More bug fixes

README.md

@@ -35,7 +35,7 @@ It only uses models where necessary, which improves speed and accuracy.
 
 ![Benchmark overall](data/images/overall.png)
 
-The above results are with marker and nougat setup so they each take ~3GB of VRAM on an A6000.
+The above results are with marker and nougat setup so they each take ~4GB of VRAM on an A6000.
 
 See [below](#benchmarks) for detailed speed and accuracy benchmarks, and instructions on how to run your own benchmarks.
 
@@ -178,7 +178,7 @@ Peak GPU memory usage during the benchmark is `4.2GB` for nougat, and `4.1GB` fo
 
 **Throughput**
 
-Marker takes about 4GB of VRAM on average per task, so you can convert 12 documents in parallel on an A6000.
+Marker takes about 4.5GB of VRAM on average per task, so you can convert 10 documents in parallel on an A6000.
 
 ![Benchmark results](data/images/per_doc.png)
 

marker/settings.py

@@ -27,7 +27,7 @@ def TORCH_DEVICE_MODEL(self) -> str:
         return "cpu"
 
     INFERENCE_RAM: int = 40 # How much VRAM each GPU has (in GB).
-    VRAM_PER_TASK: float = 4 # How much VRAM to allocate per task (in GB).  Peak marker VRAM usage is around 5GB, but avg across workers is lower.
+    VRAM_PER_TASK: float = 4.5 # How much VRAM to allocate per task (in GB).  Peak marker VRAM usage is around 5GB, but avg across workers is lower.
     DEFAULT_LANG: str = "English" # Default language we assume files to be in, should be one of the keys in TESSERACT_LANGUAGES
 
     SUPPORTED_FILETYPES: Dict = {

pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "marker-pdf"
-version = "0.2.3"
+version = "0.2.4"
 description = "Convert PDF to markdown with high speed and accuracy."
 authors = ["Vik Paruchuri <[email protected]>"]
 readme = "README.md"