Create CHANGELOG.md (#314)

* Create CHANGELOG.md

Signed-off-by: Pablo Garay <[email protected]>

* Add entries to CHANGELOG.md

Signed-off-by: Pablo Garay <[email protected]>

* Update CHANGELOG.md

Signed-off-by: Pablo Garay <[email protected]>

* Update CHANGELOG.md

Co-authored-by: oliver könig <[email protected]>
Signed-off-by: Pablo Garay <[email protected]>

* add links

---------

Signed-off-by: Pablo Garay <[email protected]>
Co-authored-by: oliver könig <[email protected]>

Author: pablo-garay

CHANGELOG.md

@@ -0,0 +1,53 @@
+# Changelog
+
+<!-- Next changelog -->
+## NVIDIA Nemo Run 0.5.0
+
+
+- Fix docs warnings [#271](https://github.com/NVIDIA-NeMo/Run/pull/271)
+- Fix docs build [#269](https://github.com/NVIDIA-NeMo/Run/pull/269)
+- Support overlapped srun commands in Slurm Ray [#263](https://github.com/NVIDIA-NeMo/Run/pull/263)
+- Refactor DGXC Lepton data mover: switch to BatchJob with auto cleanup and sleep after every run [#265](https://github.com/NVIDIA-NeMo/Run/pull/265)
+- ci: Fix nemo fw template ref after migrating to new org [#256](https://github.com/NVIDIA-NeMo/Run/pull/256)
+- Enable Nsys gpu device metrics [#257](https://github.com/NVIDIA-NeMo/Run/pull/257)
+- Sync job code in local tunnel for Slurm Ray job [#254](https://github.com/NVIDIA-NeMo/Run/pull/254)
+- Change the create dist job function to support creating a single node [#240](https://github.com/NVIDIA-NeMo/Run/pull/240)
+- Making job names match Run:ai requirements and making errors more descriptive [#255](https://github.com/NVIDIA-NeMo/Run/pull/255)
+- Support for %j in slurm log retrieval [#252](https://github.com/NVIDIA-NeMo/Run/pull/252)
+- Add KubeRay tests for Ray APIs [#249](https://github.com/NVIDIA-NeMo/Run/pull/249)
+- Upgrade skypilot executor with 0.9.2 [#246](https://github.com/NVIDIA-NeMo/Run/pull/246)  
+- Add user scoping for k8s backend and log level support for Ray APIs [#247](https://github.com/NVIDIA-NeMo/Run/pull/247)
+- Update to latest Lepton SDK [#248](https://github.com/NVIDIA-NeMo/Run/pull/248)
+- Add storage mount options to LeptonExecutor [#237](https://github.com/NVIDIA-NeMo/Run/pull/237)
+- Import guard k8s import in Ray Cluster and Job [#245](https://github.com/NVIDIA-NeMo/Run/pull/245)
+- Add RayJob and Slurm support for Ray APIs + integration with run.Experiment [#236](https://github.com/NVIDIA-NeMo/Run/pull/236)
+- ci: Enforce coverage [#238](https://github.com/NVIDIA-NeMo/Run/pull/238)
+- Fix bug with a CLI overwrite [#235](https://github.com/NVIDIA-NeMo/Run/pull/235)
+- Add LeptonExecutor support [#224](https://github.com/NVIDIA-NeMo/Run/pull/224)
+- Add cancel to docker executor [#233](https://github.com/NVIDIA-NeMo/Run/pull/233)
+- Change default log wait timeout to 10s [#232](https://github.com/NVIDIA-NeMo/Run/pull/232)
+- Add RayCluster API with Kuberay support [#222](https://github.com/NVIDIA-NeMo/Run/pull/222)
+- Add sbatch network arg [#230](https://github.com/NVIDIA-NeMo/Run/pull/230)
+- chore: Update package info [#227](https://github.com/NVIDIA-NeMo/Run/pull/227)
+- Add support for job groups for local executor [#220](https://github.com/NVIDIA-NeMo/Run/pull/220)
+- Roll back get_underlying_types change + introduce extract_constituent [#223](https://github.com/NVIDIA-NeMo/Run/pull/223)
+- Fix some bugs for --lazy in CLI [#179](https://github.com/NVIDIA-NeMo/Run/pull/179)
+- Adding support for modern type-hints [#221](https://github.com/NVIDIA-NeMo/Run/pull/221)
+- Fix bug in CLI with calling a factory-fn inside a list [#214](https://github.com/NVIDIA-NeMo/Run/pull/214)
+- Handle more edge cases in --help [#219](https://github.com/NVIDIA-NeMo/Run/pull/219)
+- Add autogenerated API reference content to the documentation [#190](https://github.com/NVIDIA-NeMo/Run/pull/190)
+- Handle Callable in --help to fix nemo llm export --help error [#217](https://github.com/NVIDIA-NeMo/Run/pull/217)
+- Ensure job directory creation for various schedulers [#216](https://github.com/NVIDIA-NeMo/Run/pull/216)
+- Adding support for ForwardRef in CLI [#176](https://github.com/NVIDIA-NeMo/Run/pull/176)
+- Add additional debug to DGXC data mover [#215](https://github.com/NVIDIA-NeMo/Run/pull/215)
+- Handle ctx in entrypoint for experiment [#213](https://github.com/NVIDIA-NeMo/Run/pull/213)
+- zozhang/dgxc executor data mover [#206](https://github.com/NVIDIA-NeMo/Run/pull/206)
+- Add support for YAML, TOML & JSON [#182](https://github.com/NVIDIA-NeMo/Run/pull/182)
+- Add clean mode for experiment to avoid printing any NeMo-Run specific logs [#208](https://github.com/NVIDIA-NeMo/Run/pull/208)
+- Fix seed for torchrun [#209](https://github.com/NVIDIA-NeMo/Run/pull/209)
+- Support torchrun multi node on local executor [#143](https://github.com/NVIDIA-NeMo/Run/pull/143)
+- Add nsys filename param [#205](https://github.com/NVIDIA-NeMo/Run/pull/205)
+- Add DGXCloudExecutor docs and update execution guide [#192](https://github.com/NVIDIA-NeMo/Run/pull/192)
+- Add --cuda-event-trace=false to nsys command [#180](https://github.com/NVIDIA-NeMo/Run/pull/180)
+
+