u002Fpretrain.sh\r\n\r\n# Stage 2: Visual instruction tuning (150K instruction data)\r\nbash scripts\u002Fv1_5\u002Ffinetune.sh\r\n```\r\n\r\n## Quantization (reduce VRAM)\r\n\r\n```python\r\n# 4-bit quantization\r\ntokenizer, model, image_processor, context_len = load_pretrained_model(\r\n    model_path=\"liuhaotian\u002Fllava-v1.5-13b\",\r\n    model_base=None,\r\n    model_name=get_model_name_from_path(\"liuhaotian\u002Fllava-v1.5-13b\"),\r\n    load_4bit=True  # Reduces VRAM ~4×\r\n)\r\n\r\n# 8-bit quantization\r\nload_8bit=True  # Reduces VRAM ~2×\r\n```\r\n\r\n## Best practices\r\n\r\n1. **Start with 7B model** - Good quality, manageable VRAM\r\n2. **Use 4-bit quantization** - Reduces VRAM significantly\r\n3. **GPU required** - CPU inference extremely slow\r\n4. **Clear prompts** - Specific questions get better answers\r\n5. **Multi-turn conversations** - Maintain conversation context\r\n6. **Temperature 0.2-0.7** - Balance creativity\u002Fconsistency\r\n7. **max_new_tokens 512-1024** - For detailed responses\r\n8. **Batch processing** - Process multiple images sequentially\r\n\r\n## Performance\r\n\r\n| Model | VRAM (FP16) | VRAM (4-bit) | Speed (tokens\u002Fs) |\r\n|-------|-------------|--------------|------------------|\r\n| 7B | ~14 GB | ~4 GB | ~20 |\r\n| 13B | ~28 GB | ~8 GB | ~12 |\r\n| 34B | ~70 GB | ~18 GB | ~5 |\r\n\r\n*On A100 GPU*\r\n\r\n## Benchmarks\r\n\r\nLLaVA achieves competitive scores on:\r\n- **VQAv2**: 78.5%\r\n- **GQA**: 62.0%\r\n- **MM-Vet**: 35.4%\r\n- **MMBench**: 64.3%\r\n\r\n## Limitations\r\n\r\n1. **Hallucinations** - May describe things not in image\r\n2. **Spatial reasoning** - Struggles with precise locations\r\n3. **Small text** - Difficulty reading fine print\r\n4. **Object counting** - Imprecise for many objects\r\n5. **VRAM requirements** - Need powerful GPU\r\n6. **Inference speed** - Slower than CLIP\r\n\r\n## Integration with frameworks\r\n\r\n### LangChain\r\n\r\n```python\r\nfrom langchain.llms.base import LLM\r\n\r\nclass LLaVALLM(LLM):\r\n    def _call(self, prompt, stop=None):\r\n        # Custom LLaVA inference\r\n        return response\r\n\r\nllm = LLaVALLM()\r\n```\r\n\r\n### Gradio App\r\n\r\n```python\r\nimport gradio as gr\r\n\r\ndef chat(image, text, history):\r\n    response = ask_llava(model, image, text)\r\n    return response\r\n\r\ndemo = gr.ChatInterface(\r\n    chat,\r\n    additional_inputs=[gr.Image(type=\"pil\")],\r\n    title=\"LLaVA Chat\"\r\n)\r\ndemo.launch()\r\n```\r\n\r\n## Resources\r\n\r\n- **GitHub**: https:\u002F\u002Fgithub.com\u002Fhaotian-liu\u002FLLaVA ⭐ 23,000+\r\n- **Paper**: https:\u002F\u002Farxiv.org\u002Fabs\u002F2304.08485\r\n- **Demo**: https:\u002F\u002Fllava.hliu.cc\r\n- **Models**: https:\u002F\u002Fhuggingface.co\u002Fliuhaotian\r\n- **License**: Apache 2.0\r\n","# LLaVA - Large Language and Vision Assistant\r\n\r\nOpen-source vision-language model for conversational image understanding.\r\n\r\n## When to use LLaVA\r\n\r\n**Use when:**\r\n- Building vision-language chatbots\r\n- Visual question answering (VQA)\r\n- Image description and captioning\r\n- Multi-turn image conversations\r\n- Visual instruction following\r\n- Document understanding with images\r\n\r\n**Metrics**:\r\n- **23,000+ GitHub stars**\r\n- GPT-4V level capabilities (targeted)\r\n- Apache 2.0 License\r\n- Multiple model sizes (7B-34B params)\r\n\r\n**Use alternatives instead**:\r\n- **GPT-4V**: Highest quality, API-based\r\n- **CLIP**: Simple zero-shot classification\r\n- **BLIP-2**: Better for captioning only\r\n- **Flamingo**: Research, not open-source\r\n\r\n## Quick start\r\n\r\n### Installation\r\n\r\n```bash\r\n# Clone repository\r\ngit clone https:\u002F\u002Fgithub.com\u002Fhaotian-liu\u002FLLaVA\r\ncd LLaVA\r\n\r\n# Install\r\npip install -e .\r\n```\r\n\r\n### Basic usage\r\n\r\n```python\r\nfrom llava.model.builder import load_pretrained_model\r\nfrom llava.mm_utils import get_model_name_from_path, process_images, tokenizer_image_token\r\nfrom llava.constants import IMAGE_TOKEN_INDEX, DEFAULT_IMAGE_TOKEN\r\nfrom llava.conversation import conv_templates\r\nfrom PIL import Image\r\nimport torch\r\n\r\n# Load model\r\nmodel_path = \"liuhaotian\u002Fllava-v1.5-7b\"\r\ntokenizer, model, image_processor, context_len = load_pretrained_model(\r\n    model_path=model_path,\r\n    model_base=None,\r\n    model_name=get_model_name_from_path(model_path)\r\n)\r\n\r\n# Load image\r\nimage = Image.open(\"image.jpg\")\r\nimage_tensor = process_images([image], image_processor, model.config)\r\nimage_tensor = image_tensor.to(model.device, dtype=torch.float16)\r\n\r\n# Create conversation\r\nconv = conv_templates[\"llava_v1\"].copy()\r\nconv.append_message(conv.roles[0], DEFAULT_IMAGE_TOKEN + \"\\nWhat is in this image?\")\r\nconv.append_message(conv.roles[1], None)\r\nprompt = conv.get_prompt()\r\n\r\n# Generate response\r\ninput_ids = tokenizer_image_token(prompt, tokenizer, IMAGE_TOKEN_INDEX, return_tensors='pt').unsqueeze(0).to(model.device)\r\n\r\nwith torch.inference_mode():\r\n    output_ids = model.generate(\r\n        input_ids,\r\n        images=image_tensor,\r\n        do_sample=True,\r\n        temperature=0.2,\r\n        max_new_tokens=512\r\n    )\r\n\r\nresponse = tokenizer.decode(output_ids[0], skip_special_tokens=True).strip()\r\nprint(response)\r\n```\r\n\r\n## Available models\r\n\r\n| Model | Parameters | VRAM | Quality |\r\n|-------|------------|------|---------|\r\n| LLaVA-v1.5-7B | 7B | ~14 GB | Good |\r\n| LLaVA-v1.5-13B | 13B | ~28 GB | Better |\r\n| LLaVA-v1.6-34B | 34B | ~70 GB | Best |\r\n\r\n```python\r\n# Load different models\r\nmodel_7b = \"liuhaotian\u002Fllava-v1.5-7b\"\r\nmodel_13b = \"liuhaotian\u002Fllava-v1.5-13b\"\r\nmodel_34b = \"liuhaotian\u002Fllava-v1.6-34b\"\r\n\r\n# 4-bit quantization for lower VRAM\r\nload_4bit = True  # Reduces VRAM by ~4×\r\n```\r\n\r\n## CLI usage\r\n\r\n```bash\r\n# Single image query\r\npython -m llava.serve.cli \\\r\n    --model-path liuhaotian\u002Fllava-v1.5-7b \\\r\n    --image-file image.jpg \\\r\n    --query \"What is in this image?\"\r\n\r\n# Multi-turn conversation\r\npython -m llava.serve.cli \\\r\n    --model-path liuhaotian\u002Fllava-v1.5-7b \\\r\n    --image-file image.jpg\r\n# Then type questions interactively\r\n```\r\n\r\n## Web UI (Gradio)\r\n\r\n```bash\r\n# Launch Gradio interface\r\npython -m llava.serve.gradio_web_server \\\r\n    --model-path liuhaotian\u002Fllava-v1.5-7b \\\r\n    --load-4bit  # Optional: reduce VRAM\r\n\r\n# Access at http:\u002F\u002Flocalhost:7860\r\n```\r\n\r\n## Multi-turn conversations\r\n\r\n```python\r\n# Initialize conversation\r\nconv = conv_templates[\"llava_v1\"].copy()\r\n\r\n# Turn 1\r\nconv.append_message(conv.roles[0], DEFAULT_IMAGE_TOKEN + \"\\nWhat is in this image?\")\r\nconv.append_message(conv.roles[1], None)\r\nresponse1 = generate(conv, model, image)  # \"A dog playing in a park\"\r\n\r\n# Turn 2\r\nconv.messages[-1][1] = response1  # Add previous response\r\nconv.append_message(conv.roles[0], \"What breed is the dog?\")\r\nconv.append_message(conv.roles[1], None)\r\nresponse2 = generate(conv, model, image)  # \"Golden Retriever\"\r\n\r\n# Turn 3\r\nconv.messages[-1][1] = response2\r\nconv.append_message(conv.roles[0], \"What time of day is it?\")\r\nconv.append_message(conv.roles[1], None)\r\nresponse3 = generate(conv, model, image)\r\n```\r\n\r\n## Common tasks\r\n\r\n### Image captioning\r\n\r\n```python\r\nquestion = \"Describe this image in detail.\"\r\nresponse = ask(model, image, question)\r\n```\r\n\r\n### Visual question answering\r\n\r\n```python\r\nquestion = \"How many people are in the image?\"\r\nresponse = ask(model, image, question)\r\n```\r\n\r\n### Object detection (textual)\r\n\r\n```python\r\nquestion = \"List all the objects you can see in this image.\"\r\nresponse = ask(model, image, question)\r\n```\r\n\r\n### Scene understanding\r\n\r\n```python\r\nquestion = \"What is happening in this scene?\"\r\nresponse = ask(model, image, question)\r\n```\r\n\r\n### Document understanding\r\n\r\n```python\r\nquestion = \"What is the main topic of this document?\"\r\nresponse = ask(model, document_image, question)\r\n```\r\n\r\n## Training custom model\r\n\r\n```bash\r\n# Stage 1: Feature alignment (558K image-caption pairs)\r\nbash scripts\u002Fv1_5\u002Fpretrain.sh\r\n\r\n# Stage 2: Visual instruction tuning (150K instruction data)\r\nbash scripts\u002Fv1_5\u002Ffinetune.sh\r\n```\r\n\r\n## Quantization (reduce VRAM)\r\n\r\n```python\r\n# 4-bit quantization\r\ntokenizer, model, image_processor, context_len = load_pretrained_model(\r\n    model_path=\"liuhaotian\u002Fllava-v1.5-13b\",\r\n    model_base=None,\r\n    model_name=get_model_name_from_path(\"liuhaotian\u002Fllava-v1.5-13b\"),\r\n    load_4bit=True  # Reduces VRAM ~4×\r\n)\r\n\r\n# 8-bit quantization\r\nload_8bit=True  # Reduces VRAM ~2×\r\n```\r\n\r\n## Best practices\r\n\r\n1. **Start with 7B model** - Good quality, manageable VRAM\r\n2. **Use 4-bit quantization** - Reduces VRAM significantly\r\n3. **GPU required** - CPU inference extremely slow\r\n4. **Clear prompts** - Specific questions get better answers\r\n5. **Multi-turn conversations** - Maintain conversation context\r\n6. **Temperature 0.2-0.7** - Balance creativity\u002Fconsistency\r\n7. **max_new_tokens 512-1024** - For detailed responses\r\n8. **Batch processing** - Process multiple images sequentially\r\n\r\n## Performance\r\n\r\n| Model | VRAM (FP16) | VRAM (4-bit) | Speed (tokens\u002Fs) |\r\n|-------|-------------|--------------|------------------|\r\n| 7B | ~14 GB | ~4 GB | ~20 |\r\n| 13B | ~28 GB | ~8 GB | ~12 |\r\n| 34B | ~70 GB | ~18 GB | ~5 |\r\n\r\n*On A100 GPU*\r\n\r\n## Benchmarks\r\n\r\nLLaVA achieves competitive scores on:\r\n- **VQAv2**: 78.5%\r\n- **GQA**: 62.0%\r\n- **MM-Vet**: 35.4%\r\n- **MMBench**: 64.3%\r\n\r\n## Limitations\r\n\r\n1. **Hallucinations** - May describe things not in image\r\n2. **Spatial reasoning** - Struggles with precise locations\r\n3. **Small text** - Difficulty reading fine print\r\n4. **Object counting** - Imprecise for many objects\r\n5. **VRAM requirements** - Need powerful GPU\r\n6. **Inference speed** - Slower than CLIP\r\n\r\n## Integration with frameworks\r\n\r\n### LangChain\r\n\r\n```python\r\nfrom langchain.llms.base import LLM\r\n\r\nclass LLaVALLM(LLM):\r\n    def _call(self, prompt, stop=None):\r\n        # Custom LLaVA inference\r\n        return response\r\n\r\nllm = LLaVALLM()\r\n```\r\n\r\n### Gradio App\r\n\r\n```python\r\nimport gradio as gr\r\n\r\ndef chat(image, text, history):\r\n    response = ask_llava(model, image, text)\r\n    return response\r\n\r\ndemo = gr.ChatInterface(\r\n    chat,\r\n    additional_inputs=[gr.Image(type=\"pil\")],\r\n    title=\"LLaVA Chat\"\r\n)\r\ndemo.launch()\r\n```\r\n\r\n## Resources\r\n\r\n- **GitHub**: https:\u002F\u002Fgithub.com\u002Fhaotian-liu\u002FLLaVA ⭐ 23,000+\r\n- **Paper**: https:\u002F\u002Farxiv.org\u002Fabs\u002F2304.08485\r\n- **Demo**: https:\u002F\u002Fllava.hliu.cc\r\n- **Models**: https:\u002F\u002Fhuggingface.co\u002Fliuhaotian\r\n- **License**: Apache 2.0\r\n",{"name":5,"author":48,"license":15,"version":49,"metadata":50,"platforms":59,"description":7,"dependencies":60},"Orchestra Research","1.0.0",{"hermes":51},{"tags":52},[53,30,54,29,28,55,27,56,57,58],"LLaVA","Multimodal","CLIP","Conversational AI","Instruction Tuning","VQA",[40,41,39],[61,62,63],"transformers","torch","pillow","main","0d1434f89f4bc6b07e9d272accd8c1013768820f","archive\u002Fskill-md-reviewed\u002Fsha256\u002F74\u002F74d97b228ac84b180022684fd12cb7d8975ad2bd970a9b75271e28531d4b7787",0,"https:\u002F\u002Fgithub.com\u002Fzapabob\u002FSkills","https:\u002F\u002Fgithub.com\u002Fzapabob","The Open University of JAPAN　2019-now\r\n\r\nTokyo University of Science　2007-2010","2026-10-26 16:29:01",{"source":8,"skillName":5,"commands":73,"prompt":83,"zipAvailable":84},[74,77,80],{"tool":75,"command":76},"npx","npx skills add zapabob\u002FSkills --skill llava",{"tool":78,"command":79},"bunx","bunx skills add zapabob\u002FSkills --skill llava",{"tool":81,"command":82},"pnpm","pnpm dlx skills add zapabob\u002FSkills --skill llava","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fzapabob\u002FSkills\u002Fblob\u002Fmain\u002Farchive\u002Fskill-md-reviewed\u002Fsha256\u002F74\u002F74d97b228ac84b180022684fd12cb7d8975ad2bd970a9b75271e28531d4b7787\u002FSKILL.md",true,[],[87,120,154,177,203,220],{"skillNo":88,"name":89,"nameZh":6,"description":90,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":91,"categories":92,"keywords":104,"publishedAt":43},"01M6GJTE1GESGGF9K028Y7753E","xurl","X\u002FTwitter via xurl CLI: post, search, DM, media, v2 API.","https:\u002F\u002Fgithub.com\u002Fzapabob\u002FSkills\u002Fblob\u002Fmain\u002Fregistry\u002Fskills\u002Fxurl\u002Fvariants\u002Fhermes-agent\u002FSKILL.md",[93,96,99],{"categoryNo":94,"categoryName":95,"taxonomyType":21,"taxonomyTypeDesc":22},"01JBSEEDCAT000000000000010","后端与接口",{"categoryNo":97,"categoryName":98,"taxonomyType":21,"taxonomyTypeDesc":22},"01JBSEEDCAT000000000000023","Skill与智能体开发",{"categoryNo":100,"categoryName":101,"taxonomyType":102,"taxonomyTypeDesc":103},"01JBSEEDCAT000000000000030","开发工程师","PROFESSION","适用职业",[89,105,41,106,107,40,39,108,109,110,111,112,113,114,115,116,64,117,118,119],"openclaw","cli","github","official","search","https","social-media","api","Media","hermes","hermes agent","twitter","social","via","post",{"skillNo":121,"name":122,"nameZh":6,"description":123,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":124,"categories":125,"keywords":137,"publishedAt":43},"01M6GJT8S1PXQ8WPXT1HSNBWH2","writing-plans","Write implementation plans: bite-sized tasks, paths, code.","https:\u002F\u002Fgithub.com\u002Fzapabob\u002FSkills\u002Fblob\u002Fmain\u002Fregistry\u002Fskills\u002Fwriting-plans\u002Fvariants\u002Fhermes-agent\u002FSKILL.md",[126,129,132,133,134],{"categoryNo":127,"categoryName":128,"taxonomyType":21,"taxonomyTypeDesc":22},"01JBSEEDCAT000000000000015","视觉设计",{"categoryNo":130,"categoryName":131,"taxonomyType":21,"taxonomyTypeDesc":22},"01JBSEEDCAT000000000000018","写作与文案",{"categoryNo":97,"categoryName":98,"taxonomyType":21,"taxonomyTypeDesc":22},{"categoryNo":100,"categoryName":101,"taxonomyType":102,"taxonomyTypeDesc":103},{"categoryNo":135,"categoryName":136,"taxonomyType":102,"taxonomyTypeDesc":103},"01JBSEEDCAT000000000000036","内容创作者",[138,139,140,141,41,142,40,143,39,144,38,145,146,147,148,149,150,151,152,153],"writing","writing plans","plans","design","review","workflow","documentation","development","code","planning","code-review","subagent","write","superpowers","paths","test-driven-development",{"skillNo":155,"name":156,"nameZh":6,"description":157,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":158,"categories":159,"keywords":161,"publishedAt":43},"01M6GJST0H5X7WHBJ3BZK0DVKH","whisper","OpenAI's general-purpose speech recognition model. Supports 99 languages, transcription, translation to English, and language identification. Six model sizes from tiny (39M params) to large (1550M params). Use for speech-to-text, podcast transcription, or multilingual audio processing. Best for robust, multilingual ASR.","https:\u002F\u002Fgithub.com\u002Fzapabob\u002FSkills\u002Fblob\u002Fmain\u002Fregistry\u002Fskills\u002Fwhisper\u002Fvariants\u002Fhermes-agent-optional\u002FSKILL.md",[160],{"categoryNo":19,"categoryName":20,"taxonomyType":21,"taxonomyTypeDesc":22},[156,41,40,162,163,37,36,164,165,166,167,168,169,170,171,172,173,174,175,176],"openai","speech","ASR","Audio","text","model","speech recognition","use","general","multilingual","translation","english","identification","processing","Audio Processing",{"skillNo":178,"name":179,"nameZh":6,"description":180,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":181,"categories":182,"keywords":186,"publishedAt":43},"01M6GJSMRQ8EETZTYMG72P70EA","weights-and-biases","W&B: log ML experiments, sweeps, model registry, dashboards.","https:\u002F\u002Fgithub.com\u002Fzapabob\u002FSkills\u002Fblob\u002Fmain\u002Fregistry\u002Fskills\u002Fweights-and-biases\u002Fvariants\u002Fhermes-agent\u002FSKILL.md",[183],{"categoryNo":184,"categoryName":185,"taxonomyType":21,"taxonomyTypeDesc":22},"01JBSEEDCAT000000000000013","分析与可视化",[187,41,188,189,39,190,191,192,193,35,194,167,195,196,197,198,199,200,201,202],"weights and biases","tensorflow","experiments","pytorch","visualization","registry","real-time","dashboards","experiment","log","time","Experiment Tracking","Model Registry","Hyperparameter Tuning","WandB","Real-Time Visualization",{"skillNo":204,"name":205,"nameZh":6,"description":206,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":207,"categories":208,"keywords":210,"publishedAt":43},"01M6GJS19KDPRY112JTR326S06","watchers","Poll RSS, JSON APIs, and GitHub with watermark dedup.","https:\u002F\u002Fgithub.com\u002Fzapabob\u002FSkills\u002Fblob\u002Fmain\u002Fregistry\u002Fskills\u002Fwatchers\u002Fvariants\u002Fhermes-agent-optional\u002FSKILL.md",[209],{"categoryNo":97,"categoryName":98,"taxonomyType":21,"taxonomyTypeDesc":22},[205,41,211,107,40,212,213,214,215,216,217,114,115,218,219],"terminal","automation","http","monitoring","cron","json","rss","devops","with",{"skillNo":221,"name":222,"nameZh":6,"description":223,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":224,"categories":225,"keywords":228,"publishedAt":43},"01M6GJRHJD2A6Q830SWF9W48BC","serving-llms-vllm","vLLM: high-throughput LLM serving, OpenAI API, quantization.","https:\u002F\u002Fgithub.com\u002Fzapabob\u002FSkills\u002Fblob\u002Fmain\u002Fregistry\u002Fskills\u002Fvllm\u002Fvariants\u002Fhermes-agent\u002FSKILL.md",[226,227],{"categoryNo":94,"categoryName":95,"taxonomyType":21,"taxonomyTypeDesc":22},{"categoryNo":100,"categoryName":101,"taxonomyType":102,"taxonomyTypeDesc":103},[229,230,231,40,232,162,233,234,235,36,236,112,237,238,239,240,241,242,243,244],"llms","vLLM","serving llms vllm","research","openai api","tensor","Inference Serving","Production","llm serving","High Throughput","PagedAttention","Continuous Batching","Quantization","Tensor Parallelism","high-throughput","throughput"]