Update comment

sstamenk · sstamenk · commit 0f0fa5de2ab4 · 2025-11-16T22:39:07.000+01:00
Signed-off-by: sstamenk &lt;strahinja.stamenkovic@amd.com&gt;
diff --git a/vllm/platforms/rocm.py b/vllm/platforms/rocm.py
@@ -185,7 +185,7 @@ class RocmPlatform(Platform):
         "petit_nvfp4",
         "torchao",
     ]
-    # bitsandbytes is not supported on GPUs with warp size 64 (gfx9)
+    # bitsandbytes quantization not supported on Instinct (warp size 64 limitation)
     if not on_gfx9():
         supported_quantization += ["bitsandbytes"]
 

Original file line number	Diff line number	Diff line change
`@@ -185,7 +185,7 @@ class RocmPlatform(Platform):`
`185`	`185`	`"petit_nvfp4",`
`186`	`186`	`"torchao",`
`187`	`187`	`]`
`188`		`- # bitsandbytes is not supported on GPUs with warp size 64 (gfx9)`
	`188`	`+ # bitsandbytes quantization not supported on Instinct (warp size 64 limitation)`
`189`	`189`	`if not on_gfx9():`
`190`	`190`	`supported_quantization += ["bitsandbytes"]`
`191`	`191`