Merge pull request #213 from VikParuchuri/dev

README.md

@@ -50,7 +50,7 @@ There's a hosted API for marker available [here](https://www.datalab.to/):
 
 - Supports PDFs, word documents, and powerpoints 
 - 1/4th the price of leading cloud-based competitors
-- Uses [modal](https://modal.com/) for high reliability without latency spikes
+- Leverages [Modal](https://modal.com/) for high reliability without latency spikes
 
 # Community
 

convert.py

@@ -1,5 +1,6 @@
 import os
 
+os.environ["PYTORCH_ENABLE_MPS_FALLBACK"] = "1" # For some reason, transformers decided to use .isin for a simple op, which is not supported on MPS
 os.environ["IN_STREAMLIT"] = "true" # Avoid multiprocessing inside surya
 os.environ["PDFTEXT_CPU_WORKERS"] = "1" # Avoid multiprocessing inside pdftext
 

convert_single.py

@@ -1,7 +1,8 @@
 import pypdfium2 # Needs to be at the top to avoid warnings
-import argparse
 import os
+os.environ["PYTORCH_ENABLE_MPS_FALLBACK"] = "1" # For some reason, transformers decided to use .isin for a simple op, which is not supported on MPS
 
+import argparse
 from marker.convert import convert_single_pdf
 from marker.logger import configure_logging
 from marker.models import load_all_models

marker/convert.py

@@ -1,6 +1,10 @@
 import warnings
 warnings.filterwarnings("ignore", category=UserWarning) # Filter torch pytree user warnings
 
+import os
+os.environ["PYTORCH_ENABLE_MPS_FALLBACK"] = "1" # For some reason, transformers decided to use .isin for a simple op, which is not supported on MPS
+
+
 import pypdfium2 as pdfium # Needs to be at the top to avoid warnings
 from PIL import Image
 

marker/models.py

@@ -1,3 +1,7 @@
+import os
+os.environ["PYTORCH_ENABLE_MPS_FALLBACK"] = "1" # For some reason, transformers decided to use .isin for a simple op, which is not supported on MPS
+
+
 from marker.postprocessors.editor import load_editing_model
 from surya.model.detection import segformer
 from texify.model.model import load_model as load_texify_model

pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "marker-pdf"
-version = "0.2.14"
+version = "0.2.15"
 description = "Convert PDF to markdown with high speed and accuracy."
 authors = ["Vik Paruchuri <github@vikas.sh>"]
 readme = "README.md"
@@ -30,9 +30,9 @@ torch = "^2.2.2" # Issue with torch 2.3.0 and vision models - https://github.com
 tqdm = "^4.66.1"
 tabulate = "^0.9.0"
 ftfy = "^6.1.1"
-texify = "^0.1.9"
+texify = "^0.1.10"
 rapidfuzz = "^3.8.1"
-surya-ocr = "^0.4.12"
+surya-ocr = "^0.4.14"
 filetype = "^1.2.0"
 regex = "^2024.4.28"
 pdftext = "^0.3.10"