1 Release published by 1 person

• v0.0.27 v0.0.27: Qwen2 models, Neuron SDK 2.20.2

  published Dec 13, 2024

4 Pull requests merged by 2 people

• Add support for granite models

  #753 merged Jan 2, 2025

• Add PixArt support

  #731 merged Dec 22, 2024

• Cache Qwen 2.5 models

  #751 merged Dec 20, 2024

• Bump TGI router version

  #748 merged Dec 13, 2024

2 Pull requests opened by 2 people

• Qwen 2.5 tutorial

  #749 opened Dec 17, 2024

• Use AWS Neuron sdk 2.21

  #754 opened Dec 23, 2024

5 Issues closed by 2 people

• stablediffusion (sdxl) ip-adapter support

  #718 closed Jan 5, 2025

• `HF_HUB_OFFLINE` environment variable not being honoured for Neuron cache

  #741 closed Dec 30, 2024

• Size mismatch while loading consolidated checkpoints trained with Tensor parallelism for custom LLama Model

  #734 closed Dec 26, 2024

• AttributeError: can't set attribute 'deepspeed_plugin'

  #735 closed Dec 26, 2024

• Could not find a matching NEFF for your HLO in this directory

  #730 closed Dec 9, 2024

3 Issues opened by 3 people

• Support for Llama 3.3 (transformers >= 4.45.0)

  #752 opened Dec 20, 2024

• Document how to use Qwen 2.5

  #750 opened Dec 17, 2024

• Support for Qwen2-VL

  #747 opened Dec 9, 2024

8 Unresolved conversations

Sometimes conversations happen on old items that aren’t yet closed. Here is a list of all the Issues and Pull Requests with unresolved conversations.

• Add support for new Black Forest's model (Flux)

  #676 commented on Dec 16, 2024 • 0 new comments

• Llama 3 8B fine tuning shows nan value as loss

  #660 commented on Dec 18, 2024 • 0 new comments

• Loading compiled fails: `model_type=bert -> transformers` being used in compiled config.

  #744 commented on Jan 3, 2025 • 0 new comments

• [DO NOT MERGE] Restore non-inlined SDXL

  #668 commented on Dec 17, 2024 • 0 new comments

• Examples for E2E Mistral training & deployment

  #673 commented on Dec 17, 2024 • 0 new comments

• Add support for LoRA with GQAQKVColumnParallelLinear

  #690 commented on Dec 17, 2024 • 0 new comments

• Make the Parallelizer class work for inference

  #715 commented on Dec 17, 2024 • 0 new comments

• Loss too big when using TP

  #740 commented on Jan 1, 2025 • 0 new comments