tocmo0nlord/axolotl
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

Built site for gh-pages

Browse Source
This commit is contained in:
Quarto GHA Workflow Runner
2026-03-03 14:34:09 +00:00
parent a5b2ace18a
commit 35e6a7e228
4 changed files with 244 additions and 259 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
search.json
View File

@@ -5572,7 +5572,7 @@
"href": "docs/models/qwen3-next.html#getting-started",
"title": "Qwen 3 Next",
"section": "Getting started",
"text": "Getting started\n\nInstall Axolotl following the installation guide. You need to install from main as Qwen3-Next is only on nightly or use our latest Docker images.\nHere is an example of how to install from main for pip:\n\n# Ensure you have Pytorch installed (Pytorch 2.6.0 min)\ngit clone https://github.com/axolotl-ai-cloud/axolotl.git\ncd axolotl\n\npip3 install packaging==26.0 setuptools==75.8.0 wheel ninja\npip3 install --no-build-isolation -e '.[flash-attn]'\n\n# Install CCE https://docs.axolotl.ai/docs/custom_integrations.html#cut-cross-entropy\npython scripts/cutcrossentropy_install.py | sh\n\nInstall Qwen3-Next transformers commit\n\npip3 uninstall -y transformers && pip3 install \"git+https://github.com/huggingface/transformers.git@b9282355bea846b54ed850a066901496b19da654\"\n\nInstall FLA for improved performance\n\npip3 uninstall -y causal-conv1d && pip3 install flash-linear-attention==0.3.2\n\nRun the finetuning example:\n\naxolotl train examples/qwen3-next/qwen3-next-80b-a3b-qlora.yaml\nThis config uses about 45.62 GiB VRAM.\nLet us know how it goes. Happy finetuning! 🚀\n\nTIPS\n\nFor inference, you can experiment with temperature: 0.7, top_p: 0.8, top_k: 20, and min_p: 0.\nYou can run a full finetuning by removing the adapter: qlora and load_in_4bit: true from the config. See Multi-GPU section below.\nRead more on how to load your own dataset at docs.\nThe dataset format follows the OpenAI Messages format as seen here.",
"text": "Getting started\n\nInstall Axolotl following the installation guide.\nInstall Cut Cross Entropy to reduce training VRAM usage.\nInstall FLA for improved performance\n\npip3 uninstall -y causal-conv1d && pip3 install flash-linear-attention==0.4.1\n\nRun the finetuning example:\n\naxolotl train examples/qwen3-next/qwen3-next-80b-a3b-qlora.yaml\nThis config uses about ~47 GiB (no target experts) and ~71GiB (target experts) VRAM.\nLet us know how it goes. Happy finetuning! 🚀\n\nTIPS\n\nFor inference, you can experiment with temperature: 0.7, top_p: 0.8, top_k: 20, and min_p: 0.\nYou can run a full finetuning by removing the adapter: qlora and load_in_4bit: true from the config. See Multi-GPU section below.\nRead more on how to load your own dataset at docs.\nThe dataset format follows the OpenAI Messages format as seen here.",
"crumbs": [
"Getting Started",
"Model Guides",