liger-kernel-nightly 0.4.2.dev20241209195727__tar.gz → 0.4.2.dev20241209195823__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-0.4.2.dev20241209195727/src/liger_kernel_nightly.egg-info → liger_kernel_nightly-0.4.2.dev20241209195823}/PKG-INFO +1 -1
- {liger_kernel_nightly-0.4.2.dev20241209195727 → liger_kernel_nightly-0.4.2.dev20241209195823}/pyproject.toml +1 -1
- {liger_kernel_nightly-0.4.2.dev20241209195727 → liger_kernel_nightly-0.4.2.dev20241209195823}/src/liger_kernel/chunked_loss/fused_linear_distillation.py +3 -1
- {liger_kernel_nightly-0.4.2.dev20241209195727 → liger_kernel_nightly-0.4.2.dev20241209195823}/src/liger_kernel/env_report.py +4 -5
- {liger_kernel_nightly-0.4.2.dev20241209195727 → liger_kernel_nightly-0.4.2.dev20241209195823/src/liger_kernel_nightly.egg-info}/PKG-INFO +1 -1
- {liger_kernel_nightly-0.4.2.dev20241209195727 → liger_kernel_nightly-0.4.2.dev20241209195823}/LICENSE +0 -0
- {liger_kernel_nightly-0.4.2.dev20241209195727 → liger_kernel_nightly-0.4.2.dev20241209195823}/NOTICE +0 -0
- {liger_kernel_nightly-0.4.2.dev20241209195727 → liger_kernel_nightly-0.4.2.dev20241209195823}/README.md +0 -0
- {liger_kernel_nightly-0.4.2.dev20241209195727 → liger_kernel_nightly-0.4.2.dev20241209195823}/setup.cfg +0 -0
- {liger_kernel_nightly-0.4.2.dev20241209195727 → liger_kernel_nightly-0.4.2.dev20241209195823}/src/liger_kernel/__init__.py +0 -0
- {liger_kernel_nightly-0.4.2.dev20241209195727 → liger_kernel_nightly-0.4.2.dev20241209195823}/src/liger_kernel/chunked_loss/__init__.py +0 -0
- {liger_kernel_nightly-0.4.2.dev20241209195727 → liger_kernel_nightly-0.4.2.dev20241209195823}/src/liger_kernel/chunked_loss/cpo_loss.py +0 -0
- {liger_kernel_nightly-0.4.2.dev20241209195727 → liger_kernel_nightly-0.4.2.dev20241209195823}/src/liger_kernel/chunked_loss/dpo_loss.py +0 -0
- {liger_kernel_nightly-0.4.2.dev20241209195727 → liger_kernel_nightly-0.4.2.dev20241209195823}/src/liger_kernel/chunked_loss/functional.py +0 -0
- {liger_kernel_nightly-0.4.2.dev20241209195727 → liger_kernel_nightly-0.4.2.dev20241209195823}/src/liger_kernel/chunked_loss/fused_linear_preference.py +0 -0
- {liger_kernel_nightly-0.4.2.dev20241209195727 → liger_kernel_nightly-0.4.2.dev20241209195823}/src/liger_kernel/chunked_loss/orpo_loss.py +0 -0
- {liger_kernel_nightly-0.4.2.dev20241209195727 → liger_kernel_nightly-0.4.2.dev20241209195823}/src/liger_kernel/chunked_loss/simpo_loss.py +0 -0
- {liger_kernel_nightly-0.4.2.dev20241209195727 → liger_kernel_nightly-0.4.2.dev20241209195823}/src/liger_kernel/ops/__init__.py +0 -0
- {liger_kernel_nightly-0.4.2.dev20241209195727 → liger_kernel_nightly-0.4.2.dev20241209195823}/src/liger_kernel/ops/cross_entropy.py +0 -0
- {liger_kernel_nightly-0.4.2.dev20241209195727 → liger_kernel_nightly-0.4.2.dev20241209195823}/src/liger_kernel/ops/experimental/embedding.py +0 -0
- {liger_kernel_nightly-0.4.2.dev20241209195727 → liger_kernel_nightly-0.4.2.dev20241209195823}/src/liger_kernel/ops/experimental/mm_int8int2.py +0 -0
- {liger_kernel_nightly-0.4.2.dev20241209195727 → liger_kernel_nightly-0.4.2.dev20241209195823}/src/liger_kernel/ops/fused_linear_cross_entropy.py +0 -0
- {liger_kernel_nightly-0.4.2.dev20241209195727 → liger_kernel_nightly-0.4.2.dev20241209195823}/src/liger_kernel/ops/fused_linear_jsd.py +0 -0
- {liger_kernel_nightly-0.4.2.dev20241209195727 → liger_kernel_nightly-0.4.2.dev20241209195823}/src/liger_kernel/ops/geglu.py +0 -0
- {liger_kernel_nightly-0.4.2.dev20241209195727 → liger_kernel_nightly-0.4.2.dev20241209195823}/src/liger_kernel/ops/group_norm.py +0 -0
- {liger_kernel_nightly-0.4.2.dev20241209195727 → liger_kernel_nightly-0.4.2.dev20241209195823}/src/liger_kernel/ops/jsd.py +0 -0
- {liger_kernel_nightly-0.4.2.dev20241209195727 → liger_kernel_nightly-0.4.2.dev20241209195823}/src/liger_kernel/ops/kl_div.py +0 -0
- {liger_kernel_nightly-0.4.2.dev20241209195727 → liger_kernel_nightly-0.4.2.dev20241209195823}/src/liger_kernel/ops/layer_norm.py +0 -0
- {liger_kernel_nightly-0.4.2.dev20241209195727 → liger_kernel_nightly-0.4.2.dev20241209195823}/src/liger_kernel/ops/qwen2vl_mrope.py +0 -0
- {liger_kernel_nightly-0.4.2.dev20241209195727 → liger_kernel_nightly-0.4.2.dev20241209195823}/src/liger_kernel/ops/rms_norm.py +0 -0
- {liger_kernel_nightly-0.4.2.dev20241209195727 → liger_kernel_nightly-0.4.2.dev20241209195823}/src/liger_kernel/ops/rope.py +0 -0
- {liger_kernel_nightly-0.4.2.dev20241209195727 → liger_kernel_nightly-0.4.2.dev20241209195823}/src/liger_kernel/ops/swiglu.py +0 -0
- {liger_kernel_nightly-0.4.2.dev20241209195727 → liger_kernel_nightly-0.4.2.dev20241209195823}/src/liger_kernel/ops/utils.py +0 -0
- {liger_kernel_nightly-0.4.2.dev20241209195727 → liger_kernel_nightly-0.4.2.dev20241209195823}/src/liger_kernel/transformers/__init__.py +0 -0
- {liger_kernel_nightly-0.4.2.dev20241209195727 → liger_kernel_nightly-0.4.2.dev20241209195823}/src/liger_kernel/transformers/auto_model.py +0 -0
- {liger_kernel_nightly-0.4.2.dev20241209195727 → liger_kernel_nightly-0.4.2.dev20241209195823}/src/liger_kernel/transformers/cross_entropy.py +0 -0
- {liger_kernel_nightly-0.4.2.dev20241209195727 → liger_kernel_nightly-0.4.2.dev20241209195823}/src/liger_kernel/transformers/experimental/embedding.py +0 -0
- {liger_kernel_nightly-0.4.2.dev20241209195727 → liger_kernel_nightly-0.4.2.dev20241209195823}/src/liger_kernel/transformers/functional.py +0 -0
- {liger_kernel_nightly-0.4.2.dev20241209195727 → liger_kernel_nightly-0.4.2.dev20241209195823}/src/liger_kernel/transformers/fused_linear_cross_entropy.py +0 -0
- {liger_kernel_nightly-0.4.2.dev20241209195727 → liger_kernel_nightly-0.4.2.dev20241209195823}/src/liger_kernel/transformers/fused_linear_jsd.py +0 -0
- {liger_kernel_nightly-0.4.2.dev20241209195727 → liger_kernel_nightly-0.4.2.dev20241209195823}/src/liger_kernel/transformers/geglu.py +0 -0
- {liger_kernel_nightly-0.4.2.dev20241209195727 → liger_kernel_nightly-0.4.2.dev20241209195823}/src/liger_kernel/transformers/group_norm.py +0 -0
- {liger_kernel_nightly-0.4.2.dev20241209195727 → liger_kernel_nightly-0.4.2.dev20241209195823}/src/liger_kernel/transformers/jsd.py +0 -0
- {liger_kernel_nightly-0.4.2.dev20241209195727 → liger_kernel_nightly-0.4.2.dev20241209195823}/src/liger_kernel/transformers/kl_div.py +0 -0
- {liger_kernel_nightly-0.4.2.dev20241209195727 → liger_kernel_nightly-0.4.2.dev20241209195823}/src/liger_kernel/transformers/layer_norm.py +0 -0
- {liger_kernel_nightly-0.4.2.dev20241209195727 → liger_kernel_nightly-0.4.2.dev20241209195823}/src/liger_kernel/transformers/model/__init__.py +0 -0
- {liger_kernel_nightly-0.4.2.dev20241209195727 → liger_kernel_nightly-0.4.2.dev20241209195823}/src/liger_kernel/transformers/model/gemma.py +0 -0
- {liger_kernel_nightly-0.4.2.dev20241209195727 → liger_kernel_nightly-0.4.2.dev20241209195823}/src/liger_kernel/transformers/model/gemma2.py +0 -0
- {liger_kernel_nightly-0.4.2.dev20241209195727 → liger_kernel_nightly-0.4.2.dev20241209195823}/src/liger_kernel/transformers/model/llama.py +0 -0
- {liger_kernel_nightly-0.4.2.dev20241209195727 → liger_kernel_nightly-0.4.2.dev20241209195823}/src/liger_kernel/transformers/model/mistral.py +0 -0
- {liger_kernel_nightly-0.4.2.dev20241209195727 → liger_kernel_nightly-0.4.2.dev20241209195823}/src/liger_kernel/transformers/model/mixtral.py +0 -0
- {liger_kernel_nightly-0.4.2.dev20241209195727 → liger_kernel_nightly-0.4.2.dev20241209195823}/src/liger_kernel/transformers/model/mllama.py +0 -0
- {liger_kernel_nightly-0.4.2.dev20241209195727 → liger_kernel_nightly-0.4.2.dev20241209195823}/src/liger_kernel/transformers/model/phi3.py +0 -0
- {liger_kernel_nightly-0.4.2.dev20241209195727 → liger_kernel_nightly-0.4.2.dev20241209195823}/src/liger_kernel/transformers/model/qwen2.py +0 -0
- {liger_kernel_nightly-0.4.2.dev20241209195727 → liger_kernel_nightly-0.4.2.dev20241209195823}/src/liger_kernel/transformers/model/qwen2_vl.py +0 -0
- {liger_kernel_nightly-0.4.2.dev20241209195727 → liger_kernel_nightly-0.4.2.dev20241209195823}/src/liger_kernel/transformers/monkey_patch.py +0 -0
- {liger_kernel_nightly-0.4.2.dev20241209195727 → liger_kernel_nightly-0.4.2.dev20241209195823}/src/liger_kernel/transformers/orpo_trainer.py +0 -0
- {liger_kernel_nightly-0.4.2.dev20241209195727 → liger_kernel_nightly-0.4.2.dev20241209195823}/src/liger_kernel/transformers/qwen2vl_mrope.py +0 -0
- {liger_kernel_nightly-0.4.2.dev20241209195727 → liger_kernel_nightly-0.4.2.dev20241209195823}/src/liger_kernel/transformers/rms_norm.py +0 -0
- {liger_kernel_nightly-0.4.2.dev20241209195727 → liger_kernel_nightly-0.4.2.dev20241209195823}/src/liger_kernel/transformers/rope.py +0 -0
- {liger_kernel_nightly-0.4.2.dev20241209195727 → liger_kernel_nightly-0.4.2.dev20241209195823}/src/liger_kernel/transformers/swiglu.py +0 -0
- {liger_kernel_nightly-0.4.2.dev20241209195727 → liger_kernel_nightly-0.4.2.dev20241209195823}/src/liger_kernel/transformers/trainer_integration.py +0 -0
- {liger_kernel_nightly-0.4.2.dev20241209195727 → liger_kernel_nightly-0.4.2.dev20241209195823}/src/liger_kernel/triton/__init__.py +0 -0
- {liger_kernel_nightly-0.4.2.dev20241209195727 → liger_kernel_nightly-0.4.2.dev20241209195823}/src/liger_kernel/triton/monkey_patch.py +0 -0
- {liger_kernel_nightly-0.4.2.dev20241209195727 → liger_kernel_nightly-0.4.2.dev20241209195823}/src/liger_kernel/utils.py +0 -0
- {liger_kernel_nightly-0.4.2.dev20241209195727 → liger_kernel_nightly-0.4.2.dev20241209195823}/src/liger_kernel_nightly.egg-info/SOURCES.txt +0 -0
- {liger_kernel_nightly-0.4.2.dev20241209195727 → liger_kernel_nightly-0.4.2.dev20241209195823}/src/liger_kernel_nightly.egg-info/dependency_links.txt +0 -0
- {liger_kernel_nightly-0.4.2.dev20241209195727 → liger_kernel_nightly-0.4.2.dev20241209195823}/src/liger_kernel_nightly.egg-info/requires.txt +0 -0
- {liger_kernel_nightly-0.4.2.dev20241209195727 → liger_kernel_nightly-0.4.2.dev20241209195823}/src/liger_kernel_nightly.egg-info/top_level.txt +0 -0
@@ -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.
|
7
|
+
version = "0.4.2.dev20241209195823"
|
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" }
|
@@ -105,7 +105,9 @@ class LigerFusedLinearDistillationBase(torch.autograd.Function):
|
|
105
105
|
|
106
106
|
hard_loss /= full_target.shape[0]
|
107
107
|
|
108
|
-
soft_loss = distillation_loss_fn(
|
108
|
+
soft_loss = distillation_loss_fn(
|
109
|
+
student_logits_chunk, teacher_logits_chunk, temperature
|
110
|
+
)
|
109
111
|
soft_loss /= full_target.shape[0]
|
110
112
|
|
111
113
|
loss = weight_hard_loss * hard_loss + weight_soft_loss * soft_loss
|
@@ -24,7 +24,7 @@ def print_env_report():
|
|
24
24
|
cuda_version = (
|
25
25
|
torch.version.cuda if torch.cuda.is_available() else "Not available"
|
26
26
|
)
|
27
|
-
print(f"CUDA version: {cuda_version}")
|
27
|
+
print(f"CUDA version: {cuda_version}")
|
28
28
|
except ImportError:
|
29
29
|
print("PyTorch: Not installed")
|
30
30
|
print("CUDA version: Unable to query")
|
@@ -42,15 +42,14 @@ def print_env_report():
|
|
42
42
|
print(f"Transformers version: {transformers.__version__}")
|
43
43
|
except ImportError:
|
44
44
|
print("Transformers: Not installed")
|
45
|
-
|
45
|
+
|
46
46
|
try:
|
47
47
|
xpu_version = (
|
48
48
|
torch.version.xpu if torch.xpu.is_available() else "XPU Not Available"
|
49
49
|
)
|
50
|
-
print(f"XPU version: {xpu_version}")
|
50
|
+
print(f"XPU version: {xpu_version}")
|
51
51
|
except ImportError:
|
52
|
-
print("XPU version: Unable to query")
|
53
|
-
|
52
|
+
print("XPU version: Unable to query")
|
54
53
|
|
55
54
|
|
56
55
|
if __name__ == "__main__":
|
File without changes
|
{liger_kernel_nightly-0.4.2.dev20241209195727 → liger_kernel_nightly-0.4.2.dev20241209195823}/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
|
File without changes
|
File without changes
|