liger-kernel-nightly 0.5.0.dev20241210032000__tar.gz → 0.5.1.dev20241210093009__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.
- {liger_kernel_nightly-0.5.0.dev20241210032000/src/liger_kernel_nightly.egg-info → liger_kernel_nightly-0.5.1.dev20241210093009}/PKG-INFO +3 -2
- {liger_kernel_nightly-0.5.0.dev20241210032000 → liger_kernel_nightly-0.5.1.dev20241210093009}/pyproject.toml +5 -2
- {liger_kernel_nightly-0.5.0.dev20241210032000 → liger_kernel_nightly-0.5.1.dev20241210093009}/src/liger_kernel/transformers/__init__.py +0 -1
- liger_kernel_nightly-0.5.1.dev20241210093009/src/liger_kernel/transformers/trainer/__init__.py +6 -0
- {liger_kernel_nightly-0.5.0.dev20241210032000/src/liger_kernel/transformers → liger_kernel_nightly-0.5.1.dev20241210093009/src/liger_kernel/transformers/trainer}/orpo_trainer.py +1 -3
- {liger_kernel_nightly-0.5.0.dev20241210032000 → liger_kernel_nightly-0.5.1.dev20241210093009/src/liger_kernel_nightly.egg-info}/PKG-INFO +3 -2
- {liger_kernel_nightly-0.5.0.dev20241210032000 → liger_kernel_nightly-0.5.1.dev20241210093009}/src/liger_kernel_nightly.egg-info/SOURCES.txt +2 -1
- {liger_kernel_nightly-0.5.0.dev20241210032000 → liger_kernel_nightly-0.5.1.dev20241210093009}/src/liger_kernel_nightly.egg-info/requires.txt +3 -1
- {liger_kernel_nightly-0.5.0.dev20241210032000 → liger_kernel_nightly-0.5.1.dev20241210093009}/LICENSE +0 -0
- {liger_kernel_nightly-0.5.0.dev20241210032000 → liger_kernel_nightly-0.5.1.dev20241210093009}/NOTICE +0 -0
- {liger_kernel_nightly-0.5.0.dev20241210032000 → liger_kernel_nightly-0.5.1.dev20241210093009}/README.md +0 -0
- {liger_kernel_nightly-0.5.0.dev20241210032000 → liger_kernel_nightly-0.5.1.dev20241210093009}/setup.cfg +0 -0
- {liger_kernel_nightly-0.5.0.dev20241210032000 → liger_kernel_nightly-0.5.1.dev20241210093009}/src/liger_kernel/__init__.py +0 -0
- {liger_kernel_nightly-0.5.0.dev20241210032000 → liger_kernel_nightly-0.5.1.dev20241210093009}/src/liger_kernel/chunked_loss/__init__.py +0 -0
- {liger_kernel_nightly-0.5.0.dev20241210032000 → liger_kernel_nightly-0.5.1.dev20241210093009}/src/liger_kernel/chunked_loss/cpo_loss.py +0 -0
- {liger_kernel_nightly-0.5.0.dev20241210032000 → liger_kernel_nightly-0.5.1.dev20241210093009}/src/liger_kernel/chunked_loss/dpo_loss.py +0 -0
- {liger_kernel_nightly-0.5.0.dev20241210032000 → liger_kernel_nightly-0.5.1.dev20241210093009}/src/liger_kernel/chunked_loss/functional.py +0 -0
- {liger_kernel_nightly-0.5.0.dev20241210032000 → liger_kernel_nightly-0.5.1.dev20241210093009}/src/liger_kernel/chunked_loss/fused_linear_distillation.py +0 -0
- {liger_kernel_nightly-0.5.0.dev20241210032000 → liger_kernel_nightly-0.5.1.dev20241210093009}/src/liger_kernel/chunked_loss/fused_linear_preference.py +0 -0
- {liger_kernel_nightly-0.5.0.dev20241210032000 → liger_kernel_nightly-0.5.1.dev20241210093009}/src/liger_kernel/chunked_loss/orpo_loss.py +0 -0
- {liger_kernel_nightly-0.5.0.dev20241210032000 → liger_kernel_nightly-0.5.1.dev20241210093009}/src/liger_kernel/chunked_loss/simpo_loss.py +0 -0
- {liger_kernel_nightly-0.5.0.dev20241210032000 → liger_kernel_nightly-0.5.1.dev20241210093009}/src/liger_kernel/env_report.py +0 -0
- {liger_kernel_nightly-0.5.0.dev20241210032000 → liger_kernel_nightly-0.5.1.dev20241210093009}/src/liger_kernel/ops/__init__.py +0 -0
- {liger_kernel_nightly-0.5.0.dev20241210032000 → liger_kernel_nightly-0.5.1.dev20241210093009}/src/liger_kernel/ops/cross_entropy.py +0 -0
- {liger_kernel_nightly-0.5.0.dev20241210032000 → liger_kernel_nightly-0.5.1.dev20241210093009}/src/liger_kernel/ops/experimental/embedding.py +0 -0
- {liger_kernel_nightly-0.5.0.dev20241210032000 → liger_kernel_nightly-0.5.1.dev20241210093009}/src/liger_kernel/ops/experimental/mm_int8int2.py +0 -0
- {liger_kernel_nightly-0.5.0.dev20241210032000 → liger_kernel_nightly-0.5.1.dev20241210093009}/src/liger_kernel/ops/fused_linear_cross_entropy.py +0 -0
- {liger_kernel_nightly-0.5.0.dev20241210032000 → liger_kernel_nightly-0.5.1.dev20241210093009}/src/liger_kernel/ops/fused_linear_jsd.py +0 -0
- {liger_kernel_nightly-0.5.0.dev20241210032000 → liger_kernel_nightly-0.5.1.dev20241210093009}/src/liger_kernel/ops/geglu.py +0 -0
- {liger_kernel_nightly-0.5.0.dev20241210032000 → liger_kernel_nightly-0.5.1.dev20241210093009}/src/liger_kernel/ops/group_norm.py +0 -0
- {liger_kernel_nightly-0.5.0.dev20241210032000 → liger_kernel_nightly-0.5.1.dev20241210093009}/src/liger_kernel/ops/jsd.py +0 -0
- {liger_kernel_nightly-0.5.0.dev20241210032000 → liger_kernel_nightly-0.5.1.dev20241210093009}/src/liger_kernel/ops/kl_div.py +0 -0
- {liger_kernel_nightly-0.5.0.dev20241210032000 → liger_kernel_nightly-0.5.1.dev20241210093009}/src/liger_kernel/ops/layer_norm.py +0 -0
- {liger_kernel_nightly-0.5.0.dev20241210032000 → liger_kernel_nightly-0.5.1.dev20241210093009}/src/liger_kernel/ops/qwen2vl_mrope.py +0 -0
- {liger_kernel_nightly-0.5.0.dev20241210032000 → liger_kernel_nightly-0.5.1.dev20241210093009}/src/liger_kernel/ops/rms_norm.py +0 -0
- {liger_kernel_nightly-0.5.0.dev20241210032000 → liger_kernel_nightly-0.5.1.dev20241210093009}/src/liger_kernel/ops/rope.py +0 -0
- {liger_kernel_nightly-0.5.0.dev20241210032000 → liger_kernel_nightly-0.5.1.dev20241210093009}/src/liger_kernel/ops/swiglu.py +0 -0
- {liger_kernel_nightly-0.5.0.dev20241210032000 → liger_kernel_nightly-0.5.1.dev20241210093009}/src/liger_kernel/ops/utils.py +0 -0
- {liger_kernel_nightly-0.5.0.dev20241210032000 → liger_kernel_nightly-0.5.1.dev20241210093009}/src/liger_kernel/transformers/auto_model.py +0 -0
- {liger_kernel_nightly-0.5.0.dev20241210032000 → liger_kernel_nightly-0.5.1.dev20241210093009}/src/liger_kernel/transformers/cross_entropy.py +0 -0
- {liger_kernel_nightly-0.5.0.dev20241210032000 → liger_kernel_nightly-0.5.1.dev20241210093009}/src/liger_kernel/transformers/experimental/embedding.py +0 -0
- {liger_kernel_nightly-0.5.0.dev20241210032000 → liger_kernel_nightly-0.5.1.dev20241210093009}/src/liger_kernel/transformers/functional.py +0 -0
- {liger_kernel_nightly-0.5.0.dev20241210032000 → liger_kernel_nightly-0.5.1.dev20241210093009}/src/liger_kernel/transformers/fused_linear_cross_entropy.py +0 -0
- {liger_kernel_nightly-0.5.0.dev20241210032000 → liger_kernel_nightly-0.5.1.dev20241210093009}/src/liger_kernel/transformers/fused_linear_jsd.py +0 -0
- {liger_kernel_nightly-0.5.0.dev20241210032000 → liger_kernel_nightly-0.5.1.dev20241210093009}/src/liger_kernel/transformers/geglu.py +0 -0
- {liger_kernel_nightly-0.5.0.dev20241210032000 → liger_kernel_nightly-0.5.1.dev20241210093009}/src/liger_kernel/transformers/group_norm.py +0 -0
- {liger_kernel_nightly-0.5.0.dev20241210032000 → liger_kernel_nightly-0.5.1.dev20241210093009}/src/liger_kernel/transformers/jsd.py +0 -0
- {liger_kernel_nightly-0.5.0.dev20241210032000 → liger_kernel_nightly-0.5.1.dev20241210093009}/src/liger_kernel/transformers/kl_div.py +0 -0
- {liger_kernel_nightly-0.5.0.dev20241210032000 → liger_kernel_nightly-0.5.1.dev20241210093009}/src/liger_kernel/transformers/layer_norm.py +0 -0
- {liger_kernel_nightly-0.5.0.dev20241210032000 → liger_kernel_nightly-0.5.1.dev20241210093009}/src/liger_kernel/transformers/model/__init__.py +0 -0
- {liger_kernel_nightly-0.5.0.dev20241210032000 → liger_kernel_nightly-0.5.1.dev20241210093009}/src/liger_kernel/transformers/model/gemma.py +0 -0
- {liger_kernel_nightly-0.5.0.dev20241210032000 → liger_kernel_nightly-0.5.1.dev20241210093009}/src/liger_kernel/transformers/model/gemma2.py +0 -0
- {liger_kernel_nightly-0.5.0.dev20241210032000 → liger_kernel_nightly-0.5.1.dev20241210093009}/src/liger_kernel/transformers/model/llama.py +0 -0
- {liger_kernel_nightly-0.5.0.dev20241210032000 → liger_kernel_nightly-0.5.1.dev20241210093009}/src/liger_kernel/transformers/model/mistral.py +0 -0
- {liger_kernel_nightly-0.5.0.dev20241210032000 → liger_kernel_nightly-0.5.1.dev20241210093009}/src/liger_kernel/transformers/model/mixtral.py +0 -0
- {liger_kernel_nightly-0.5.0.dev20241210032000 → liger_kernel_nightly-0.5.1.dev20241210093009}/src/liger_kernel/transformers/model/mllama.py +0 -0
- {liger_kernel_nightly-0.5.0.dev20241210032000 → liger_kernel_nightly-0.5.1.dev20241210093009}/src/liger_kernel/transformers/model/phi3.py +0 -0
- {liger_kernel_nightly-0.5.0.dev20241210032000 → liger_kernel_nightly-0.5.1.dev20241210093009}/src/liger_kernel/transformers/model/qwen2.py +0 -0
- {liger_kernel_nightly-0.5.0.dev20241210032000 → liger_kernel_nightly-0.5.1.dev20241210093009}/src/liger_kernel/transformers/model/qwen2_vl.py +0 -0
- {liger_kernel_nightly-0.5.0.dev20241210032000 → liger_kernel_nightly-0.5.1.dev20241210093009}/src/liger_kernel/transformers/monkey_patch.py +0 -0
- {liger_kernel_nightly-0.5.0.dev20241210032000 → liger_kernel_nightly-0.5.1.dev20241210093009}/src/liger_kernel/transformers/qwen2vl_mrope.py +0 -0
- {liger_kernel_nightly-0.5.0.dev20241210032000 → liger_kernel_nightly-0.5.1.dev20241210093009}/src/liger_kernel/transformers/rms_norm.py +0 -0
- {liger_kernel_nightly-0.5.0.dev20241210032000 → liger_kernel_nightly-0.5.1.dev20241210093009}/src/liger_kernel/transformers/rope.py +0 -0
- {liger_kernel_nightly-0.5.0.dev20241210032000 → liger_kernel_nightly-0.5.1.dev20241210093009}/src/liger_kernel/transformers/swiglu.py +0 -0
- {liger_kernel_nightly-0.5.0.dev20241210032000 → liger_kernel_nightly-0.5.1.dev20241210093009}/src/liger_kernel/transformers/trainer_integration.py +0 -0
- {liger_kernel_nightly-0.5.0.dev20241210032000 → liger_kernel_nightly-0.5.1.dev20241210093009}/src/liger_kernel/triton/__init__.py +0 -0
- {liger_kernel_nightly-0.5.0.dev20241210032000 → liger_kernel_nightly-0.5.1.dev20241210093009}/src/liger_kernel/triton/monkey_patch.py +0 -0
- {liger_kernel_nightly-0.5.0.dev20241210032000 → liger_kernel_nightly-0.5.1.dev20241210093009}/src/liger_kernel/utils.py +0 -0
- {liger_kernel_nightly-0.5.0.dev20241210032000 → liger_kernel_nightly-0.5.1.dev20241210093009}/src/liger_kernel_nightly.egg-info/dependency_links.txt +0 -0
- {liger_kernel_nightly-0.5.0.dev20241210032000 → liger_kernel_nightly-0.5.1.dev20241210093009}/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: 0.5.
|
3
|
+
Version: 0.5.1.dev20241210093009
|
4
4
|
Summary: Efficient Triton kernels for LLM Training
|
5
5
|
License: BSD 2-CLAUSE LICENSE
|
6
6
|
Copyright 2024 LinkedIn Corporation
|
@@ -34,9 +34,10 @@ Requires-Dist: torch>=2.1.2
|
|
34
34
|
Requires-Dist: triton>=2.3.1
|
35
35
|
Provides-Extra: transformers
|
36
36
|
Requires-Dist: transformers~=4.0; extra == "transformers"
|
37
|
+
Provides-Extra: trl
|
38
|
+
Requires-Dist: trl>=0.11.0; extra == "trl"
|
37
39
|
Provides-Extra: dev
|
38
40
|
Requires-Dist: transformers>=4.44.2; extra == "dev"
|
39
|
-
Requires-Dist: trl>=0.11.0; extra == "dev"
|
40
41
|
Requires-Dist: matplotlib>=3.7.2; extra == "dev"
|
41
42
|
Requires-Dist: flake8>=4.0.1.1; extra == "dev"
|
42
43
|
Requires-Dist: black>=24.4.2; extra == "dev"
|
@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
|
|
4
4
|
|
5
5
|
[project]
|
6
6
|
name = "liger_kernel_nightly"
|
7
|
-
version = "0.5.
|
7
|
+
version = "0.5.1.dev20241210093009"
|
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" }
|
@@ -20,9 +20,12 @@ transformers = [
|
|
20
20
|
"transformers~=4.0"
|
21
21
|
]
|
22
22
|
|
23
|
+
trl = [
|
24
|
+
"trl>=0.11.0",
|
25
|
+
]
|
26
|
+
|
23
27
|
dev = [
|
24
28
|
"transformers>=4.44.2",
|
25
|
-
"trl>=0.11.0",
|
26
29
|
"matplotlib>=3.7.2",
|
27
30
|
"flake8>=4.0.1.1",
|
28
31
|
"black>=24.4.2",
|
@@ -22,7 +22,6 @@ from liger_kernel.transformers.monkey_patch import ( # noqa: F401
|
|
22
22
|
apply_liger_kernel_to_qwen2,
|
23
23
|
apply_liger_kernel_to_qwen2_vl,
|
24
24
|
)
|
25
|
-
from liger_kernel.transformers.orpo_trainer import LigerORPOTrainer # noqa: F401
|
26
25
|
from liger_kernel.transformers.rms_norm import LigerRMSNorm # noqa: F401
|
27
26
|
from liger_kernel.transformers.rope import liger_rotary_pos_emb # noqa: F401
|
28
27
|
from liger_kernel.transformers.swiglu import ( # noqa: F401
|
@@ -76,9 +76,7 @@ class LigerORPOTrainer(ORPOTrainer):
|
|
76
76
|
padding_value=self.padding_value,
|
77
77
|
device=self.accelerator.device,
|
78
78
|
)
|
79
|
-
|
80
|
-
# import pdb; pdb.set_trace()
|
81
|
-
# torch.distributed.barrier()
|
79
|
+
|
82
80
|
model_kwargs = (
|
83
81
|
{
|
84
82
|
"decoder_input_ids": self._shift_right(
|
@@ -1,6 +1,6 @@
|
|
1
1
|
Metadata-Version: 2.1
|
2
2
|
Name: liger_kernel_nightly
|
3
|
-
Version: 0.5.
|
3
|
+
Version: 0.5.1.dev20241210093009
|
4
4
|
Summary: Efficient Triton kernels for LLM Training
|
5
5
|
License: BSD 2-CLAUSE LICENSE
|
6
6
|
Copyright 2024 LinkedIn Corporation
|
@@ -34,9 +34,10 @@ Requires-Dist: torch>=2.1.2
|
|
34
34
|
Requires-Dist: triton>=2.3.1
|
35
35
|
Provides-Extra: transformers
|
36
36
|
Requires-Dist: transformers~=4.0; extra == "transformers"
|
37
|
+
Provides-Extra: trl
|
38
|
+
Requires-Dist: trl>=0.11.0; extra == "trl"
|
37
39
|
Provides-Extra: dev
|
38
40
|
Requires-Dist: transformers>=4.44.2; extra == "dev"
|
39
|
-
Requires-Dist: trl>=0.11.0; extra == "dev"
|
40
41
|
Requires-Dist: matplotlib>=3.7.2; extra == "dev"
|
41
42
|
Requires-Dist: flake8>=4.0.1.1; extra == "dev"
|
42
43
|
Requires-Dist: black>=24.4.2; extra == "dev"
|
@@ -41,7 +41,6 @@ src/liger_kernel/transformers/jsd.py
|
|
41
41
|
src/liger_kernel/transformers/kl_div.py
|
42
42
|
src/liger_kernel/transformers/layer_norm.py
|
43
43
|
src/liger_kernel/transformers/monkey_patch.py
|
44
|
-
src/liger_kernel/transformers/orpo_trainer.py
|
45
44
|
src/liger_kernel/transformers/qwen2vl_mrope.py
|
46
45
|
src/liger_kernel/transformers/rms_norm.py
|
47
46
|
src/liger_kernel/transformers/rope.py
|
@@ -58,6 +57,8 @@ src/liger_kernel/transformers/model/mllama.py
|
|
58
57
|
src/liger_kernel/transformers/model/phi3.py
|
59
58
|
src/liger_kernel/transformers/model/qwen2.py
|
60
59
|
src/liger_kernel/transformers/model/qwen2_vl.py
|
60
|
+
src/liger_kernel/transformers/trainer/__init__.py
|
61
|
+
src/liger_kernel/transformers/trainer/orpo_trainer.py
|
61
62
|
src/liger_kernel/triton/__init__.py
|
62
63
|
src/liger_kernel/triton/monkey_patch.py
|
63
64
|
src/liger_kernel_nightly.egg-info/PKG-INFO
|
File without changes
|
{liger_kernel_nightly-0.5.0.dev20241210032000 → liger_kernel_nightly-0.5.1.dev20241210093009}/NOTICE
RENAMED
File without changes
|
File without changes
|
File without changes
|
File without changes
|
File without changes
|
File without changes
|
File without changes
|
File without changes
|
File without changes
|
File without changes
|
File without changes
|
File without changes
|
File without changes
|
File without changes
|
File without changes
|
File without changes
|
File without changes
|
File without changes
|
File without changes
|
File without changes
|
File without changes
|
File without changes
|
File without changes
|
File without changes
|
File without changes
|
File without changes
|
File without changes
|
File without changes
|
File without changes
|
File without changes
|
File without changes
|
File without changes
|
File without changes
|
File without changes
|
File without changes
|
File without changes
|
File without changes
|
File without changes
|
File without changes
|
File without changes
|
File without changes
|
File without changes
|
File without changes
|
File without changes
|
File without changes
|
File without changes
|
File without changes
|
File without changes
|
File without changes
|
File without changes
|
File without changes
|
File without changes
|
File without changes
|
File without changes
|
File without changes
|
File without changes
|
File without changes
|
File without changes
|
File without changes
|
File without changes
|