Reduce CTX_CHECKPOINTS from 32 to 16 in env file to free ~16 GiB VRAM and prevent OOM crashes
This commit is contained in:
1 parent
8c391fe120
commit
9fd82dc6e1
1 file changed
+1
-1
@@ -52,7 +52,7 @@ BATCH_SIZE=2048
|
|||||||
UBATCH_SIZE=2048
|
UBATCH_SIZE=2048
|
||||||
CACHE_TYPE_K=f16
|
CACHE_TYPE_K=f16
|
||||||
CACHE_TYPE_V=f16
|
CACHE_TYPE_V=f16
|
||||||
CTX_CHECKPOINTS=32
|
CTX_CHECKPOINTS=16
|
||||||
CACHE_RAM=8192
|
CACHE_RAM=8192
|
||||||
|
|
||||||
# Native MTP draft (depth-4, greedy, ROCm-safe CPU argmax fast path).
|
# Native MTP draft (depth-4, greedy, ROCm-safe CPU argmax fast path).
|
||||||
|
|||||||
Reference in new issue
Block a user