Update src/context/VLMContext.tsx

src/context/VLMContext.tsx

@@ -42,7 +42,7 @@ export const VLMProvider: React.FC<React.PropsWithChildren> = ({ children }) =>
           onProgress?.("Processor loaded. Loading model...");
           modelRef.current = await AutoModelForImageTextToText.from_pretrained(MODEL_ID, {
             dtype: {
-              embed_tokens: "fp16",
+              embed_tokens: "float32",
               vision_encoder: "q4",
               decoder_model_merged: "q4",
             },