x-transformers 2.2.4__tar.gz → 2.2.6__tar.gz

{x_transformers-2.2.4 → x_transformers-2.2.6}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: x-transformers
-Version: 2.2.4
+Version: 2.2.6
 Summary: X-Transformers
 Project-URL: Homepage, https://pypi.org/project/x-transformers/
 Project-URL: Repository, https://github.com/lucidrains/x-transformers

{x_transformers-2.2.4 → x_transformers-2.2.6}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "x-transformers"
-version = "2.2.4"
+version = "2.2.6"
 description = "X-Transformers"
 authors = [
     { name = "Phil Wang", email = "lucidrains@gmail.com" }

{x_transformers-2.2.4 → x_transformers-2.2.6}/x_transformers/entropy_based_tokenizer.py

@@ -6,7 +6,11 @@ import torch.nn.functional as F
 from torch.nn import Module
 from torch.nn.utils.rnn import pad_sequence
 
-from x_transformers.x_transformers import Decoder, TransformerWrapper
+from x_transformers.x_transformers import (
+    Decoder,
+    TransformerWrapper,
+    calc_entropy
+)
 
 import einx
 from einops import repeat, rearrange, pack, unpack
@@ -27,12 +31,10 @@ def default(v, d):
 class EntropyBasedTokenizer(Module):
     def __init__(
         self,
-        decoder: TransformerWrapper,
+        decoder: Module,
         entropy_threshold: float
     ):
         super().__init__()
-        assert isinstance(decoder.attn_layers, Decoder)
-
         self.decoder = decoder
         self.entropy_threshold = entropy_threshold
 
@@ -41,7 +43,8 @@ class EntropyBasedTokenizer(Module):
         self,
         seq,            # Float['b n'] | Float['n']
         lens = None,    # Int['b']
-        return_segmented_seq = False
+        return_segmented_seq = False,
+        decoder_forward_kwargs: dict = dict()
     ):
         no_batch_dim = seq.ndim == 1
         seq, maybe_batch_ps = pack((seq,), '* n')
@@ -55,9 +58,9 @@ class EntropyBasedTokenizer(Module):
 
         # forward through a small trained decoder and get the entropies of the logits
 
-        _, intermediates = self.decoder(seq, return_logit_entropies = True)
+        logits = self.decoder(seq, **decoder_forward_kwargs)
 
-        entropies = intermediates.logit_entropies
+        entropies = calc_entropy(logits)
 
         # get length mask for boundaries