From c575be59b22c931bf4afe596497b714aa6c10a91 Mon Sep 17 00:00:00 2001 From: NanoCode012 Date: Tue, 5 Aug 2025 17:18:20 +0700 Subject: [PATCH] feat: update readme instructions to include CCE installation --- examples/devstral/README.md | 8 +++++++- examples/hunyuan/README.md | 12 ++++++++++-- examples/magistral/README.md | 8 +++++++- examples/voxtral/README.md | 3 +++ 4 files changed, 27 insertions(+), 4 deletions(-) diff --git a/examples/devstral/README.md b/examples/devstral/README.md index 1cf2e2cec..5a9b74898 100644 --- a/examples/devstral/README.md +++ b/examples/devstral/README.md @@ -23,7 +23,13 @@ pip3 install packaging==23.2 setuptools==75.8.0 wheel ninja pip3 install --no-build-isolation -e '.[flash-attn]' ``` -2. Run the finetuning example: +2. Install [Cut Cross Entropy](https://docs.axolotl.ai/docs/custom_integrations.html#cut-cross-entropy) to reduce training VRAM usage + +```bash +python scripts/cutcrossentropy_install.py | sh +``` + +3. Run the finetuning example: ```bash axolotl train examples/devstral/devstral-small-qlora.yml diff --git a/examples/hunyuan/README.md b/examples/hunyuan/README.md index 8bf18191c..2a7d89e2b 100644 --- a/examples/hunyuan/README.md +++ b/examples/hunyuan/README.md @@ -13,9 +13,11 @@ Tencent released a family of opensource models called HunYuan with varying param git clone https://github.com/axolotl-ai-cloud/axolotl.git cd axolotl - pip3 install packaging==23.2 setuptools==75.8.0 wheel ninja pip3 install --no-build-isolation -e '.[flash-attn]' + +# Install CCE https://docs.axolotl.ai/docs/custom_integrations.html#cut-cross-entropy +python scripts/cutcrossentropy_install.py | sh ``` 2. Please install HunYuan's [transformers PR](https://github.com/huggingface/transformers/pull/39606) @@ -25,7 +27,13 @@ pip3 uninstall transformers pip3 install git+https://github.com/huggingface/transformers@06b8c1323b366ecb5b8f8d7768f3a8b73e82f4cb ``` -3. Run the finetuning example: +3. Install [Cut Cross Entropy](https://docs.axolotl.ai/docs/custom_integrations.html#cut-cross-entropy) to reduce training VRAM usage + +```bash +python scripts/cutcrossentropy_install.py | sh +``` + +4. Run the finetuning example: ```bash axolotl train examples/hunyuan/hunyuan-v1-dense-qlora.yaml diff --git a/examples/magistral/README.md b/examples/magistral/README.md index 865f872d9..1547654cc 100644 --- a/examples/magistral/README.md +++ b/examples/magistral/README.md @@ -21,7 +21,13 @@ pip3 install packaging==23.2 setuptools==75.8.0 wheel ninja pip3 install --no-build-isolation -e '.[flash-attn]' ``` -2. Run the finetuning example: +2. Install [Cut Cross Entropy](https://docs.axolotl.ai/docs/custom_integrations.html#cut-cross-entropy) to reduce training VRAM usage + +```bash +python scripts/cutcrossentropy_install.py | sh +``` + +3. Run the finetuning example: ```bash axolotl train examples/magistral/magistral-small-qlora.yaml diff --git a/examples/voxtral/README.md b/examples/voxtral/README.md index 669ebbe55..6d939a54f 100644 --- a/examples/voxtral/README.md +++ b/examples/voxtral/README.md @@ -25,6 +25,9 @@ pip3 install --no-build-isolation -e '.[flash-attn]' # audio pip3 install librosa==0.11.0 pip3 install 'mistral_common[audio]==1.8.3' + +# Install CCE https://docs.axolotl.ai/docs/custom_integrations.html#cut-cross-entropy +python scripts/cutcrossentropy_install.py | sh ``` 3. Run the finetuning example: