switch to moondream for caption

app.py

@@ -8,7 +8,8 @@ hf_token = os.environ.get('HF_TOKEN')
 
 from gradio_client import Client, handle_file
 
-clipi_client = Client("fffiloni/CLIP-Interrogator-2")
+client = Client("fffiloni/moondream2", token=hf_token)
+
 
 from transformers import AutoTokenizer, AutoModelForCausalLM
 
@@ -17,10 +18,6 @@ model_path = "meta-llama/Llama-2-7b-chat-hf"
 tokenizer = AutoTokenizer.from_pretrained(model_path, use_fast=False, use_auth_token=hf_token)
 model = AutoModelForCausalLM.from_pretrained(model_path, use_auth_token=hf_token).half().cuda()
 
-#client = Client("https://fffiloni-test-llama-api-debug.hf.space/", hf_token=hf_token)
-
-clipi_client = Client("https://fffiloni-clip-interrogator-2.hf.space/")
-
 # FLUX
 
 import numpy as np
@@ -344,18 +341,16 @@ def find_best_perfumes_from_json(data: Union[str, dict], top_n: int = 5, thresho
     
 def infer(image_input):
 
-    gr.Info('Calling CLIP Interrogator ...')
+    gr.Info('Calling Moondream model for caption ...')
 
     yield None, None, None, None
-    clipi_result = clipi_client.predict(
+    moondream_result = client.predict(
 		image=handle_file(image_input),
-		mode="best",
-		best_max_flavors=4,
-		api_name="/clipi2"
+		api_name="/infer"
     )
-    print(clipi_result)
+    print(moondream_result)
 
-    llama_q = clipi_result
+    llama_q = moondream_result
     
     gr.Info('Calling Llama2 ...')
     result = llama_gen_fragrance(llama_q)