huggingface/candle
1
0
Fork 0
mirror of https://github.com/huggingface/candle.git synced 2025-06-20 04:00:28 +00:00
Code Issues Packages Projects Releases Wiki Activity

Support for attention bias in gemma + refactor things a bit. (#1744)

Browse Source 
* Support for attention bias in gemma + refactor things a bit.

* Fix the cuda tests.
This commit is contained in:
Laurent Mazare
2024-02-22 09:35:28 +01:00
committed by GitHub
parent 8013b50829
commit c753f72c85
8 changed files with 62 additions and 88 deletions
Download Patch File Download Diff File Expand all files Collapse all files

12
candle-transformers/src/models/bigcode.rs
View File

@ -1,15 +1,5 @@
use candle::{DType, Device, IndexOp, Result, Tensor, D};
use candle_nn::{embedding, Embedding, LayerNorm, Linear, Module, VarBuilder};
fn linear(size1: usize, size2: usize, bias: bool, vb: VarBuilder) -> Result<Linear> {
let weight = vb.get((size2, size1), "weight")?;
let bias = if bias {
Some(vb.get(size2, "bias")?)
} else {
None
};
Ok(Linear::new(weight, bias))
}
use candle_nn::{embedding, linear_b as linear, Embedding, LayerNorm, Linear, Module, VarBuilder};
fn layer_norm(size: usize, eps: f64, vb: VarBuilder) -> Result<LayerNorm> {
let weight = vb.get(size, "weight")?;