amd
/

DeepSeek-R1-MXFP4

8-bit precision

Model card Files Files and versions

ColinZ22 commited on 5 days ago

Commit

a7e208e

·

verified ·

1 Parent(s): d31023d

Update README.md

Files changed (1) hide show

README.md +1 -2

README.md CHANGED Viewed

@@ -102,8 +102,7 @@ python3 -m sglang.launch_server \
     --model amd/DeepSeek-R1-MXFP4 \
     --tp 8  \
     --trust-remote-code  \
-    --n-share-experts-fusion 8 \
-    --disable-radix-cache
 # Evaluating
 lm_eval --model local-completions \

     --model amd/DeepSeek-R1-MXFP4 \
     --tp 8  \
     --trust-remote-code  \
+    --attention-backend aiter
 # Evaluating
 lm_eval --model local-completions \