From 14b810c4519d00ecf03d70445793a2e15482ff61 Mon Sep 17 00:00:00 2001 From: GassiGiuseppe Date: Mon, 6 Oct 2025 21:41:45 +0200 Subject: [PATCH] WIP NanoSocratesEmbedder for batching --- .../Libs/Embedder/Classes/NanoSocratesEmbedder.py | 8 +++++--- 1 file changed, 5 insertions(+), 3 deletions(-) diff --git a/Project_Model/Libs/Embedder/Classes/NanoSocratesEmbedder.py b/Project_Model/Libs/Embedder/Classes/NanoSocratesEmbedder.py index 0377331..96315ed 100644 --- a/Project_Model/Libs/Embedder/Classes/NanoSocratesEmbedder.py +++ b/Project_Model/Libs/Embedder/Classes/NanoSocratesEmbedder.py @@ -1,5 +1,7 @@ import torch from ..Utils import fixed_positional_encoding + +# WIP FOR BATCHING class NanoSocratesEmbedder(torch.nn.Module): def __init__( @@ -13,20 +15,20 @@ class NanoSocratesEmbedder(torch.nn.Module): embedding_size ) - def forward(self, tokenized_sentence: list[int]) -> torch.Tensor: + def forward(self, tokenized_sentence: list[list[int]]) -> torch.Tensor: TOKENIZED_TENSOR = torch.tensor(tokenized_sentence) computed_embeddings: torch.Tensor = self.__embedder(TOKENIZED_TENSOR) - SENTENCE_LENGHT, EMBEDDING_SIZE = computed_embeddings.shape + _ ,SENTENCE_LENGHT, EMBEDDING_SIZE = computed_embeddings.shape # for batching POSITIONAL_ENCODINGS = fixed_positional_encoding( SENTENCE_LENGHT, EMBEDDING_SIZE ) - computed_embeddings = computed_embeddings + POSITIONAL_ENCODINGS + computed_embeddings = computed_embeddings + POSITIONAL_ENCODINGS.unsqueeze(0) # for batching return computed_embeddings