candle-flash-attn-v1
| Crates.io | candle-flash-attn-v1 |
| lib.rs | candle-flash-attn-v1 |
| version | 0.0.1 |
| created_at | 2025-04-15 11:23:22.869699+00 |
| updated_at | 2025-04-15 11:23:22.869699+00 |
| description | Flash attention V1 layer for the candle ML framework. |
| homepage | https://github.com/huggingface/candle-extensions/candle-flash-attn-v1/ |
| repository | https://github.com/huggingface/candle-extensions/ |
| max_upload_size | |
| id | 1634308 |
| size | 25,601,223 |
Nicolas Patry (Narsil)
documentation
https://docs.rs/candle-flash-attn-v1
README
Candle Flash Attention v1 Layer
Flash Attention v2 does not support Turing GPUs (T4, RTX 2080). This layer can be used in replacement of the official flash attention Candle layer in the meantime.