Revise rfork blog images to PNG format (#272)

Signed-off-by: Anqi Shen <[email protected]>

Author: amysaq2023

blog/2025-12-10-rfork.md

@@ -2,7 +2,7 @@
 title: "Let Tensors Fly — Accelerating Large Model Weight Loading with R-Fork"
 author: "Ant Group DeepXPU Team, SGLang Team"
 date: "December 10, 2025"
-previewImg: /images/blog/rfork/preview.svg
+previewImg: /images/blog/rfork/preview.png
 ---
 
 ## TL;DR
@@ -45,7 +45,7 @@ From the data flow analysis, we observe that weight tensors are stored on each G
 To maximize the utilization of InfiniBand NIC's bandwidth, we design a per GPU-pair data transfer strategy: a local GPU directly transfers data to/from its paired remote GPU. This design effectively bypasses the PCIe bottleneck between GPU and CPU, enabling high-throughput communication without relying on CPU or host memory.
 The data flow of loading weights from remote SGLang instance is the following:  
 
-<img src="/images/blog/rfork/design.svg" style="display:block; margin-top: auto; margin-left: auto; margin-right: auto; margin-bottom: auto; width: 100%; image-orientation: none;"></img>  
+<img src="/images/blog/rfork/design.png" style="display:block; margin-top: auto; margin-left: auto; margin-right: auto; margin-bottom: auto; width: 100%; image-orientation: none;"></img>  
 
 
 ## Implementation
@@ -117,7 +117,7 @@ python -m sglang.launch_server [args] \
 
 We evaluated the performance of launching a new SGLang instance equipped with eight NVIDIA H20 GPUs, while loading the DeepSeek-R1 model from different sources.  
 
-<img src="/images/blog/rfork/performance.svg" style="display:block; margin-top: auto; margin-left: auto; margin-right: auto; margin-bottom: auto; width: 100%; image-orientation: none;"></img>  
+<img src="/images/blog/rfork/performance.png" style="display:block; margin-top: auto; margin-left: auto; margin-right: auto; margin-bottom: auto; width: 100%; image-orientation: none;"></img>  
 
 Registering the memory region can be overlapped with other initialization phases to further optimize total boot-up time.