跳转到主要内容
此内容尚未提供您的语言版本,正在以英文显示。

Rwkv Architecture

技能 已验证 活跃

RNN+Transformer hybrid with O(n) inference. Linear time, infinite context, no KV cache. Train like GPT (parallel), infer like RNN (sequential). Linux Foundation AI project. Production at Windows, Office, NeMo. RWKV-7 (March 2025). Models up to 14B parameters.

目的

To empower AI agents to autonomously conduct AI research by providing a comprehensive library of specialized skills covering the entire research lifecycle, from ideation to paper writing.

功能

  • Comprehensive AI research skill library
  • Autoresearch orchestration layer
  • Expert-level guidance with real code examples
  • Detailed troubleshooting guides
  • Production-ready workflows

使用场景

  • Fine-tuning LLMs with custom data
  • Optimizing inference latency
  • Learning about transformer architectures
  • Scaling training to large GPU clusters

非目标

  • Replacing core LLM functionalities
  • Providing generic task automation
  • Offering basic tutorials without practical application

安装

请先添加 Marketplace

/plugin marketplace add Orchestra-Research/AI-Research-SKILLs
/plugin install AI-Research-SKILLs@ai-research-skills

质量评分

已验证
99 /100
1 day ago 分析

信任信号

最近提交17 days ago
星标8.3k
许可证MIT
状态
查看源代码

类似扩展

Rwkv Architecture

96

RNN+Transformer hybrid with O(n) inference. Linear time, infinite context, no KV cache. Train like GPT (parallel), infer like RNN (sequential). Linux Foundation AI project. Production at Windows, Office, NeMo. RWKV-7 (March 2025). Models up to 14B parameters.

技能
davila7

Mamba Architecture

99

State-space model with O(n) complexity vs Transformers' O(n²). 5× faster inference, million-token sequences, no KV cache. Selective SSM with hardware-aware design. Mamba-1 (d_state=16) and Mamba-2 (d_state=128, multi-head). Models 130M-2.8B on HuggingFace.

技能
Orchestra-Research

Mamba Architecture

95

State-space model with O(n) complexity vs Transformers' O(n²). 5× faster inference, million-token sequences, no KV cache. Selective SSM with hardware-aware design. Mamba-1 (d_state=16) and Mamba-2 (d_state=128, multi-head). Models 130M-2.8B on HuggingFace.

技能
davila7

TorchTitan Distributed LLM Pretraining

99

Provides PyTorch-native distributed LLM pretraining using torchtitan with 4D parallelism (FSDP2, TP, PP, CP). Use when pretraining Llama 3.1, DeepSeek V3, or custom models at scale from 8 to 512+ GPUs with Float8, torch.compile, and distributed checkpointing.

技能
Orchestra-Research

Implementing Llms Litgpt

98

Implements and trains LLMs using Lightning AI's LitGPT with 20+ pretrained architectures (Llama, Gemma, Phi, Qwen, Mistral). Use when need clean model implementations, educational understanding of architectures, or production fine-tuning with LoRA/QLoRA. Single-file implementations, no abstraction layers.

技能
Orchestra-Research

Distributed Llm Pretraining Torchtitan

98

Provides PyTorch-native distributed LLM pretraining using torchtitan with 4D parallelism (FSDP2, TP, PP, CP). Use when pretraining Llama 3.1, DeepSeek V3, or custom models at scale from 8 to 512+ GPUs with Float8, torch.compile, and distributed checkpointing.

技能
davila7