[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M6QF2CBYYHX94KMQA1QKX0AS":3,"skill-same:01M6QF2CBYYHX94KMQA1QKX0AS":83},{"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":54,"commitSha":55,"skillDirectory":56,"forksCount":14,"repositoryHtmlUrl":57,"repositoryHomepageUrl":58,"ownerHtmlUrl":59,"ownerBlogUrl":60,"ownerDescription":61,"updatedAt":62,"install":63,"resources":76},"01M6QF2CBYYHX94KMQA1QKX0AS","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.","Tashima-Tarsh\u002FDisha6.6","Tashima-Tarsh","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F269906498?v=4","01M6PBH98X03XERH0HFT4PFB1P","NITISH KUMAR","01M6PBH9DA0M6KM3Z12P848F0K",0,"MIT","https:\u002F\u002Fgithub.com\u002FTashima-Tarsh\u002FDisha6.6\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fkal-llava\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],"Vision-Language","language models","for","pillow","Instruction Tuning","Conversational AI","Vicuna","Image Chat","Visual Question Answering","VQA","image","image analysis","visual","chat","clip","use","question","multimodal","assistant","2026-11-10 10:12:30",false,"---\nname: llava\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.\nlicense: MIT\nmetadata:\n  version: 1.0.0\n  category: multimodal-and-emerging\n  maintainer: Kalaris Labs\n  tags: LLaVA, Vision-Language, Multimodal, Visual Question Answering, Image Chat, CLIP, Vicuna, Conversational AI, Instruction Tuning, VQA\n  dependencies: transformers, torch, pillow\n---\n\n# LLaVA - Large Language and Vision Assistant\n\nOpen-source vision-language model for conversational image understanding.\n\n## When to use LLaVA\n\n**Use when:**\n- Building vision-language chatbots\n- Visual question answering (VQA)\n- Image description and captioning\n- Multi-turn image conversations\n- Visual instruction following\n- Document understanding with images\n\n**Metrics**:\n- GPT-4V level capabilities (targeted)\n- Apache 2.0 License\n- Multiple model sizes (7B-34B params)\n\n**Use alternatives instead**:\n- **GPT-4V**: Highest quality, API-based\n- **CLIP**: Simple zero-shot classification\n- **BLIP-2**: Better for captioning only\n- **Flamingo**: Research, not open-source\n\n## Quick start\n\n### Installation\n\n```bash\n# Clone repository\ngit clone https:\u002F\u002Fgithub.com\u002Fhaotian-liu\u002FLLaVA\ncd LLaVA\n\n# Install\npip install -e .\n```\n\n### Basic usage\n\n```python\nfrom llava.model.builder import load_pretrained_model\nfrom llava.mm_utils import get_model_name_from_path, process_images, tokenizer_image_token\nfrom llava.constants import IMAGE_TOKEN_INDEX, DEFAULT_IMAGE_TOKEN\nfrom llava.conversation import conv_templates\nfrom PIL import Image\nimport torch\n\n# Load model\nmodel_path = \"liuhaotian\u002Fllava-v1.5-7b\"\ntokenizer, model, image_processor, context_len = load_pretrained_model(\n    model_path=model_path,\n    model_base=None,\n    model_name=get_model_name_from_path(model_path)\n)\n\n# Load image\nimage = Image.open(\"image.jpg\")\nimage_tensor = process_images([image], image_processor, model.config)\nimage_tensor = image_tensor.to(model.device, dtype=torch.float16)\n\n# Create conversation\nconv = conv_templates[\"llava_v1\"].copy()\nconv.append_message(conv.roles[0], DEFAULT_IMAGE_TOKEN + \"\\nWhat is in this image?\")\nconv.append_message(conv.roles[1], None)\nprompt = conv.get_prompt()\n\n# Generate response\ninput_ids = tokenizer_image_token(prompt, tokenizer, IMAGE_TOKEN_INDEX, return_tensors='pt').unsqueeze(0).to(model.device)\n\nwith torch.inference_mode():\n    output_ids = model.generate(\n        input_ids,\n        images=image_tensor,\n        do_sample=True,\n        temperature=0.2,\n        max_new_tokens=512\n    )\n\nresponse = tokenizer.decode(output_ids[0], skip_special_tokens=True).strip()\nprint(response)\n```\n\n## Available models\n\n| Model | Parameters | VRAM | Quality |\n|-------|------------|------|---------|\n| LLaVA-v1.5-7B | 7B | ~14 GB | Good |\n| LLaVA-v1.5-13B | 13B | ~28 GB | Better |\n| LLaVA-v1.6-34B | 34B | ~70 GB | Best |\n\n```python\n# Load different models\nmodel_7b = \"liuhaotian\u002Fllava-v1.5-7b\"\nmodel_13b = \"liuhaotian\u002Fllava-v1.5-13b\"\nmodel_34b = \"liuhaotian\u002Fllava-v1.6-34b\"\n\n# 4-bit quantization for lower VRAM\nload_4bit = True  # Reduces VRAM by ~4×\n```\n\n## CLI usage\n\n```bash\n# Single image query\npython -m llava.serve.cli \\\n    --model-path liuhaotian\u002Fllava-v1.5-7b \\\n    --image-file image.jpg \\\n    --query \"What is in this image?\"\n\n# Multi-turn conversation\npython -m llava.serve.cli \\\n    --model-path liuhaotian\u002Fllava-v1.5-7b \\\n    --image-file image.jpg\n# Then type questions interactively\n```\n\n## Web UI (Gradio)\n\n```bash\n# Launch Gradio interface\npython -m llava.serve.gradio_web_server \\\n    --model-path liuhaotian\u002Fllava-v1.5-7b \\\n    --load-4bit  # Optional: reduce VRAM\n\n# Access at http:\u002F\u002Flocalhost:7860\n```\n\n## Multi-turn conversations\n\n```python\n# Initialize conversation\nconv = conv_templates[\"llava_v1\"].copy()\n\n# Turn 1\nconv.append_message(conv.roles[0], DEFAULT_IMAGE_TOKEN + \"\\nWhat is in this image?\")\nconv.append_message(conv.roles[1], None)\nresponse1 = generate(conv, model, image)  # \"A dog playing in a park\"\n\n# Turn 2\nconv.messages[-1][1] = response1  # Add previous response\nconv.append_message(conv.roles[0], \"What breed is the dog?\")\nconv.append_message(conv.roles[1], None)\nresponse2 = generate(conv, model, image)  # \"Golden Retriever\"\n\n# Turn 3\nconv.messages[-1][1] = response2\nconv.append_message(conv.roles[0], \"What time of day is it?\")\nconv.append_message(conv.roles[1], None)\nresponse3 = generate(conv, model, image)\n```\n\n## Common tasks\n\n### Image captioning\n\n```python\nquestion = \"Describe this image in detail.\"\nresponse = ask(model, image, question)\n```\n\n### Visual question answering\n\n```python\nquestion = \"How many people are in the image?\"\nresponse = ask(model, image, question)\n```\n\n### Object detection (textual)\n\n```python\nquestion = \"List all the objects you can see in this image.\"\nresponse = ask(model, image, question)\n```\n\n### Scene understanding\n\n```python\nquestion = \"What is happening in this scene?\"\nresponse = ask(model, image, question)\n```\n\n### Document understanding\n\n```python\nquestion = \"What is the main topic of this document?\"\nresponse = ask(model, document_image, question)\n```\n\n## Training custom model\n\n```bash\n# Stage 1: Feature alignment (558K image-caption pairs)\nbash scripts\u002Fv1_5\u002Fpretrain.sh\n\n# Stage 2: Visual instruction tuning (150K instruction data)\nbash scripts\u002Fv1_5\u002Ffinetune.sh\n```\n\n## Quantization (reduce VRAM)\n\n```python\n# 4-bit quantization\ntokenizer, model, image_processor, context_len = load_pretrained_model(\n    model_path=\"liuhaotian\u002Fllava-v1.5-13b\",\n    model_base=None,\n    model_name=get_model_name_from_path(\"liuhaotian\u002Fllava-v1.5-13b\"),\n    load_4bit=True  # Reduces VRAM ~4×\n)\n\n# 8-bit quantization\nload_8bit=True  # Reduces VRAM ~2×\n```\n\n## Best practices\n\n1. **Start with 7B model** - Good quality, manageable VRAM\n2. **Use 4-bit quantization** - Reduces VRAM significantly\n3. **GPU required** - CPU inference extremely slow\n4. **Clear prompts** - Specific questions get better answers\n5. **Multi-turn conversations** - Maintain conversation context\n6. **Temperature 0.2-0.7** - Balance creativity\u002Fconsistency\n7. **max_new_tokens 512-1024** - For detailed responses\n8. **Batch processing** - Process multiple images sequentially\n\n## Performance\n\n| Model | VRAM (FP16) | VRAM (4-bit) | Speed (tokens\u002Fs) |\n|-------|-------------|--------------|------------------|\n| 7B | ~14 GB | ~4 GB | ~20 |\n| 13B | ~28 GB | ~8 GB | ~12 |\n| 34B | ~70 GB | ~18 GB | ~5 |\n\n*On A100 GPU*\n\n## Benchmarks\n\nLLaVA achieves competitive scores on:\n- **VQAv2**: 78.5%\n- **GQA**: 62.0%\n- **MM-Vet**: 35.4%\n- **MMBench**: 64.3%\n\n## Limitations\n\n1. **Hallucinations** - May describe things not in image\n2. **Spatial reasoning** - Struggles with precise locations\n3. **Small text** - Difficulty reading fine print\n4. **Object counting** - Imprecise for many objects\n5. **VRAM requirements** - Need powerful GPU\n6. **Inference speed** - Slower than CLIP\n\n## Integration with frameworks\n\n### LangChain\n\n```python\nfrom langchain.llms.base import LLM\n\nclass LLaVALLM(LLM):\n    def _call(self, prompt, stop=None):\n        # Custom LLaVA inference\n        return response\n\nllm = LLaVALLM()\n```\n\n### Gradio App\n\n```python\nimport gradio as gr\n\ndef chat(image, text, history):\n    response = ask_llava(model, image, text)\n    return response\n\ndemo = gr.ChatInterface(\n    chat,\n    additional_inputs=[gr.Image(type=\"pil\")],\n    title=\"LLaVA Chat\"\n)\ndemo.launch()\n```\n\n## Resources\n\n- **GitHub**: https:\u002F\u002Fgithub.com\u002Fhaotian-liu\u002FLLaVA ⭐ 23,000+\n- **Paper**: https:\u002F\u002Farxiv.org\u002Fabs\u002F2304.08485\n- **Demo**: https:\u002F\u002Fllava.hliu.cc\n- **Models**: https:\u002F\u002Fhuggingface.co\u002Fliuhaotian\n- **License**: Apache 2.0\n\n## Agent operating procedure\n\n1. **Check the environment.** Confirm GPU memory, model checkpoint and licence, framework versions and input formats.\n2. **Pin down the inputs.** Confirm formats, identifiers and parameters from the data or the user. Ask rather than guess any value that changes the result.\n3. **Run a small version first.** Run inference on a single sample at low resolution or short length first.\n4. **Execute the full task** using the instructions and references above.\n5. **Validate the result.** Inspect outputs qualitatively and with task metrics; check licence restrictions for generated content.\n6. **Report.** State what was run (versions, commands, parameters), what was checked, and what is still uncertain.\n\n| If this happens | Do this |\n|---|---|\n| Out-of-memory on large inputs | Reduce resolution or sequence length, use half precision, or tile the input. |\n| A function, flag or endpoint in these instructions is missing in the installed version | Check the installed version's own documentation (`help()`, `--help`, official docs), adapt, and tell the user. Never invent an API. |\n| A required input, identifier or parameter is ambiguous | Ask the user, or state the assumption explicitly before running. |\n\n**Integrity rules**\n\n- Never fabricate results, parameters, identifiers, citations or statistics. If something cannot be run or verified, say so plainly.\n- Check model and dataset licences before commercial or published use.\n- Treat version-specific details here as possibly outdated: confirm them against the official documentation for the installed version.\n- Ask before actions that cost money, consume shared GPUs or cloud quota, touch personal or patient data, or cannot be undone.\n\n## Related skills\n\n- `clip`: OpenAI's model connecting vision and language.\n- `blip-2-vision-language`: Explains how to use Salesforce BLIP-2 (Q-Former bridging a frozen image encoder and an LLM such as OPT or FlanT5) through HuggingFace Trans…\n- `fine-tuning-with-trl`: Fine-tune LLMs using reinforcement learning with TRL - SFT for instruction tuning, DPO for preference alignment, PPO\u002FGRPO for reward optimi…\n","# LLaVA - Large Language and Vision Assistant\n\nOpen-source vision-language model for conversational image understanding.\n\n## When to use LLaVA\n\n**Use when:**\n- Building vision-language chatbots\n- Visual question answering (VQA)\n- Image description and captioning\n- Multi-turn image conversations\n- Visual instruction following\n- Document understanding with images\n\n**Metrics**:\n- GPT-4V level capabilities (targeted)\n- Apache 2.0 License\n- Multiple model sizes (7B-34B params)\n\n**Use alternatives instead**:\n- **GPT-4V**: Highest quality, API-based\n- **CLIP**: Simple zero-shot classification\n- **BLIP-2**: Better for captioning only\n- **Flamingo**: Research, not open-source\n\n## Quick start\n\n### Installation\n\n```bash\n# Clone repository\ngit clone https:\u002F\u002Fgithub.com\u002Fhaotian-liu\u002FLLaVA\ncd LLaVA\n\n# Install\npip install -e .\n```\n\n### Basic usage\n\n```python\nfrom llava.model.builder import load_pretrained_model\nfrom llava.mm_utils import get_model_name_from_path, process_images, tokenizer_image_token\nfrom llava.constants import IMAGE_TOKEN_INDEX, DEFAULT_IMAGE_TOKEN\nfrom llava.conversation import conv_templates\nfrom PIL import Image\nimport torch\n\n# Load model\nmodel_path = \"liuhaotian\u002Fllava-v1.5-7b\"\ntokenizer, model, image_processor, context_len = load_pretrained_model(\n    model_path=model_path,\n    model_base=None,\n    model_name=get_model_name_from_path(model_path)\n)\n\n# Load image\nimage = Image.open(\"image.jpg\")\nimage_tensor = process_images([image], image_processor, model.config)\nimage_tensor = image_tensor.to(model.device, dtype=torch.float16)\n\n# Create conversation\nconv = conv_templates[\"llava_v1\"].copy()\nconv.append_message(conv.roles[0], DEFAULT_IMAGE_TOKEN + \"\\nWhat is in this image?\")\nconv.append_message(conv.roles[1], None)\nprompt = conv.get_prompt()\n\n# Generate response\ninput_ids = tokenizer_image_token(prompt, tokenizer, IMAGE_TOKEN_INDEX, return_tensors='pt').unsqueeze(0).to(model.device)\n\nwith torch.inference_mode():\n    output_ids = model.generate(\n        input_ids,\n        images=image_tensor,\n        do_sample=True,\n        temperature=0.2,\n        max_new_tokens=512\n    )\n\nresponse = tokenizer.decode(output_ids[0], skip_special_tokens=True).strip()\nprint(response)\n```\n\n## Available models\n\n| Model | Parameters | VRAM | Quality |\n|-------|------------|------|---------|\n| LLaVA-v1.5-7B | 7B | ~14 GB | Good |\n| LLaVA-v1.5-13B | 13B | ~28 GB | Better |\n| LLaVA-v1.6-34B | 34B | ~70 GB | Best |\n\n```python\n# Load different models\nmodel_7b = \"liuhaotian\u002Fllava-v1.5-7b\"\nmodel_13b = \"liuhaotian\u002Fllava-v1.5-13b\"\nmodel_34b = \"liuhaotian\u002Fllava-v1.6-34b\"\n\n# 4-bit quantization for lower VRAM\nload_4bit = True  # Reduces VRAM by ~4×\n```\n\n## CLI usage\n\n```bash\n# Single image query\npython -m llava.serve.cli \\\n    --model-path liuhaotian\u002Fllava-v1.5-7b \\\n    --image-file image.jpg \\\n    --query \"What is in this image?\"\n\n# Multi-turn conversation\npython -m llava.serve.cli \\\n    --model-path liuhaotian\u002Fllava-v1.5-7b \\\n    --image-file image.jpg\n# Then type questions interactively\n```\n\n## Web UI (Gradio)\n\n```bash\n# Launch Gradio interface\npython -m llava.serve.gradio_web_server \\\n    --model-path liuhaotian\u002Fllava-v1.5-7b \\\n    --load-4bit  # Optional: reduce VRAM\n\n# Access at http:\u002F\u002Flocalhost:7860\n```\n\n## Multi-turn conversations\n\n```python\n# Initialize conversation\nconv = conv_templates[\"llava_v1\"].copy()\n\n# Turn 1\nconv.append_message(conv.roles[0], DEFAULT_IMAGE_TOKEN + \"\\nWhat is in this image?\")\nconv.append_message(conv.roles[1], None)\nresponse1 = generate(conv, model, image)  # \"A dog playing in a park\"\n\n# Turn 2\nconv.messages[-1][1] = response1  # Add previous response\nconv.append_message(conv.roles[0], \"What breed is the dog?\")\nconv.append_message(conv.roles[1], None)\nresponse2 = generate(conv, model, image)  # \"Golden Retriever\"\n\n# Turn 3\nconv.messages[-1][1] = response2\nconv.append_message(conv.roles[0], \"What time of day is it?\")\nconv.append_message(conv.roles[1], None)\nresponse3 = generate(conv, model, image)\n```\n\n## Common tasks\n\n### Image captioning\n\n```python\nquestion = \"Describe this image in detail.\"\nresponse = ask(model, image, question)\n```\n\n### Visual question answering\n\n```python\nquestion = \"How many people are in the image?\"\nresponse = ask(model, image, question)\n```\n\n### Object detection (textual)\n\n```python\nquestion = \"List all the objects you can see in this image.\"\nresponse = ask(model, image, question)\n```\n\n### Scene understanding\n\n```python\nquestion = \"What is happening in this scene?\"\nresponse = ask(model, image, question)\n```\n\n### Document understanding\n\n```python\nquestion = \"What is the main topic of this document?\"\nresponse = ask(model, document_image, question)\n```\n\n## Training custom model\n\n```bash\n# Stage 1: Feature alignment (558K image-caption pairs)\nbash scripts\u002Fv1_5\u002Fpretrain.sh\n\n# Stage 2: Visual instruction tuning (150K instruction data)\nbash scripts\u002Fv1_5\u002Ffinetune.sh\n```\n\n## Quantization (reduce VRAM)\n\n```python\n# 4-bit quantization\ntokenizer, model, image_processor, context_len = load_pretrained_model(\n    model_path=\"liuhaotian\u002Fllava-v1.5-13b\",\n    model_base=None,\n    model_name=get_model_name_from_path(\"liuhaotian\u002Fllava-v1.5-13b\"),\n    load_4bit=True  # Reduces VRAM ~4×\n)\n\n# 8-bit quantization\nload_8bit=True  # Reduces VRAM ~2×\n```\n\n## Best practices\n\n1. **Start with 7B model** - Good quality, manageable VRAM\n2. **Use 4-bit quantization** - Reduces VRAM significantly\n3. **GPU required** - CPU inference extremely slow\n4. **Clear prompts** - Specific questions get better answers\n5. **Multi-turn conversations** - Maintain conversation context\n6. **Temperature 0.2-0.7** - Balance creativity\u002Fconsistency\n7. **max_new_tokens 512-1024** - For detailed responses\n8. **Batch processing** - Process multiple images sequentially\n\n## Performance\n\n| Model | VRAM (FP16) | VRAM (4-bit) | Speed (tokens\u002Fs) |\n|-------|-------------|--------------|------------------|\n| 7B | ~14 GB | ~4 GB | ~20 |\n| 13B | ~28 GB | ~8 GB | ~12 |\n| 34B | ~70 GB | ~18 GB | ~5 |\n\n*On A100 GPU*\n\n## Benchmarks\n\nLLaVA achieves competitive scores on:\n- **VQAv2**: 78.5%\n- **GQA**: 62.0%\n- **MM-Vet**: 35.4%\n- **MMBench**: 64.3%\n\n## Limitations\n\n1. **Hallucinations** - May describe things not in image\n2. **Spatial reasoning** - Struggles with precise locations\n3. **Small text** - Difficulty reading fine print\n4. **Object counting** - Imprecise for many objects\n5. **VRAM requirements** - Need powerful GPU\n6. **Inference speed** - Slower than CLIP\n\n## Integration with frameworks\n\n### LangChain\n\n```python\nfrom langchain.llms.base import LLM\n\nclass LLaVALLM(LLM):\n    def _call(self, prompt, stop=None):\n        # Custom LLaVA inference\n        return response\n\nllm = LLaVALLM()\n```\n\n### Gradio App\n\n```python\nimport gradio as gr\n\ndef chat(image, text, history):\n    response = ask_llava(model, image, text)\n    return response\n\ndemo = gr.ChatInterface(\n    chat,\n    additional_inputs=[gr.Image(type=\"pil\")],\n    title=\"LLaVA Chat\"\n)\ndemo.launch()\n```\n\n## Resources\n\n- **GitHub**: https:\u002F\u002Fgithub.com\u002Fhaotian-liu\u002FLLaVA ⭐ 23,000+\n- **Paper**: https:\u002F\u002Farxiv.org\u002Fabs\u002F2304.08485\n- **Demo**: https:\u002F\u002Fllava.hliu.cc\n- **Models**: https:\u002F\u002Fhuggingface.co\u002Fliuhaotian\n- **License**: Apache 2.0\n\n## Agent operating procedure\n\n1. **Check the environment.** Confirm GPU memory, model checkpoint and licence, framework versions and input formats.\n2. **Pin down the inputs.** Confirm formats, identifiers and parameters from the data or the user. Ask rather than guess any value that changes the result.\n3. **Run a small version first.** Run inference on a single sample at low resolution or short length first.\n4. **Execute the full task** using the instructions and references above.\n5. **Validate the result.** Inspect outputs qualitatively and with task metrics; check licence restrictions for generated content.\n6. **Report.** State what was run (versions, commands, parameters), what was checked, and what is still uncertain.\n\n| If this happens | Do this |\n|---|---|\n| Out-of-memory on large inputs | Reduce resolution or sequence length, use half precision, or tile the input. |\n| A function, flag or endpoint in these instructions is missing in the installed version | Check the installed version's own documentation (`help()`, `--help`, official docs), adapt, and tell the user. Never invent an API. |\n| A required input, identifier or parameter is ambiguous | Ask the user, or state the assumption explicitly before running. |\n\n**Integrity rules**\n\n- Never fabricate results, parameters, identifiers, citations or statistics. If something cannot be run or verified, say so plainly.\n- Check model and dataset licences before commercial or published use.\n- Treat version-specific details here as possibly outdated: confirm them against the official documentation for the installed version.\n- Ask before actions that cost money, consume shared GPUs or cloud quota, touch personal or patient data, or cannot be undone.\n\n## Related skills\n\n- `clip`: OpenAI's model connecting vision and language.\n- `blip-2-vision-language`: Explains how to use Salesforce BLIP-2 (Q-Former bridging a frozen image encoder and an LLM such as OPT or FlanT5) through HuggingFace Trans…\n- `fine-tuning-with-trl`: Fine-tune LLMs using reinforcement learning with TRL - SFT for instruction tuning, DPO for preference alignment, PPO\u002FGRPO for reward optimi…\n",{"name":5,"license":15,"metadata":48,"description":7},{"tags":49,"version":50,"category":51,"maintainer":52,"dependencies":53},"LLaVA, Vision-Language, Multimodal, Visual Question Answering, Image Chat, CLIP, Vicuna, Conversational AI, Instruction Tuning, VQA","1.0.0","multimodal-and-emerging","Kalaris Labs","transformers, torch, pillow","main","0da4485785ee599388900be50db1635489d53056",".agents\u002Fskills\u002Fkal-llava","https:\u002F\u002Fgithub.com\u002FTashima-Tarsh\u002FDisha6.6","https:\u002F\u002Fdisha6.6.thenitishkr.in\u002F","https:\u002F\u002Fgithub.com\u002FTashima-Tarsh","www.thenitishkr","thenitishkr | AI • Cyber-security • Research\r\n","2026-11-08 03:33:25",{"source":8,"skillName":5,"commands":64,"prompt":74,"zipAvailable":75},[65,68,71],{"tool":66,"command":67},"npx","npx skills add Tashima-Tarsh\u002FDisha6.6 --skill llava",{"tool":69,"command":70},"bunx","bunx skills add Tashima-Tarsh\u002FDisha6.6 --skill llava",{"tool":72,"command":73},"pnpm","pnpm dlx skills add Tashima-Tarsh\u002FDisha6.6 --skill llava","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FTashima-Tarsh\u002FDisha6.6\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fkal-llava\u002FSKILL.md",true,[77],{"path":78,"resourceType":79,"resourceTypeDesc":80,"sizeBytes":81,"sourceHtmlUrl":82},"references\u002Ftraining.md","REFERENCE","参考文档",4738,"https:\u002F\u002Fgithub.com\u002FTashima-Tarsh\u002FDisha6.6\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fkal-llava\u002Freferences\u002Ftraining.md",[84,111,135,162,187,211],{"skillNo":85,"name":86,"nameZh":6,"description":87,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":88,"categories":89,"keywords":91,"publishedAt":43},"01M6QRD07D4TGNDK6VJ4EHF8BV","youtube-full","Use when the user needs YouTube transcripts, video search, channel browsing, playlist extraction, or content monitoring. Trigger phrases: 'get the transcript for', 'search YouTube for', 'what are the latest videos on', 'list this playlist', 'monitor this channel', or any request involving a YouTube URL, video ID, or @handle. Do NOT use for downloading video or audio files, YouTube engagement data (likes, comments), or private\u002Fage-restricted videos.","https:\u002F\u002Fgithub.com\u002FTashima-Tarsh\u002FDisha6.6\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fyoutube-full\u002FSKILL.md",[90],{"categoryNo":19,"categoryName":20,"taxonomyType":21,"taxonomyTypeDesc":22},[92,93,94,95,96,97,98,99,100,101,102,103,104,105,106,107,26,108,109,110],"youtube full","youtube","full","transcript","search","monitoring","data","Video","Audio","list","files","do","get","age","trigger","URL","browsing","request","youtube transcripts",{"skillNo":112,"name":113,"nameZh":6,"description":114,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":115,"categories":116,"keywords":117,"publishedAt":43},"01M6QRCJ3112VYPA97YG7KGMC2","x-twitter-growth","X\u002FTwitter growth engine for building audience, crafting viral content, and analyzing engagement. Use when the user wants to grow on X\u002FTwitter, write tweets or threads, analyze their X profile, research competitors on X, plan a posting strategy, or optimize engagement. Complements social-content (generic multi-platform) with X-specific depth: algorithm mechanics, thread engineering, reply strategy, profile optimization, and competitive intelligence via web search.","https:\u002F\u002Fgithub.com\u002FTashima-Tarsh\u002FDisha6.6\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fx-twitter-growth\u002FSKILL.md",[],[118,119,120,121,96,122,123,124,125,126,127,128,129,130,39,131,132,133,134,26],"twitter","x twitter growth","growth","research","content","engineering","plan","web search","competitors","analyze","engine","write","strategy","platform","grow","posting","competitive",{"skillNo":136,"name":137,"nameZh":6,"description":138,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":139,"categories":140,"keywords":144,"publishedAt":43},"01M6QRBN3EHQNJY42RERP7A149","write-a-skill","Create new agent skills with proper structure, progressive disclosure, and bundled resources. Use when user wants to create, write, build, or author a new skill.","https:\u002F\u002Fgithub.com\u002FTashima-Tarsh\u002FDisha6.6\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fwrite-a-skill\u002FSKILL.md",[141],{"categoryNo":142,"categoryName":143,"taxonomyType":21,"taxonomyTypeDesc":22},"01JBSEEDCAT000000000000023","Skill与智能体开发",[145,129,146,147,148,149,150,151,39,152,153,54,154,155,156,157,158,159,160,161],"write a skill","github","create","https","productivity","matt pocock","direct","example","build","new skill","structure","new agent","resources","with","tree","progressive disclosure","author",{"skillNo":163,"name":164,"nameZh":6,"description":165,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":166,"categories":167,"keywords":168,"publishedAt":43},"01M6QRB6G8P33PNGAF9Y69EJ9T","wrap-up","Close out a launched Claude Managed Agent — recap every primitive the founder now owns, regenerate the single-file overview page, and suggest the next 1-2 upgrades. Use when the user says \"wrap up\", \"close this out\", \"what do I own now\", \"give me the summary\", \"recap the agent\", or when the orchestrator routes phase=wrap-up. primitives_inventory.py tables everything owned (agent, environment, session, memory, outcome, deployment); overview_page.py regenerates a self-contained .\u002Fmy-agent\u002Fagent-overview.html; upgrade_suggester.py ranks the next moves from recorded deferrals plus standing hardening steps. Companion to run-without-you; the last stop before phase=done.","https:\u002F\u002Fgithub.com\u002FTashima-Tarsh\u002FDisha6.6\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fwrap-up\u002FSKILL.md",[],[169,170,171,172,173,174,175,176,177,39,178,103,179,180,181,182,183,184,185,186],"wrap up","up","wrap","gemini","deployment","run","file","hardening","closeout","stop","upgrade","founder","close","companion","you","tables","directions","phase",{"skillNo":188,"name":189,"nameZh":6,"description":190,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":191,"categories":192,"keywords":194,"publishedAt":43},"01M6QRA5AYC1RBFZ39GTWGS3N8","workflow-builder","Design and write deterministic multi-agent workflow scripts (.js files in .claude\u002Fworkflows\u002F) for Claude Code's Workflow tool. Use when a user wants to build, create, author, scaffold, or run a custom Claude Code workflow, orchestrate sub-agents (fan-out, pipeline, loop, judge-panel), or automate a repeatable multi-step task across fresh-context agents.","https:\u002F\u002Fgithub.com\u002FTashima-Tarsh\u002FDisha6.6\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fworkflow-builder\u002FSKILL.md",[193],{"categoryNo":142,"categoryName":143,"taxonomyType":21,"taxonomyTypeDesc":22},[195,196,197,198,199,147,174,200,201,202,203,204,129,205,102,206,207,208,209,210],"workflow","workflow builder","builder","design","workflows","code","orchestrate","pipeline","loop","multi-agent","context","across","fresh","scripts","judge","claude code workflow",{"skillNo":212,"name":213,"nameZh":6,"description":214,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":215,"categories":216,"keywords":217,"publishedAt":43},"01M6QR7QMCZESKQA0ERW1E657M","weekly-review","Use when someone wants to run a weekly review, close open loops, audit stalled projects and commitments, get their system back to trusted, restart a lapsed review habit, or says \"\u002Fcs:weekly-review\". Walks David Allen's three-phase loop — GET CLEAR, GET CURRENT, GET CREATIVE — with deterministic scripts that inventory open loops, gate the checklist with named gaps, and score commitment health 0-100.","https:\u002F\u002Fgithub.com\u002FTashima-Tarsh\u002FDisha6.6\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Fweekly-review\u002FSKILL.md",[],[218,219,220,221,222,223,224,225,226,203,227,228,229,230,231,232,104,233,234,186],"review","weekly review","weekly","deterministic scripts","reflect","feeds","inventory","open","team","projects","project-management","scanner","conversation","back","project","sprint","cadence"]