From 34e9c949015513d05bd4e98c9389847c3e098ad3 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Fran=C3=A7ois=20Hernandez?= Date: Wed, 18 Dec 2024 19:02:26 +0100 Subject: [PATCH] 0.0.3 (#165) --- CHANGELOG.md | 39 ++++++++++++++++++++++++++++++++++++++- README.md | 5 ++--- setup.py | 2 +- 3 files changed, 41 insertions(+), 5 deletions(-) diff --git a/CHANGELOG.md b/CHANGELOG.md index dd29feaa..d9ea5518 100644 --- a/CHANGELOG.md +++ b/CHANGELOG.md @@ -2,6 +2,43 @@ This is just a centralised version of the Github automatically generated Release changelogs. +## 0.0.3 + +* [patch] minor fixes for 0.0.2 by @francoishernandez in https://github.com/eole-nlp/eole/pull/109 +* **Rework handling of special tokens** by @francoishernandez in https://github.com/eole-nlp/eole/pull/45 +* [patch] get_transforms_cls after update_config_with_checkpoint by @francoishernandez in https://github.com/eole-nlp/eole/pull/110 +* [patch] get_transforms_cls after update_config_with_checkpoint BIS by @francoishernandez in https://github.com/eole-nlp/eole/pull/111 +* Updated translator.py to handle updated special token logic when computing alignments by @dameikle in https://github.com/eole-nlp/eole/pull/113 +* clearer log by @vince62s in https://github.com/eole-nlp/eole/pull/112 +* fix training tensor parallel by @vince62s in https://github.com/eole-nlp/eole/pull/115 +* restore all_reduce directly but with detach.clone first - fix #115 by @vince62s in https://github.com/eole-nlp/eole/pull/116 +* **Initial support for Metal Performance Shaders (MPS)** by @dameikle in https://github.com/eole-nlp/eole/pull/98 +* Manage `share_decoder_embeddings` in `convert_HF`, misc fixes and improvements by @francoishernandez in https://github.com/eole-nlp/eole/pull/121 +* Deduce share_decoder_embeddings from HF tie_word_embeddings flag by @francoishernandez in https://github.com/eole-nlp/eole/pull/123 +* [docs] Upgrading docusaurus packages, should fix dependabot warnings by @francoishernandez in https://github.com/eole-nlp/eole/pull/124 +* **add estimator in decoder-only** + clean code by @vince62s in https://github.com/eole-nlp/eole/pull/120 +* fineweb10B/gpt2 recipe, and supporting changes by @francoishernandez in https://github.com/eole-nlp/eole/pull/32 +* enable pure bf16 training by @vince62s in https://github.com/eole-nlp/eole/pull/133 +* Update WMT17 recipe with working tokenization transforms examples by @francoishernandez in https://github.com/eole-nlp/eole/pull/129 +* fixes #131, module 'eole.utils' has no attribute 'distributed' error when training multi-gpu by @isanvicente in https://github.com/eole-nlp/eole/pull/132 +* add estimator in greedy inference by @vince62s in https://github.com/eole-nlp/eole/pull/135 +* Some QOL config/saving improvements by @francoishernandez in https://github.com/eole-nlp/eole/pull/134 +* fix #136. Updated eole/bin/model/average_models.py to work with safetensors model format. by @isanvicente in https://github.com/eole-nlp/eole/pull/137 +* fix head dim in rope by @vince62s in https://github.com/eole-nlp/eole/pull/140 +* fix autocast at scoring when doing AMP by @vince62s in https://github.com/eole-nlp/eole/pull/141 +* Some minor fixes by @francoishernandez in https://github.com/eole-nlp/eole/pull/143 +* fix lora lm head by @vince62s in https://github.com/eole-nlp/eole/pull/142 +* fix missing pad change by @vince62s in https://github.com/eole-nlp/eole/pull/148 +* flash_attn_func does not support padding mask maybe we need to drop a… by @vince62s in https://github.com/eole-nlp/eole/pull/149 +* fix maybe_retranslate when number of newline does not match by @vince62s in https://github.com/eole-nlp/eole/pull/150 +* **Supporting HF tokenizers** by @francoishernandez in https://github.com/eole-nlp/eole/pull/122 +* **Model Validator Recipe** by @francoishernandez in https://github.com/eole-nlp/eole/pull/146 +* apply bytefallback at detok (onmt_tokenize with sentencepiece) by @vince62s in https://github.com/eole-nlp/eole/pull/155 +* patch eos_token_id list handling by @francoishernandez in https://github.com/eole-nlp/eole/pull/158 +* **Compile and Ctranslate2 support** by @vince62s in https://github.com/eole-nlp/eole/pull/161 +* Move predict config update from model loading to config validation by @francoishernandez in https://github.com/eole-nlp/eole/pull/163 +* EuroLLM Gradio (web based) translator 35 languages to 35 languages by @vince62s in https://github.com/eole-nlp/eole/pull/164 + ## 0.0.2 * Refactor position encoding configuration by @vince62s in https://github.com/eole-nlp/eole/pull/60 @@ -66,4 +103,4 @@ This is just a centralised version of the Github automatically generated Release * Add Recipe to train a cometkiwi-like encoder model (which can be used to score sentence pairs) by @vince62s in https://github.com/eole-nlp/eole/pull/53 * Simplify __init__ files, remove some unused code by @francoishernandez in https://github.com/eole-nlp/eole/pull/52 -**Full Changelog**: https://github.com/eole-nlp/eole/commits/0.0.1rc1 \ No newline at end of file +**Full Changelog**: https://github.com/eole-nlp/eole/commits/0.0.1rc1 diff --git a/README.md b/README.md index e05371fb..e5ac5d6a 100644 --- a/README.md +++ b/README.md @@ -53,16 +53,15 @@ To facilitate setup and reproducibility, we provide Docker images via the GitHub You can customize the workflow and build your own images based on specific needs using `build.sh` and `Dockerfile` in the `docker` directory of the repository. -There are two images with CUDA 11.8 and 12.1 prebuilt, change the -cudaXX.X to your desired version when pulling the Docker images To pull the Docker image: ```bash -docker pull ghcr.io/eole-nlp/eole:0.0.2-torch2.3.0-ubuntu22.04-cuda12.1 +docker pull ghcr.io/eole-nlp/eole:0.0.3-torch2.5.1-ubuntu22.04-cuda12.4 ``` Example one-liner to run a container and open a bash shell within it: ```bash -docker run --rm -it --runtime=nvidia ghcr.io/eole-nlp/eole:0.0.2-torch2.3.0-ubuntu22.04-cuda12.1 +docker run --rm -it --runtime=nvidia ghcr.io/eole-nlp/eole:0.0.3-torch2.5.1-ubuntu22.04-cuda12.4 ``` > **Note**: Ensure you have the [Nvidia Container Toolkit](https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/latest/install-guide.html) (formerly nvidia-docker) installed to take advantage of CUDA/GPU features. diff --git a/setup.py b/setup.py index ec890c7c..432421aa 100644 --- a/setup.py +++ b/setup.py @@ -11,7 +11,7 @@ description="Open language modeling toolkit based on PyTorch", long_description=long_description, long_description_content_type="text/markdown", - version="0.0.2", + version="0.0.3", packages=find_packages(), project_urls={ "Source": "https://github.com/eole-nlp/eole/",