A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 15, 12:15 UTC
[Doc] Update notes for H2O-VL and Gemma3 (#17219)
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
Cyrus Leung committed
May 6, 2025 at 15:51 UTC
63ced7b43f56c4f81b73b0ad176e820f70b2e782
1 file changed
+1
-6
docs/source/models/supported_models.md
+1
-6
@@ -1118,11 +1118,6 @@ See [this page](#generative-models) for more information on how to use generativ
1118
<sup>E</sup> Pre-computed embeddings can be inputted for this modality.
1119
<sup>+</sup> Multiple items can be inputted per text prompt for this modality.
1120
1121
-:::{important}
1122
-Pan-and-scan image pre-processing is currently supported on V0 (but not V1).
1123
-You can enable it by passing `--mm-processor-kwargs '{"do_pan_and_scan": true}'`.
1124
-:::
1125
-
1121
:::{warning}
1122
Both V0 and V1 support `Gemma3ForConditionalGeneration` for text-only inputs.
1123
However, there are differences in how they handle text + image inputs:
@@ -1142,7 +1137,7 @@ This limitation exists because the model's mixed attention pattern (bidirectiona
1137
:::
1138
1139
:::{note}
1145
-`h2oai/h2ovl-mississippi-2b` will be available in V1 once we support backends other than FlashAttention.
1140
+`h2oai/h2ovl-mississippi-2b` will be available in V1 once we support head size 80.
1141
:::
1142
1143
:::{note}