fixed pixtral

app.py

@@ -55,8 +55,8 @@ if "pipeline" not in st.session_state:
         reranker_model="monovlm",
         device=device,
         invoice_json_path=invoice_json_path,
-        gpu_memory_utilization=0.8,
-        max_tokens=4096,
+        gpu_memory_utilization=0.65,
+        max_tokens=2048,
     )
 pipeline = st.session_state.pipeline
 
@@ -114,6 +114,11 @@ def display_invoice(image_path):
         return None
 
 
+def resize_image_and_save(img: Image, image_path: str, max_width: int = 800, max_height: int = 800):
+    img.thumbnail((max_width, max_height))
+    img.save(image_path, "PNG", quality=85)
+
+
 st.title("Upload FNOL photo")
 
 col1, col2, col3 = st.columns([4, 1, 4])
@@ -124,7 +129,7 @@ with col1:
         try:
             image = Image.open(uploaded_image)
             image_path = f"{output_folder}/{str(uuid4())[:5]}.png"
-            image.save(image_path)
+            resize_image_and_save(image, image_path)
             print(f"Image: {image_path}")
             st.image(image, caption="Uploaded photo", width=300)
             st.session_state["image"] = image_path

src/model.py

@@ -9,10 +9,10 @@ def encode_image(image_path: str):
 
 
 class Pixtral:
-    def __init__(self, max_model_len=4096, max_tokens=4096, gpu_memory_utilization=0.95, temperature=0.4):
+    def __init__(self, max_model_len=4096, max_tokens=2048, gpu_memory_utilization=0.65, temperature=0.35):
         self.model_name = "mistralai/Pixtral-12B-2409"
 
-        self.sampling_params = SamplingParams(max_tokens=max_tokens, temperature=temperature, top_p=0.8, top_k=40)
+        self.sampling_params = SamplingParams(max_tokens=max_tokens, temperature=temperature)
 
         self.llm = LLM(
             model=self.model_name,

src/prompts.py

@@ -4,6 +4,7 @@ The invoice should include details on the parts needing replacement, labor hours
 Include only repair and cost information, with no contact details or extra data.
 
 Important: Use the example invoice as a reference only. Do not copy it directly, but you can use labor costs and part names as a reference, if they are relevant to the image.
+Pay attention to little details which are included in reference invoice. It's the granularity level which you must follow (but, again, do not copy all the details directly, only if they are indeed needed).
 
 The invoice should include a list of items, where each item has the following fields:
 
@@ -26,16 +27,16 @@ Gesamtkosten (Total Cost, EUR)
 ---
 
 **Guidelines for Generation:**
-1. **Unique Response**: Generate reasonable variations in part types, costs, and labor hours based on the damage visible in the provided image.
-2. **Check Calculations**: Ensure that **Gesamtkosten** for each item is calculated as:
+1. **Adhere to Granularity**: Ensure that each damaged component is itemized in the invoice. Follow the level of granularity in the example invoice, avoiding broad categories like "front end" or "bumper assembly."
+2. **Realistic Labor and Costs**: Base labor hours and part costs on the visible damage described or implied in the image. Use reasonable values consistent with the invoice example.
+3. **Check Calculations**: Ensure that **Gesamtkosten** for each item is calculated as:
    \[
    Gesamtkosten = Teilkosten + (Arbeitsstunden x Arbeitskosten)
    \]
    The **Gesamtsumme** should reflect the sum of all **Gesamtkosten** entries.
+4. **Unique Response**: Generate variations in part names, costs, and labor hours specific to the image, while adhering to the reference invoice’s level of granularity.
 
-Generate a detailed and realistic invoice tailored specifically to the image, without replicating the example.
-
-OUTPUT REQUIREMENTS:
+**OUTPUT REQUIREMENTS:**
 Your output should be a table in markdown format WITHOUT ANY ADDITIONAL COMMENTS. The format of a table is provided in reference invoice examle.
 """.format