-
Notifications
You must be signed in to change notification settings - Fork 0
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
玄界で用いた環境 #3
base: retrieva
Are you sure you want to change the base?
玄界で用いた環境 #3
Conversation
Make API backwards compatible, and add underscore before internally used method See merge request ADLR/megatron-lm!1829
ci: Globally set default MR A100 cluster See merge request ADLR/megatron-lm!1839
Fix TikTokenizer decoding case See merge request ADLR/megatron-lm!1827
Update README.md See merge request ADLR/megatron-lm!1842
tests: Refactor t5 tests See merge request ADLR/megatron-lm!1802
ci: Restart JET on more failure types See merge request ADLR/megatron-lm!1848
ci: Send single failure per message See merge request ADLR/megatron-lm!1847
ci: Deprecate JET flavor See merge request ADLR/megatron-lm!1836
chore: Add datasets owner See merge request ADLR/megatron-lm!1850
…ependent training.
Distributed optimizer support for TE/Apex-independent training. See merge request ADLR/megatron-lm!1751
Fix Encoder-Decoder Pipeline Parallelism Semantics See merge request ADLR/megatron-lm!1819
ci: Fix process groups and flaky tests See merge request ADLR/megatron-lm!1822
tests: Refactor bert tests See merge request ADLR/megatron-lm!1806
Support cpu initialization in rope See merge request ADLR/megatron-lm!1845
…t BaseException:"
Fix all instances of bare "except:" or "except BaseException:" See merge request ADLR/megatron-lm!1756
…hen pipelining occurred
tests: Stop convergence training See merge request ADLR/megatron-lm!2022
ci: CI on CI-branches only on schedule See merge request ADLR/megatron-lm!2023
ci: Clean nodes See merge request ADLR/megatron-lm!2024
ci: Nicer formatting of notifier See merge request ADLR/megatron-lm!2025
ci: Fix H100 label See merge request ADLR/megatron-lm!2028
Multimodal - add ChartQA evaluation See merge request ADLR/megatron-lm!1967
Co-authored-by: Huy Vu2 <huvu@login-eos01.eos.clusters.nvidia.com>
Adding ModelType.encoder_and_decoder to T5 model See merge request ADLR/megatron-lm!2000
Co-authored-by: Mike Chrzanowski <mchrzanowski@draco-oci-dc-01.cm.cluster>
Fix T5 Layer Construction See merge request ADLR/megatron-lm!1989
ci: Check git-version for autoformat.sh See merge request ADLR/megatron-lm!2032
Fix TE versions See merge request ADLR/megatron-lm!1935
Co-authored-by: Shiqing Fan <shiqingf@nvidia.com>
Allgather dispatcher optimization Closes NVIDIA#279 See merge request ADLR/megatron-lm!1884
Co-authored-by: Ali Taghibakhshi <ataghibakhsh@login-eos01.eos.clusters.nvidia.com> Co-authored-by: Mikołaj Błaż <mblaz@nvidia.com>
Alit/dist ckpt mamba See merge request ADLR/megatron-lm!1933
Co-authored-by: 1195343015 <1195343015@qq.com>
Update dist-checkpointing to PyTorch 2.4 See merge request ADLR/megatron-lm!1970
…bject to ShardedTensor Factory
Distributed Checkpointing: Add NonPersistentObject to ShardedTensor Factory See merge request ADLR/megatron-lm!1988
Important Review skippedDraft detected. Please check the settings in the CodeRabbit UI or the You can disable this status message by setting the Thank you for using CodeRabbit. We offer it for free to the OSS community and would appreciate your support in helping us grow. If you find it useful, would you consider giving us a shout-out on your favorite social media? 🪧 TipsChatThere are 3 ways to chat with CodeRabbit:
Note: Be mindful of the bot's finite context window. It's strongly recommended to break down tasks such as reading entire modules into smaller chunks. For a focused discussion, use review comments to chat about specific files and their changes, instead of using the PR comments. CodeRabbit Commands (Invoked using PR comments)
Other keywords and placeholders
CodeRabbit Configuration File (
|
玄界で用いたバージョンをPRの形で保存します。
NVIDIA/Megatron-LMの方に追従し、FlashAttention及びRoPEを使えるようにしました。