huggingface/candle
1
0
Fork 0
mirror of https://github.com/huggingface/candle.git synced 2025-06-17 02:58:50 +00:00
Code Issues Packages Projects Releases Wiki Activity

Merge pull request #469 from huggingface/fix_llama_v1

Browse Source 
Fixing llamav1
This commit is contained in:
Nicolas Patry
2023-08-16 17:47:40 +02:00
committed by GitHub
parent 2e206e269d 102fa4c2e3
commit fa4590d7fd
1 changed files with 2 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
candle-examples/examples/llama/model.rs
View File

@ -13,7 +13,7 @@ pub struct LlamaConfig {
pub vocab_size: usize, pub vocab_size: usize,
pub num_hidden_layers: usize, pub num_hidden_layers: usize,
pub num_attention_heads: usize, pub num_attention_heads: usize,
pub num_key_value_heads: usize, pub num_key_value_heads: Option<usize>,
pub rms_norm_eps: f64, pub rms_norm_eps: f64,
} }
@ -25,7 +25,7 @@ impl LlamaConfig {
vocab_size: self.vocab_size, vocab_size: self.vocab_size,
num_hidden_layers: self.num_hidden_layers, num_hidden_layers: self.num_hidden_layers,
num_attention_heads: self.num_attention_heads, num_attention_heads: self.num_attention_heads,
num_key_value_heads: self.num_key_value_heads, num_key_value_heads: self.num_key_value_heads.unwrap_or(self.num_attention_heads),
rms_norm_eps: self.rms_norm_eps, rms_norm_eps: self.rms_norm_eps,
use_flash_attn, use_flash_attn,
} }