summaryrefslogtreecommitdiff
diff options
context:
space:
mode:
authorLaurent Mazare <laurent.mazare@gmail.com>2023-10-23 10:26:56 +0100
committerGitHub <noreply@github.com>2023-10-23 10:26:56 +0100
commit25c3cc4149304a4f6eec93b2f88aa9c241f8f696 (patch)
treeef7d927db25084511e09c621acd927f5f7650c78
parenta11af79e23179ccdd52fb97873c1bfa7b979edb4 (diff)
downloadcandle-25c3cc4149304a4f6eec93b2f88aa9c241f8f696.tar.gz
candle-25c3cc4149304a4f6eec93b2f88aa9c241f8f696.tar.bz2
candle-25c3cc4149304a4f6eec93b2f88aa9c241f8f696.zip
Mention the flash-attention restriction in the readme. (#1158)
-rw-r--r--candle-examples/examples/stable-diffusion/README.md3
1 files changed, 3 insertions, 0 deletions
diff --git a/candle-examples/examples/stable-diffusion/README.md b/candle-examples/examples/stable-diffusion/README.md
index ee83b3f9..b8736a2a 100644
--- a/candle-examples/examples/stable-diffusion/README.md
+++ b/candle-examples/examples/stable-diffusion/README.md
@@ -50,6 +50,9 @@ cached.
Enabling flash-attention requires both a feature flag, `--feature flash-attn`
and using the command line flag `--use-flash-attn`.
+Note that flash-attention-v2 is only compatible with Ampere, Ada, or Hopper GPUs
+(e.g., A100/H100, RTX 3090/4090).
+
## Image to Image Pipeline
...