bugfix

app/qwenvl.py

@@ -37,7 +37,7 @@ class Qwen2VLClass:
 		# max_pixels = 1280*28*28
 		# processor = AutoProcessor.from_pretrained("Qwen/Qwen2-VL-7B-Instruct", min_pixels=min_pixels, max_pixels=max_pixels)
 		
-	def inference(self, prompt, image=None, videofile=None):
+	def inference(self, prompt, image, videofile):
 		content = [
 			{
 				"type":"text",
@@ -88,10 +88,10 @@ class Qwen2VLClass:
 		generated_ids_trimmed = [
 			out_ids[len(in_ids) :] for in_ids, out_ids in zip(inputs.input_ids, generated_ids)
 		]
-		output_text = processor.batch_decode(
+		output_text = self.processor.batch_decode(
 			generated_ids_trimmed, skip_special_tokens=True, clean_up_tokenization_spaces=False
 		)
-		return output_text
+		return output_text[0]
 
 def main():
 	config = getConfig()

app/test.py

@@ -1,3 +1,4 @@
+import time
 import requests
 import base64
 
@@ -21,9 +22,11 @@ while True:
 	i = input()
 	if p == '' or i == '':
 		continue
+	t1 = time.time()
 	ret = requests.get('http://pd4e.com:10090/api',
 					params={
 						'prompt':p,
 						'image':file2b64(i)
 					})
-	print(ret.text)
+	t2 = time.time()
+	print(ret.text, t2 - t1, 'seconds')

conf/config.json

@@ -1,7 +1,7 @@
 {
 	"password_key":"!@#$%^&*(*&^%$QWERTYUIqwertyui234567",
-	"modelname":"Qwen/Qwen2-VL-7B-Instruct",
 	"modelname":"Qwen/Qwen2-VL-2B-Instruct",
+	"modelname":"Qwen/Qwen2-VL-7B-Instruct",
 	"logger":{
 		"name":"qwenvl",
 		"levelname":"info",

wwwroot/api/index.dspy

@@ -1,3 +1,3 @@
 info(f'{params_kw=}')
-return inference(params_kw.prompt, image=params_kw.image, video=params_kw.video)
+return await inference(params_kw.prompt, params_kw.image, params_kw.video)