Skip to content

Releases: exalsius/diloco-training

Release list

v0.2.1

v0.2.1 Latest
Latest

Choose a tag to compare

View all tags
@srnbckr srnbckr released this 02 Dec 12:39
v0.2.1
df0b781
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

0.2.1 (2025-12-01)

Bug Fixes

  • add incremental log updates about parameter synchronization (#63) (8b04ecd)
  • introduce detailed logging for diloco outer optimizer synchronization (#60) (1b920b0)
Assets 2
Loading

v0.2.0

v0.2.0

Choose a tag to compare

View all tags
@srnbckr srnbckr released this 30 Nov 17:50
v0.2.0
d6b4795
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

0.2.0 (2025-11-30)

Features

  • adjust dockerfiles to support heterogeneous amd/nvidia training (#58) (3cd838b)
Loading

v0.1.0

v0.1.0

Choose a tag to compare

View all tags
@srnbckr srnbckr released this 12 Nov 13:00
v0.1.0
56a4e03
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

0.1.0 (2025-11-12)

Features

  • add a TrainingConfig model and new entrypoint script (6bb6171)
  • add TrainingConfig model and start_training script (3fe2d37)
  • add TrainingConfig model and start_training script (c851033)
  • introduce model_cache_dir and dataset_cache_dir configuration (#56) (4fbf953)

Bug Fixes

  • add deleted uv.lock and install etcd (3e94611)
  • add latest changes from inside the pod (948a21c)
  • add pre-commit as a dev dependency (dd817b6)
  • adjust heterogenous profiling implementation with new training interface (decd023)
  • adjust heterogenous training profiling with new DistributedTraining interface (de75de4)
  • adjust main.py script with TrainingConfig (d86aa48)
  • adjust main.py script with TrainingConfig (54d9e17)
  • ensure that PosixPath is serializable (a7fedab)
  • fix logging (31228ab)
  • handle comma-separated values for experiment tags (82d70cb)
  • initialize sum_local_steps in DistributedTrainer (f97e621)
  • integrate missing parameters into TrainingConfig (3071e5f)
  • introduce model_name variable initializatoin in DistributedTrainer (358b1a2)
  • Makefile: Correctly declare .PHONY in Makefile (3b18fd9)
  • only log to wandb if it is initialized (e3dc41d)
  • only log to wandb if it is initialized (1226eb0)
  • refactor DistributedTrainer to support TrainingConfig (e66659e)
  • refactor DistributedTrainer to support TrainingConfig (70a1d6a)
  • wandb: show real steps on x-axis (#57) (5c8be44)

Documentation

  • add Apache 2.0 license (2c879f1)
  • add example config and torchrun script (fecc8cc)
  • add heterogenous gpu support highlight (c00cd37)
  • add hf model upload params to example config (303222c)
  • fix whitespaces in license (bebd3b1)
  • remove unnessary prerequiste section (c9753ca)
  • rewrite readme (448c853)
Loading