Remove 'torch_dtype' parameter

Turns out float16 is not supported on CPU, thank you https://stackoverflow.com/a/76126998

app.py

@@ -9,7 +9,7 @@ moondream_id = "vikhyatk/moondream2"
 moondream_revision = "2024-04-02"
 moondream_tokenizer = AutoTokenizer.from_pretrained(moondream_id, revision=moondream_revision)
 moondream: PreTrainedModel = AutoModelForCausalLM.from_pretrained(
-    moondream_id, trust_remote_code=True, revision=moondream_revision, torch_dtype="auto"
+    moondream_id, trust_remote_code=True, revision=moondream_revision
 )
 moondream.eval()