perf(magic_pdf): optimize batch ratio calculation for GPU

- Update GPU memory check and batch ratio calculation logic
- Add support for virtual VRAM size environment variable
- Improve logging for GPU memory and batch ratio

magic_pdf/model/doc_analyze_by_custom_model.py

@@ -175,9 +175,9 @@ def doc_analyze(
             npu_support = True
 
     if torch.cuda.is_available() and device != 'cpu' or npu_support:
-        gpu_memory = get_vram(device)
-        if gpu_memory is not None and gpu_memory >= 7.5:
-            batch_ratio = int((gpu_memory-5) // 1)
+        gpu_memory = int(os.getenv("virtual_vram_size", round(get_vram(device))))
+        if gpu_memory is not None and gpu_memory >= 8:
+            batch_ratio = int(gpu_memory-5)
             if batch_ratio >= 1:
                 logger.info(f'gpu_memory: {gpu_memory} GB, batch_ratio: {batch_ratio}')
                 batch_model = BatchAnalyze(model=custom_model, batch_ratio=batch_ratio)