d9d 0.1.0__tar.gz → 0.1.1__tar.gz

{d9d-0.1.0 → d9d-0.1.1}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: d9d
-Version: 0.1.0
+Version: 0.1.1
 Summary: d9d - d[istribute]d - distributed training framework based on PyTorch that tries to be efficient yet hackable
 License: Apache-2.0
 Author: Maksim Afanasyev
@@ -38,12 +38,32 @@ Project-URL: Issues, https://github.com/d9d-project/d9d/issues
 Project-URL: Repository, https://github.com/d9d-project/d9d
 Description-Content-Type: text/markdown
 
+---
+title: Home
+---
+
 # The d9d Project
 
 **d9d** is a distributed training framework built on top of PyTorch 2.0. It aims to be hackable, modular, and efficient, designed to scale from single-GPU debugging to massive clusters running 6D-Parallelism.
 
 [LET'S START TRAINING 🚀](https://d9d-project.github.io/d9d/)
 
+## Installation
+
+Just use your favourite package manager:
+```bash
+pip install d9d
+poetry add d9d
+uv add d9d
+```
+
+### Extras
+
+* `d9d[aim]`: [Aim](https://aimstack.io/) experiment tracker integration.
+* `d9d[visualization]`: Plotting libraries required to some advanced visualization functionality.
+* `d9d[moe]`: Efficient Mixture of Experts GPU kernels. You should build and install some dependencies manually before installation: [DeepEP](https://github.com/deepseek-ai/DeepEP), [grouped-gemm](https://github.com/fanshiqing/grouped_gemm/).
+* `d9d[cce]`: Efficient Fused Cross-Entropy kernels. You should build and install some dependencies manually before installation: [Cut Cross Entropy](https://github.com/apple/ml-cross-entropy).
+
 ## Why another framework?
 
 Distributed training frameworks such as **Megatron-LM** are monolithic in the way you run a script from the command line to train any of a set of *predefined* models, using *predefined* regimes. While powerful, these systems can be difficult to hack and integrate into novel research workflows. Their focus is often on providing a complete, end-to-end solution, which can limit flexibility for experimentally-driven research.

{d9d-0.1.0 → d9d-0.1.1}/README.md

@@ -1,9 +1,29 @@
+---
+title: Home
+---
+
 # The d9d Project
 
 **d9d** is a distributed training framework built on top of PyTorch 2.0. It aims to be hackable, modular, and efficient, designed to scale from single-GPU debugging to massive clusters running 6D-Parallelism.
 
 [LET'S START TRAINING 🚀](https://d9d-project.github.io/d9d/)
 
+## Installation
+
+Just use your favourite package manager:
+```bash
+pip install d9d
+poetry add d9d
+uv add d9d
+```
+
+### Extras
+
+* `d9d[aim]`: [Aim](https://aimstack.io/) experiment tracker integration.
+* `d9d[visualization]`: Plotting libraries required to some advanced visualization functionality.
+* `d9d[moe]`: Efficient Mixture of Experts GPU kernels. You should build and install some dependencies manually before installation: [DeepEP](https://github.com/deepseek-ai/DeepEP), [grouped-gemm](https://github.com/fanshiqing/grouped_gemm/).
+* `d9d[cce]`: Efficient Fused Cross-Entropy kernels. You should build and install some dependencies manually before installation: [Cut Cross Entropy](https://github.com/apple/ml-cross-entropy).
+
 ## Why another framework?
 
 Distributed training frameworks such as **Megatron-LM** are monolithic in the way you run a script from the command line to train any of a set of *predefined* models, using *predefined* regimes. While powerful, these systems can be difficult to hack and integrate into novel research workflows. Their focus is often on providing a complete, end-to-end solution, which can limit flexibility for experimentally-driven research.

{d9d-0.1.0 → d9d-0.1.1}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "d9d"
-version = "0.1.0"
+version = "0.1.1"
 description = "d9d - d[istribute]d - distributed training framework based on PyTorch that tries to be efficient yet hackable"
 authors = [
     {name = "Maksim Afanasyev", email = "mr.applexz@gmail.com"}
@@ -66,9 +66,9 @@ cce = ["cut-cross-entropy"]
 ruff = "^0.14.13"
 mypy = "^1.19.1"
 pre-commit = "^4.5.1"
+python-semantic-release = "^10.5.3"
 
 
-# Group containing dependencies for running tests
 [tool.poetry.group.test.dependencies]
 pytest = "^9.0.2"
 coverage = "^7.13.1"
@@ -78,7 +78,6 @@ matplotlib = "^3.10.8"  # for triton benchmarks
 transformers = "^4.57.6"  # for comparing models
 
 
-# Group containing dependencies for building documentation website
 [tool.poetry.group.docs.dependencies]
 mkdocs = "^1.6.1"
 mkdocstrings = {extras = ["python"], version = "^1.0.1"}
@@ -94,7 +93,6 @@ deep-ep = {path = "packages/deep_ep-1.2.1+9af0e0d-cp311-cp311-linux_x86_64.whl"}
 nv-grouped-gemm = {path = "packages/nv_grouped_gemm-1.1.4.post8-cp311-cp311-linux_x86_64.whl"}
 
 
-# Group containing dependencies for running example scripts
 [tool.poetry.group.examples.dependencies]
 datasets = "^4.5.0"
 
@@ -265,3 +263,30 @@ module = [
     "d9d.module.block.moe.communications.deepep.*"
 ]
 ignore_errors = true
+
+[tool.semantic_release]
+allow_zero_version = true
+major_on_zero = true
+version_toml = ["pyproject.toml:project.version"]
+build_command = "pip install poetry && poetry build"
+commit_parser = "conventional"
+
+[tool.semantic_release.commit_parser_options]
+minor_tags = ["feat"]
+patch_tags = ["fix", "perf"]
+parse_squash_commits = true
+ignore_merge_commits = true
+
+[tool.semantic_release.changelog]
+exclude_commit_patterns = [
+    '''chore(?:\([^)]*?\))?: .+''',
+    '''ci(?:\([^)]*?\))?: .+''',
+    '''refactor(?:\([^)]*?\))?: .+''',
+    '''style(?:\([^)]*?\))?: .+''',
+    '''test(?:\([^)]*?\))?: .+''',
+    '''build\((?!deps\): .+)''',
+    '''Initial [Cc]ommit.*''',
+]
+
+[tool.semantic_release.remote]
+type = "github"