@@ -20,7 +20,7 @@ https://github.com/ARM-software/Tool-Solutions/tree/r25.08
2020
2121### Added
2222- Adds https://github.com/pytorch/pytorch/pull/159859 , a WIP LUT implmentation of bf16 GELU
23- this gives an ~ 8x end-to-end speedup for llama3.2 and an ~ 1.8x speedup for attention (both on 16 threads).
23+ this gives an ~ 8x speedup on GELU and an ~ 1.8x speedup for attention for llama3.2 11B Vision (both on 16 threads).
2424- Adds https://github.com/pytorch/pytorch/pull/158250 , to integrate INT4->BF16 via KleidiAI, with fallback.
2525- Adds https://github.com/pytorch/pytorch/pull/160080 , a VLA PoC for PyTorch.
2626 This includes an optimised SVE implementation of exp().
@@ -35,6 +35,7 @@ https://github.com/ARM-software/Tool-Solutions/tree/r25.08
3535 - TORCH_AO_HASH to 8d4a5d83d7be4d7807feabe38d37704c92d40900, from main, August 1st.
3636 - KLEIDIAI_HASH to 8ca226712975f24f13f71d04cda039a0ee9f9e2f, v1.12 from main.
3737- Update the examples/transformers_llm_text_gen.py to use the new quantizer api Int8DynamicActivationIntxWeightConfig.
38+ - Updares huggingface_hub to 0.34.0.
3839- Deleted torchchat_llm_text_gen.py.
3940- Removed Dockerfile lines cloning TorchChat repo and setting safe.directory.
4041
@@ -43,7 +44,7 @@ https://github.com/ARM-software/Tool-Solutions/tree/r25.08
4344
4445### Fixed
4546- Updates various Python packages to address known vulnerabilies with a high CVSS score:
46- - Updates Transformers to 4.50 as a mitigation for CVE-2025 -2099.
47+ - Updates Transformers to 4.55.2, this also provides a mitigation for CVE-2025 -2099.
4748 - Updates Wheel version to 0.38.0 as a mitigation for CVE-2022 -40898.
4849 - Updates setup-tools to 78.1.1 as a mitigation for CVE-2025 -47273 and CVE-2024 -6345.
4950 - Updates Torchvison to 0.23 to avoid the need to use ` --extra-index-url `
0 commit comments