huggingface/candle
1
0
Fork 0
mirror of https://github.com/huggingface/candle.git synced 2025-06-16 10:38:54 +00:00
Code Issues Packages Projects Releases Wiki Activity

Use flash-attn for mistral. (#1004)

Browse Source
This commit is contained in:
Laurent Mazare
2023-09-30 13:15:10 +02:00
committed by GitHub
parent 87e3a4e175
commit 4021272875
2 changed files with 41 additions and 9 deletions
Download Patch File Download Diff File Expand all files Collapse all files

5
candle-examples/examples/mistral/main.rs
View File

@ -113,6 +113,9 @@ struct Args {
#[arg(long)]
tracing: bool,
#[arg(long)]
use_flash_attn: bool,
#[arg(long)]
prompt: String,
@ -207,7 +210,7 @@ fn main() -> Result<()> {
let tokenizer = Tokenizer::from_file(tokenizer_filename).map_err(E::msg)?;
let start = std::time::Instant::now();
let config = Config::config_7b_v0_1();
let config = Config::config_7b_v0_1(args.use_flash_attn);
let device = candle_examples::device(args.cpu)?;
let dtype = if device.is_cuda() {
DType::BF16