x-transformers 2.11.22__tar.gz → 2.11.23__tar.gz

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.

Potentially problematic release.

This version of x-transformers might be problematic. Click here for more details.

Files changed (68) hide show
  1. {x_transformers-2.11.22 → x_transformers-2.11.23}/PKG-INFO +1 -1
  2. {x_transformers-2.11.22 → x_transformers-2.11.23}/pyproject.toml +1 -1
  3. {x_transformers-2.11.22 → x_transformers-2.11.23}/tests/test_x_transformers.py +1 -0
  4. {x_transformers-2.11.22 → x_transformers-2.11.23}/x_transformers/x_transformers.py +2 -1
  5. {x_transformers-2.11.22 → x_transformers-2.11.23}/.github/FUNDING.yml +0 -0
  6. {x_transformers-2.11.22 → x_transformers-2.11.23}/.github/workflows/python-publish.yml +0 -0
  7. {x_transformers-2.11.22 → x_transformers-2.11.23}/.github/workflows/python-test.yaml +0 -0
  8. {x_transformers-2.11.22 → x_transformers-2.11.23}/.gitignore +0 -0
  9. {x_transformers-2.11.22 → x_transformers-2.11.23}/LICENSE +0 -0
  10. {x_transformers-2.11.22 → x_transformers-2.11.23}/README.md +0 -0
  11. {x_transformers-2.11.22 → x_transformers-2.11.23}/data/README.md +0 -0
  12. {x_transformers-2.11.22 → x_transformers-2.11.23}/data/enwik8.gz +0 -0
  13. {x_transformers-2.11.22 → x_transformers-2.11.23}/images/all-attention.png +0 -0
  14. {x_transformers-2.11.22 → x_transformers-2.11.23}/images/attention-on-attention.png +0 -0
  15. {x_transformers-2.11.22 → x_transformers-2.11.23}/images/cosine-sim-attention.png +0 -0
  16. {x_transformers-2.11.22 → x_transformers-2.11.23}/images/deepnorm.png +0 -0
  17. {x_transformers-2.11.22 → x_transformers-2.11.23}/images/dynamic-pos-bias-linear.png +0 -0
  18. {x_transformers-2.11.22 → x_transformers-2.11.23}/images/dynamic-pos-bias-log.png +0 -0
  19. {x_transformers-2.11.22 → x_transformers-2.11.23}/images/dynamic-pos-bias-sinusoidal.png +0 -0
  20. {x_transformers-2.11.22 → x_transformers-2.11.23}/images/dynamic-pos-bias.png +0 -0
  21. {x_transformers-2.11.22 → x_transformers-2.11.23}/images/enhanced-recurrence.png +0 -0
  22. {x_transformers-2.11.22 → x_transformers-2.11.23}/images/fcm.png +0 -0
  23. {x_transformers-2.11.22 → x_transformers-2.11.23}/images/ffglu.png +0 -0
  24. {x_transformers-2.11.22 → x_transformers-2.11.23}/images/flash-attention.png +0 -0
  25. {x_transformers-2.11.22 → x_transformers-2.11.23}/images/gate_values.png +0 -0
  26. {x_transformers-2.11.22 → x_transformers-2.11.23}/images/gating.png +0 -0
  27. {x_transformers-2.11.22 → x_transformers-2.11.23}/images/length-extrapolation-scale.png +0 -0
  28. {x_transformers-2.11.22 → x_transformers-2.11.23}/images/macaron-1.png +0 -0
  29. {x_transformers-2.11.22 → x_transformers-2.11.23}/images/macaron-2.png +0 -0
  30. {x_transformers-2.11.22 → x_transformers-2.11.23}/images/memory-transformer.png +0 -0
  31. {x_transformers-2.11.22 → x_transformers-2.11.23}/images/normformer.png +0 -0
  32. {x_transformers-2.11.22 → x_transformers-2.11.23}/images/pia.png +0 -0
  33. {x_transformers-2.11.22 → x_transformers-2.11.23}/images/qknorm-analysis.png +0 -0
  34. {x_transformers-2.11.22 → x_transformers-2.11.23}/images/resi_dual.png +0 -0
  35. {x_transformers-2.11.22 → x_transformers-2.11.23}/images/residual_attn.png +0 -0
  36. {x_transformers-2.11.22 → x_transformers-2.11.23}/images/rezero.png +0 -0
  37. {x_transformers-2.11.22 → x_transformers-2.11.23}/images/rotary.png +0 -0
  38. {x_transformers-2.11.22 → x_transformers-2.11.23}/images/sandwich-2.png +0 -0
  39. {x_transformers-2.11.22 → x_transformers-2.11.23}/images/sandwich.png +0 -0
  40. {x_transformers-2.11.22 → x_transformers-2.11.23}/images/sandwich_norm.png +0 -0
  41. {x_transformers-2.11.22 → x_transformers-2.11.23}/images/scalenorm.png +0 -0
  42. {x_transformers-2.11.22 → x_transformers-2.11.23}/images/talking-heads.png +0 -0
  43. {x_transformers-2.11.22 → x_transformers-2.11.23}/images/topk-attention.png +0 -0
  44. {x_transformers-2.11.22 → x_transformers-2.11.23}/images/xval.png +0 -0
  45. {x_transformers-2.11.22 → x_transformers-2.11.23}/train_belief_state.py +0 -0
  46. {x_transformers-2.11.22 → x_transformers-2.11.23}/train_copy.py +0 -0
  47. {x_transformers-2.11.22 → x_transformers-2.11.23}/train_entropy_tokenizer.py +0 -0
  48. {x_transformers-2.11.22 → x_transformers-2.11.23}/train_enwik8.py +0 -0
  49. {x_transformers-2.11.22 → x_transformers-2.11.23}/train_free.py +0 -0
  50. {x_transformers-2.11.22 → x_transformers-2.11.23}/train_gpt_vae.py +0 -0
  51. {x_transformers-2.11.22 → x_transformers-2.11.23}/train_length_extrapolate.py +0 -0
  52. {x_transformers-2.11.22 → x_transformers-2.11.23}/train_parity.py +0 -0
  53. {x_transformers-2.11.22 → x_transformers-2.11.23}/train_with_muon.py +0 -0
  54. {x_transformers-2.11.22 → x_transformers-2.11.23}/x_transformers/__init__.py +0 -0
  55. {x_transformers-2.11.22 → x_transformers-2.11.23}/x_transformers/attend.py +0 -0
  56. {x_transformers-2.11.22 → x_transformers-2.11.23}/x_transformers/autoregressive_wrapper.py +0 -0
  57. {x_transformers-2.11.22 → x_transformers-2.11.23}/x_transformers/belief_state_wrapper.py +0 -0
  58. {x_transformers-2.11.22 → x_transformers-2.11.23}/x_transformers/continuous.py +0 -0
  59. {x_transformers-2.11.22 → x_transformers-2.11.23}/x_transformers/dpo.py +0 -0
  60. {x_transformers-2.11.22 → x_transformers-2.11.23}/x_transformers/entropy_based_tokenizer.py +0 -0
  61. {x_transformers-2.11.22 → x_transformers-2.11.23}/x_transformers/free_transformer.py +0 -0
  62. {x_transformers-2.11.22 → x_transformers-2.11.23}/x_transformers/gpt_vae.py +0 -0
  63. {x_transformers-2.11.22 → x_transformers-2.11.23}/x_transformers/multi_input.py +0 -0
  64. {x_transformers-2.11.22 → x_transformers-2.11.23}/x_transformers/neo_mlp.py +0 -0
  65. {x_transformers-2.11.22 → x_transformers-2.11.23}/x_transformers/nonautoregressive_wrapper.py +0 -0
  66. {x_transformers-2.11.22 → x_transformers-2.11.23}/x_transformers/up_wrapper.py +0 -0
  67. {x_transformers-2.11.22 → x_transformers-2.11.23}/x_transformers/xl_autoregressive_wrapper.py +0 -0
  68. {x_transformers-2.11.22 → x_transformers-2.11.23}/x_transformers/xval.py +0 -0
{x_transformers-2.11.22 → x_transformers-2.11.23}/PKG-INFO RENAMED
@@ -1,6 +1,6 @@
1
1
  Metadata-Version: 2.4
2
2
  Name: x-transformers
3
- Version: 2.11.22
3
+ Version: 2.11.23
4
4
  Summary: X-Transformers
5
5
  Project-URL: Homepage, https://pypi.org/project/x-transformers/
6
6
  Project-URL: Repository, https://github.com/lucidrains/x-transformers
{x_transformers-2.11.22 → x_transformers-2.11.23}/pyproject.toml RENAMED
@@ -1,6 +1,6 @@
1
1
  [project]
2
2
  name = "x-transformers"
3
- version = "2.11.22"
3
+ version = "2.11.23"
4
4
  description = "X-Transformers"
5
5
  authors = [
6
6
  { name = "Phil Wang", email = "lucidrains@gmail.com" }
{x_transformers-2.11.22 → x_transformers-2.11.23}/tests/test_x_transformers.py RENAMED
@@ -1478,6 +1478,7 @@ def test_belief_attn(
1478
1478
  dim = 512,
1479
1479
  depth = 6,
1480
1480
  heads = 8,
1481
+ attn_kv_heads = 4,
1481
1482
  rotary_pos_emb = True,
1482
1483
  attn_orthog_projected_values = orthog_project,
1483
1484
  attn_orthog_projected_values_per_head = orthog_project_per_head
{x_transformers-2.11.22 → x_transformers-2.11.23}/x_transformers/x_transformers.py RENAMED
@@ -1431,6 +1431,7 @@ class Attention(Module):
1431
1431
  assert divisible_by(heads, kv_heads)
1432
1432
 
1433
1433
  self.kv_heads = kv_heads
1434
+ self.groups = heads // kv_heads
1434
1435
 
1435
1436
  q_dim = dim_head * heads
1436
1437
  k_dim = dim_head * kv_heads
@@ -2077,7 +2078,7 @@ class Attention(Module):
2077
2078
 
2078
2079
  if self.orthog_projected_values or self.orthog_projected_values_per_head:
2079
2080
  orthog_projected = []
2080
- v_for_proj = self.merge_heads(orig_values)
2081
+ v_for_proj = repeat(orig_values, 'b h n d -> b n (g h d)', g = self.groups)
2081
2082
 
2082
2083
  if self.orthog_projected_values:
2083
2084
  projected = orthog_project(out, v_for_proj)