From dc03395bd6ac35ae277fae0c6718bd2c8560ec1f Mon Sep 17 00:00:00 2001
From: Michael Kranzlein <8162250+mkranzlein@users.noreply.github.com>
Date: Wed, 27 Sep 2023 12:07:12 -0400
Subject: [PATCH] Planning chunk+sentence behavior

---
 src/hipool/models.py | 10 ++++++++++
 1 file changed, 10 insertions(+)

diff --git a/src/hipool/models.py b/src/hipool/models.py
index d33ee2c..0597ea1 100644
--- a/src/hipool/models.py
+++ b/src/hipool/models.py
@@ -70,6 +70,16 @@ def forward(self, ids: list[Integer[Tensor, "_ d"]],
             token_type_ids: A list of varied-length tensors token_type_ids.
               All 0s.
         """
+        
+        # Get hipool embedding
+        
+        # Forward pass happens on one or more documents
+        # One is the minimum because hipool needs all of the document's chunks
+        # Pipeline: send document through bert sentence by sentence
+        
+        # Chunking approaches: equal number of sentences, equal number of tokens,
+        #   unequal number of sentences that approximates an equal number of tokens
+        
 
         # Pad such that each sequence has the same number of chunks
         # Padding chunks c-dim vectors, where all the input ids are 0, which is