update README

README-CN.md

@@ -6,21 +6,25 @@ Tensorlink 是一款基于CUDA API-Forwarding实现的分布式计算框架，
 
 # 示例
 PS: 场景中展示的系统中没有实体显卡，使用tensolink链接处于另一个子网下的4090显卡
-
-## 场景一：Stable Diffusion 使用远端显卡加速
+## 场景一：Comfy-UI 使用远端显卡加速
+图中展示了Comfy-UI使用tensorlink进行计算的效果
+![alt text](assets/6.gif)
+## 场景二：Stable Diffusion 使用远端显卡加速
 图中展示了SD使用tensorlink进行计算的效果
 ![alt text](assets/3.gif)
-## 场景二：Transformer LLM 使用远端显卡推理
+## 场景三：Transformer LLM 使用远端显卡推理
 图中展示了trnasformers框架使用tensorlink进行模型推理的效果
 ![alt text](assets/4.gif)
-## 场景三：Cinma4D Octane插件 使用远端显卡渲染
+## 场景四：Cinma4D Octane插件 使用远端显卡渲染
 ![alt text](assets/5.gif)
 # 路线图
 
 - CUDA Runtime API Hook ✅
 - CUDA Driver API Hook ✅
 - CUDA cuBLAS Hook ✅
 - CUDA cuDNN Hook ✅
+- 支持缓存传输数据，提升加载速度 ✅
+- 释放进入空闲状态的显存，支持按需加载module ✅
 - 支持客户端多进程 ✅
 - 支持ZSTD数据压缩 ✅
 - 支持Light(TCP+UDP)与Native(TCP+TCP)两种通信协议 ✅

README.md

@@ -8,20 +8,24 @@ Tensorlink is a distributed computing framework based on CUDA API-Forwarding. Wh
 
 Note: The system shown in the scenarios does not have a physical GPU. It uses Tensorlink to connect to a 4090 GPU in another subnet.
 
+## Scenario 1: Comfy-UI Accelerated by Remote GPU
 
-## Scenario 1: Stable Diffusion Accelerated by Remote GPU
+The image shows the effect of Comfy-UI using Tensorlink for computation.
+![alt text](assets/6.gif)
+
+## Scenario 2: Stable Diffusion Accelerated by Remote GPU
 
 The image shows the effect of SD using Tensorlink for computation.
 ![alt text](assets/3.gif)
 
 
-## Scenario 2: Transformer LLM Inference Using Remote GPU
+## Scenario 3: Transformer LLM Inference Using Remote GPU
 
 The image shows the effect of the transformers framework using Tensorlink for model inference.
 ![alt text](assets/4.gif)
 
 
-## Scenario 3: Cinema4D Octane Plugin Remote GPU Rendering
+## Scenario 4: Cinema4D Octane Plugin Remote GPU Rendering
 ![alt text](assets/5.gif)
 
 # Roadmap
@@ -30,6 +34,8 @@ The image shows the effect of the transformers framework using Tensorlink for mo
 - CUDA Driver API Hook ✅
 - CUDA cuBLAS Hook ✅
 - CUDA cuDNN Hook ✅
+- Support caching transmitted data to enhance loading speed ✅
+- Release memory that has entered idle state, supporting on-demand module loading. ✅
 - Support for Client Multi-Process ✅
 - Support for ZSTD Data Compression ✅
 - Support for Light (TCP+UDP) and Native (TCP+TCP) Communication Protocols ✅