liger-kernel-nightly 0.4.2.dev20241210001927__tar.gz → 0.4.2.dev20241210002150__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.
Files changed (69) hide show
  1. {liger_kernel_nightly-0.4.2.dev20241210001927/src/liger_kernel_nightly.egg-info → liger_kernel_nightly-0.4.2.dev20241210002150}/PKG-INFO +1 -1
  2. {liger_kernel_nightly-0.4.2.dev20241210001927 → liger_kernel_nightly-0.4.2.dev20241210002150}/pyproject.toml +1 -1
  3. {liger_kernel_nightly-0.4.2.dev20241210001927 → liger_kernel_nightly-0.4.2.dev20241210002150}/src/liger_kernel/ops/cross_entropy.py +5 -4
  4. {liger_kernel_nightly-0.4.2.dev20241210001927 → liger_kernel_nightly-0.4.2.dev20241210002150/src/liger_kernel_nightly.egg-info}/PKG-INFO +1 -1
  5. {liger_kernel_nightly-0.4.2.dev20241210001927 → liger_kernel_nightly-0.4.2.dev20241210002150}/LICENSE +0 -0
  6. {liger_kernel_nightly-0.4.2.dev20241210001927 → liger_kernel_nightly-0.4.2.dev20241210002150}/NOTICE +0 -0
  7. {liger_kernel_nightly-0.4.2.dev20241210001927 → liger_kernel_nightly-0.4.2.dev20241210002150}/README.md +0 -0
  8. {liger_kernel_nightly-0.4.2.dev20241210001927 → liger_kernel_nightly-0.4.2.dev20241210002150}/setup.cfg +0 -0
  9. {liger_kernel_nightly-0.4.2.dev20241210001927 → liger_kernel_nightly-0.4.2.dev20241210002150}/src/liger_kernel/__init__.py +0 -0
  10. {liger_kernel_nightly-0.4.2.dev20241210001927 → liger_kernel_nightly-0.4.2.dev20241210002150}/src/liger_kernel/chunked_loss/__init__.py +0 -0
  11. {liger_kernel_nightly-0.4.2.dev20241210001927 → liger_kernel_nightly-0.4.2.dev20241210002150}/src/liger_kernel/chunked_loss/cpo_loss.py +0 -0
  12. {liger_kernel_nightly-0.4.2.dev20241210001927 → liger_kernel_nightly-0.4.2.dev20241210002150}/src/liger_kernel/chunked_loss/dpo_loss.py +0 -0
  13. {liger_kernel_nightly-0.4.2.dev20241210001927 → liger_kernel_nightly-0.4.2.dev20241210002150}/src/liger_kernel/chunked_loss/functional.py +0 -0
  14. {liger_kernel_nightly-0.4.2.dev20241210001927 → liger_kernel_nightly-0.4.2.dev20241210002150}/src/liger_kernel/chunked_loss/fused_linear_distillation.py +0 -0
  15. {liger_kernel_nightly-0.4.2.dev20241210001927 → liger_kernel_nightly-0.4.2.dev20241210002150}/src/liger_kernel/chunked_loss/fused_linear_preference.py +0 -0
  16. {liger_kernel_nightly-0.4.2.dev20241210001927 → liger_kernel_nightly-0.4.2.dev20241210002150}/src/liger_kernel/chunked_loss/orpo_loss.py +0 -0
  17. {liger_kernel_nightly-0.4.2.dev20241210001927 → liger_kernel_nightly-0.4.2.dev20241210002150}/src/liger_kernel/chunked_loss/simpo_loss.py +0 -0
  18. {liger_kernel_nightly-0.4.2.dev20241210001927 → liger_kernel_nightly-0.4.2.dev20241210002150}/src/liger_kernel/env_report.py +0 -0
  19. {liger_kernel_nightly-0.4.2.dev20241210001927 → liger_kernel_nightly-0.4.2.dev20241210002150}/src/liger_kernel/ops/__init__.py +0 -0
  20. {liger_kernel_nightly-0.4.2.dev20241210001927 → liger_kernel_nightly-0.4.2.dev20241210002150}/src/liger_kernel/ops/experimental/embedding.py +0 -0
  21. {liger_kernel_nightly-0.4.2.dev20241210001927 → liger_kernel_nightly-0.4.2.dev20241210002150}/src/liger_kernel/ops/experimental/mm_int8int2.py +0 -0
  22. {liger_kernel_nightly-0.4.2.dev20241210001927 → liger_kernel_nightly-0.4.2.dev20241210002150}/src/liger_kernel/ops/fused_linear_cross_entropy.py +0 -0
  23. {liger_kernel_nightly-0.4.2.dev20241210001927 → liger_kernel_nightly-0.4.2.dev20241210002150}/src/liger_kernel/ops/fused_linear_jsd.py +0 -0
  24. {liger_kernel_nightly-0.4.2.dev20241210001927 → liger_kernel_nightly-0.4.2.dev20241210002150}/src/liger_kernel/ops/geglu.py +0 -0
  25. {liger_kernel_nightly-0.4.2.dev20241210001927 → liger_kernel_nightly-0.4.2.dev20241210002150}/src/liger_kernel/ops/group_norm.py +0 -0
  26. {liger_kernel_nightly-0.4.2.dev20241210001927 → liger_kernel_nightly-0.4.2.dev20241210002150}/src/liger_kernel/ops/jsd.py +0 -0
  27. {liger_kernel_nightly-0.4.2.dev20241210001927 → liger_kernel_nightly-0.4.2.dev20241210002150}/src/liger_kernel/ops/kl_div.py +0 -0
  28. {liger_kernel_nightly-0.4.2.dev20241210001927 → liger_kernel_nightly-0.4.2.dev20241210002150}/src/liger_kernel/ops/layer_norm.py +0 -0
  29. {liger_kernel_nightly-0.4.2.dev20241210001927 → liger_kernel_nightly-0.4.2.dev20241210002150}/src/liger_kernel/ops/qwen2vl_mrope.py +0 -0
  30. {liger_kernel_nightly-0.4.2.dev20241210001927 → liger_kernel_nightly-0.4.2.dev20241210002150}/src/liger_kernel/ops/rms_norm.py +0 -0
  31. {liger_kernel_nightly-0.4.2.dev20241210001927 → liger_kernel_nightly-0.4.2.dev20241210002150}/src/liger_kernel/ops/rope.py +0 -0
  32. {liger_kernel_nightly-0.4.2.dev20241210001927 → liger_kernel_nightly-0.4.2.dev20241210002150}/src/liger_kernel/ops/swiglu.py +0 -0
  33. {liger_kernel_nightly-0.4.2.dev20241210001927 → liger_kernel_nightly-0.4.2.dev20241210002150}/src/liger_kernel/ops/utils.py +0 -0
  34. {liger_kernel_nightly-0.4.2.dev20241210001927 → liger_kernel_nightly-0.4.2.dev20241210002150}/src/liger_kernel/transformers/__init__.py +0 -0
  35. {liger_kernel_nightly-0.4.2.dev20241210001927 → liger_kernel_nightly-0.4.2.dev20241210002150}/src/liger_kernel/transformers/auto_model.py +0 -0
  36. {liger_kernel_nightly-0.4.2.dev20241210001927 → liger_kernel_nightly-0.4.2.dev20241210002150}/src/liger_kernel/transformers/cross_entropy.py +0 -0
  37. {liger_kernel_nightly-0.4.2.dev20241210001927 → liger_kernel_nightly-0.4.2.dev20241210002150}/src/liger_kernel/transformers/experimental/embedding.py +0 -0
  38. {liger_kernel_nightly-0.4.2.dev20241210001927 → liger_kernel_nightly-0.4.2.dev20241210002150}/src/liger_kernel/transformers/functional.py +0 -0
  39. {liger_kernel_nightly-0.4.2.dev20241210001927 → liger_kernel_nightly-0.4.2.dev20241210002150}/src/liger_kernel/transformers/fused_linear_cross_entropy.py +0 -0
  40. {liger_kernel_nightly-0.4.2.dev20241210001927 → liger_kernel_nightly-0.4.2.dev20241210002150}/src/liger_kernel/transformers/fused_linear_jsd.py +0 -0
  41. {liger_kernel_nightly-0.4.2.dev20241210001927 → liger_kernel_nightly-0.4.2.dev20241210002150}/src/liger_kernel/transformers/geglu.py +0 -0
  42. {liger_kernel_nightly-0.4.2.dev20241210001927 → liger_kernel_nightly-0.4.2.dev20241210002150}/src/liger_kernel/transformers/group_norm.py +0 -0
  43. {liger_kernel_nightly-0.4.2.dev20241210001927 → liger_kernel_nightly-0.4.2.dev20241210002150}/src/liger_kernel/transformers/jsd.py +0 -0
  44. {liger_kernel_nightly-0.4.2.dev20241210001927 → liger_kernel_nightly-0.4.2.dev20241210002150}/src/liger_kernel/transformers/kl_div.py +0 -0
  45. {liger_kernel_nightly-0.4.2.dev20241210001927 → liger_kernel_nightly-0.4.2.dev20241210002150}/src/liger_kernel/transformers/layer_norm.py +0 -0
  46. {liger_kernel_nightly-0.4.2.dev20241210001927 → liger_kernel_nightly-0.4.2.dev20241210002150}/src/liger_kernel/transformers/model/__init__.py +0 -0
  47. {liger_kernel_nightly-0.4.2.dev20241210001927 → liger_kernel_nightly-0.4.2.dev20241210002150}/src/liger_kernel/transformers/model/gemma.py +0 -0
  48. {liger_kernel_nightly-0.4.2.dev20241210001927 → liger_kernel_nightly-0.4.2.dev20241210002150}/src/liger_kernel/transformers/model/gemma2.py +0 -0
  49. {liger_kernel_nightly-0.4.2.dev20241210001927 → liger_kernel_nightly-0.4.2.dev20241210002150}/src/liger_kernel/transformers/model/llama.py +0 -0
  50. {liger_kernel_nightly-0.4.2.dev20241210001927 → liger_kernel_nightly-0.4.2.dev20241210002150}/src/liger_kernel/transformers/model/mistral.py +0 -0
  51. {liger_kernel_nightly-0.4.2.dev20241210001927 → liger_kernel_nightly-0.4.2.dev20241210002150}/src/liger_kernel/transformers/model/mixtral.py +0 -0
  52. {liger_kernel_nightly-0.4.2.dev20241210001927 → liger_kernel_nightly-0.4.2.dev20241210002150}/src/liger_kernel/transformers/model/mllama.py +0 -0
  53. {liger_kernel_nightly-0.4.2.dev20241210001927 → liger_kernel_nightly-0.4.2.dev20241210002150}/src/liger_kernel/transformers/model/phi3.py +0 -0
  54. {liger_kernel_nightly-0.4.2.dev20241210001927 → liger_kernel_nightly-0.4.2.dev20241210002150}/src/liger_kernel/transformers/model/qwen2.py +0 -0
  55. {liger_kernel_nightly-0.4.2.dev20241210001927 → liger_kernel_nightly-0.4.2.dev20241210002150}/src/liger_kernel/transformers/model/qwen2_vl.py +0 -0
  56. {liger_kernel_nightly-0.4.2.dev20241210001927 → liger_kernel_nightly-0.4.2.dev20241210002150}/src/liger_kernel/transformers/monkey_patch.py +0 -0
  57. {liger_kernel_nightly-0.4.2.dev20241210001927 → liger_kernel_nightly-0.4.2.dev20241210002150}/src/liger_kernel/transformers/orpo_trainer.py +0 -0
  58. {liger_kernel_nightly-0.4.2.dev20241210001927 → liger_kernel_nightly-0.4.2.dev20241210002150}/src/liger_kernel/transformers/qwen2vl_mrope.py +0 -0
  59. {liger_kernel_nightly-0.4.2.dev20241210001927 → liger_kernel_nightly-0.4.2.dev20241210002150}/src/liger_kernel/transformers/rms_norm.py +0 -0
  60. {liger_kernel_nightly-0.4.2.dev20241210001927 → liger_kernel_nightly-0.4.2.dev20241210002150}/src/liger_kernel/transformers/rope.py +0 -0
  61. {liger_kernel_nightly-0.4.2.dev20241210001927 → liger_kernel_nightly-0.4.2.dev20241210002150}/src/liger_kernel/transformers/swiglu.py +0 -0
  62. {liger_kernel_nightly-0.4.2.dev20241210001927 → liger_kernel_nightly-0.4.2.dev20241210002150}/src/liger_kernel/transformers/trainer_integration.py +0 -0
  63. {liger_kernel_nightly-0.4.2.dev20241210001927 → liger_kernel_nightly-0.4.2.dev20241210002150}/src/liger_kernel/triton/__init__.py +0 -0
  64. {liger_kernel_nightly-0.4.2.dev20241210001927 → liger_kernel_nightly-0.4.2.dev20241210002150}/src/liger_kernel/triton/monkey_patch.py +0 -0
  65. {liger_kernel_nightly-0.4.2.dev20241210001927 → liger_kernel_nightly-0.4.2.dev20241210002150}/src/liger_kernel/utils.py +0 -0
  66. {liger_kernel_nightly-0.4.2.dev20241210001927 → liger_kernel_nightly-0.4.2.dev20241210002150}/src/liger_kernel_nightly.egg-info/SOURCES.txt +0 -0
  67. {liger_kernel_nightly-0.4.2.dev20241210001927 → liger_kernel_nightly-0.4.2.dev20241210002150}/src/liger_kernel_nightly.egg-info/dependency_links.txt +0 -0
  68. {liger_kernel_nightly-0.4.2.dev20241210001927 → liger_kernel_nightly-0.4.2.dev20241210002150}/src/liger_kernel_nightly.egg-info/requires.txt +0 -0
  69. {liger_kernel_nightly-0.4.2.dev20241210001927 → liger_kernel_nightly-0.4.2.dev20241210002150}/src/liger_kernel_nightly.egg-info/top_level.txt +0 -0
@@ -1,6 +1,6 @@
1
1
  Metadata-Version: 2.1
2
2
  Name: liger_kernel_nightly
3
- Version: 0.4.2.dev20241210001927
3
+ Version: 0.4.2.dev20241210002150
4
4
  Summary: Efficient Triton kernels for LLM Training
5
5
  License: BSD 2-CLAUSE LICENSE
6
6
  Copyright 2024 LinkedIn Corporation
@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
4
4
 
5
5
  [project]
6
6
  name = "liger_kernel_nightly"
7
- version = "0.4.2.dev20241210001927"
7
+ version = "0.4.2.dev20241210002150"
8
8
  description = "Efficient Triton kernels for LLM Training"
9
9
  urls = { "Homepage" = "https://github.com/linkedin/Liger-Kernel" }
10
10
  readme = { file = "README.md", content-type = "text/markdown" }
@@ -285,11 +285,12 @@ def cross_entropy_forward(
285
285
  num_warps=32 if not is_hip() else 16,
286
286
  )
287
287
 
288
- loss = torch.sum(loss_1d)
289
- if return_z_loss == _TRUE.value:
290
- z_loss = torch.sum(z_loss_1d)
288
+ if reduction == "none":
289
+ loss = loss_1d
290
+ z_loss = z_loss_1d if return_z_loss == _TRUE.value else None
291
291
  else:
292
- z_loss = None
292
+ loss = torch.sum(loss_1d)
293
+ z_loss = torch.sum(z_loss_1d) if return_z_loss == _TRUE.value else None
293
294
 
294
295
  return loss, z_loss, _input
295
296
 
@@ -1,6 +1,6 @@
1
1
  Metadata-Version: 2.1
2
2
  Name: liger_kernel_nightly
3
- Version: 0.4.2.dev20241210001927
3
+ Version: 0.4.2.dev20241210002150
4
4
  Summary: Efficient Triton kernels for LLM Training
5
5
  License: BSD 2-CLAUSE LICENSE
6
6
  Copyright 2024 LinkedIn Corporation