diff --git a/README.md b/README.md
index 2f2435ec..0cf03ea9 100755
--- a/README.md
+++ b/README.md
@@ -14,7 +14,12 @@ Pytorch original implementation of TransCoder in [Unsupervised Translation of Pr
 - [submitit](https://pypi.org/project/submitit/) (to run the preprocessing pipeline on remote machine) 
 - [six](https://pypi.org/project/six/) 
 - [sacrebleu](https://pypi.org/project/sacrebleu/) (`pip install sacrebleu=="1.2.11"`)
- 
+
+All this dependencies can be installed using :
+```
+$ pip install -r requirement.txt
+```
+
 If your `libclang.so` is not in `/usr/lib/llvm-7/lib/`, replace the path to `libclang.so` to the correct path in `clang.cindex.Config.set_library_path('path_to_libclang')` in `code_tokenizer.py` 
 
 If you run the data preprocessing pipeline, you will have to compile fastBPE. Go in XLM/tools/fastBPE and carry out the steps described in the [ReadMe](https://github.com/facebookresearch/TransCoder/blob/master/XLM/tools/README.md).
diff --git a/requirements.txt b/requirements.txt
new file mode 100644
index 00000000..f76a2c1e
--- /dev/null
+++ b/requirements.txt
@@ -0,0 +1,10 @@
+numpy==1.19.4
+Cython==0.29.21
+torch==1.7.1
+fastBPE==0.1.0
+moses==0.10.0
+apex==0.9.10.dev0
+libclang==10.0.1.0
+submitit==1.1.5
+six==1.15.0
+sacrebleu==1.2.11