cat concatenates ReplicatedSharedTensor (#326)

* `cat` concatenates `ReplicatedSharedTensor`

- refer to issue #324
- `cat` can concatenates `ShareTensor`s, but it is not workable for `ReplicatedSharedTensor`.
- added `cat_replicatedShare_tensor` in *apy.py* and *static.py* for the concatenation of `ReplicatedSharedTensor`.

* Saving a variable and formatting

Format
The following commands have been run:
- `python -m black tensor/static.py`
- `python -m black api.py`
- `isort tensor/static.py`

Save number of replicated shares as a variable
- Q: Could you save the len(shares[0].shares) in a variable and then use it here and on line 140? [x]*done*

Amended docstring
- `shares` is a tuple of `ReplicatedSharedTensor`s
- the empty blank is deleted in `cat_replicatedShare_tensor`

* For Python3.7 that doesn't support `math.prod`

- python 3.7 cannot support `math.prod`, only python 3.8+ can.

* Remove python3.7

* Remove tutorial python3.7

Co-authored-by: George Muraru <[email protected]>

.github/workflows/tutorials.yml

@@ -22,7 +22,7 @@ jobs:
     strategy:
       max-parallel: 3
       matrix:
-        python-version: [3.7, 3.8, 3.9]
+        python-version: [3.8, 3.9]
 
     steps:
       - uses: actions/checkout@v2

CONTRIBUTING.md

@@ -65,7 +65,7 @@ If you are new to the project and want to get into the code, we recommend pickin
 Before you get started you will need a few things installed depending on your operating system.
 
 - OS Package Manager
-- Python 3.7+
+- Python 3.8+
 - git
 
 ### OSes
@@ -117,7 +117,7 @@ $ brew install git
 
 ## Python Versions
 
-This project supports Python 3.7+, however, if you are contributing it can help to be able to switch between python versions to fix issues or bugs that relate to a specific python version. Depending on your operating system there are a number of ways to install different versions of python however one of the easiest is with the `pyenv` tool. Additionally, as we will be frequently be installing and changing python packages for this project we should isolate it from your system python and other projects you have using a virtualenv.
+This project supports Python 3.8+, however, if you are contributing it can help to be able to switch between python versions to fix issues or bugs that relate to a specific python version. Depending on your operating system there are a number of ways to install different versions of python however one of the easiest is with the `pyenv` tool. Additionally, as we will be frequently be installing and changing python packages for this project we should isolate it from your system python and other projects you have using a virtualenv.
 
 ### MacOS
 
@@ -152,10 +152,10 @@ $  pyenv install --list | grep 3.9
 3.9.4
 ```
 
-Wow, there are lots of options, lets install 3.7.
+Wow, there are lots of options, lets install 3.8.
 
 ```
-$ pyenv install 3.7.9
+$ pyenv install 3.8.0
 ```
 
 Now, lets see what versions are installed:
@@ -461,7 +461,7 @@ $ pydocstyle .
 
 ### Imports Formatting
 
-We use isort to automatically format the python imports. 
+We use isort to automatically format the python imports.
 Run isort manually like this:
 
 ```

src/sympc/api.py

@@ -262,6 +262,10 @@
         "sympc.tensor.replicatedshare_tensor.ReplicatedSharedTensor.repeat",
         "sympc.tensor.replicatedshare_tensor.ReplicatedSharedTensor",
     ),
+    (
+        "sympc.tensor.static.cat_replicatedShare_tensor",
+        "sympc.tensor.replicatedshare_tensor.ReplicatedSharedTensor",
+    ),
 ]
 
 allowed_external_attrs = [

src/sympc/protocol/falcon/falcon.py

@@ -545,7 +545,7 @@ def private_compare(x: List[MPCTensor], r: torch.Tensor) -> MPCTensor:
             c[i] = u[i] + 1 + w
             w += x[i] ^ r_i
 
-        d = m * math.prod(c)
+            d = m * math.prod(c)
 
         d_val = d.reconstruct(decode=False)  # plaintext d.
         d_val[d_val != 0] = 1  # making all non zero values as 1.

src/sympc/tensor/static.py

@@ -15,8 +15,10 @@
 import numpy as np
 import torch
 
+import sympc.protocol as protocol
 from sympc.session import get_session
 from sympc.tensor.mpc_tensor import MPCTensor
+from sympc.tensor.replicatedshare_tensor import ReplicatedSharedTensor
 from sympc.tensor.share_tensor import ShareTensor
 from sympc.utils import parallel_execution
 
@@ -90,8 +92,12 @@ def cat(tensors: List, dim: int = 0) -> MPCTensor:
         )
     )
 
-    stack_shares = parallel_execution(cat_share_tensor, session.parties)(args)
-    from sympc.tensor import MPCTensor
+    if isinstance(session.protocol, protocol.FSS):
+        stack_shares = parallel_execution(cat_share_tensor, session.parties)(args)
+    elif isinstance(session.protocol, protocol.Falcon):
+        stack_shares = parallel_execution(cat_replicatedShare_tensor, session.parties)(
+            args
+        )
 
     expected_shape = torch.cat(
         [torch.empty(each_tensor.shape) for each_tensor in tensors], dim=dim
@@ -118,6 +124,34 @@ def cat_share_tensor(session_uuid_str: str, *shares: Tuple[ShareTensor]) -> Shar
     return result
 
 
+def cat_replicatedShare_tensor(
+    session_uuid_str: str, *shares: Tuple[ReplicatedSharedTensor]
+) -> ReplicatedSharedTensor:
+    """Helper method that performs torch.cat on the replicated shares of the Tensors.
+
+    Args:
+        session_uuid_str (str): UUID to identify the session on each party side.
+        shares (Tuple[ReplicatedSharedTensor]): Replicated shares of the tensors to be concatenated.
+
+    Returns:
+        ReplicatedSharedTensor: Respective replicated shares after concatenation
+    """
+    session = get_session(session_uuid_str)
+    result = ReplicatedSharedTensor(
+        session_uuid=UUID(session_uuid_str), config=session.config
+    )
+
+    num_shares = len(shares[0].shares)
+
+    cat_result = [torch.tensor([]).type(torch.LongTensor) for _ in range(num_shares)]
+    for share in shares:
+        for i in range(num_shares):
+            cat_result[i] = torch.cat([cat_result[i], share.shares[i]])
+
+    result.shares = cat_result
+    return result
+
+
 def helper_argmax(
     x: MPCTensor,
     dim: Optional[Union[int, Tuple[int]]] = None,