test input

app.py

@@ -18,14 +18,14 @@ from transformers import AutoModelForCausalLM
 
 models = {
     #"OS-Copilot/OS-Atlas-Base-7B": Qwen2VLForConditionalGeneration.from_pretrained("OS-Copilot/OS-Atlas-Base-7B", torch_dtype="auto", device_map="auto"),
-    "deepseek-ai/deepseek-vl2-tiny": AutoModelForCausalLM.from_pretrained("deepseek-ai/deepseek-vl2-tiny", trust_remote_code=True),
+    #"deepseek-ai/deepseek-vl2-tiny": AutoModelForCausalLM.from_pretrained("deepseek-ai/deepseek-vl2-tiny", trust_remote_code=True),
     "deepseek-ai/deepseek-vl2-small": AutoModelForCausalLM.from_pretrained("deepseek-ai/deepseek-vl2-small", trust_remote_code=True),
     #"deepseek-ai/deepseek-vl2": AutoModelForCausalLM.from_pretrained("deepseek-ai/deepseek-vl2", trust_remote_code=True)
 }
 
 processors = {
     #"OS-Copilot/OS-Atlas-Base-7B": AutoProcessor.from_pretrained("OS-Copilot/OS-Atlas-Base-7B")
-    "deepseek-ai/deepseek-vl2-tiny": DeepseekVLV2Processor.from_pretrained("deepseek-ai/deepseek-vl2-tiny",),
+    #"deepseek-ai/deepseek-vl2-tiny": DeepseekVLV2Processor.from_pretrained("deepseek-ai/deepseek-vl2-tiny",),
     "deepseek-ai/deepseek-vl2-small": DeepseekVLV2Processor.from_pretrained("deepseek-ai/deepseek-vl2-small",),
     #"deepseek-ai/deepseek-vl2": DeepseekVLV2Processor.from_pretrained("deepseek-ai/deepseek-vl2",),
 }
@@ -62,14 +62,13 @@ def rescale_bounding_boxes(bounding_boxes, original_width, original_height, scal
     return rescaled_boxes
 
 
-def deepseek(image, text_input):
+def deepseek(image, text_input, model_id):
     print("helloe!!!!")
     # specify the path to the model
-    model_path = "deepseek-ai/deepseek-vl2-tiny"
-    vl_chat_processor: DeepseekVLV2Processor = DeepseekVLV2Processor.from_pretrained(model_path)
+    vl_chat_processor: DeepseekVLV2Processor = processors[model_id]
     tokenizer = vl_chat_processor.tokenizer
 
-    vl_gpt: DeepseekVLV2ForCausalLM = AutoModelForCausalLM.from_pretrained(model_path, trust_remote_code=True)
+    vl_gpt: DeepseekVLV2ForCausalLM = models[model_id]
     vl_gpt = vl_gpt.to(torch.bfloat16).cuda().eval()
 
     ## single image conversation example
@@ -123,7 +122,7 @@ def deepseek(image, text_input):
 @spaces.GPU
 def run_example(image, text_input, model_id="OS-Copilot/OS-Atlas-Base-7B"):
 
-    return deepseek(image, text_input)
+    return deepseek(image, text_input, model_id)
     
 
 def run_example_old(image, text_input, model_id="OS-Copilot/OS-Atlas-Base-7B"):