huggingface/candle
1
0
Fork 0
mirror of https://github.com/huggingface/candle.git synced 2025-06-16 18:48:51 +00:00
Code Issues Packages Projects Releases Wiki Activity

fix: qwen2 lm_head loading #2443 (#2445)

Browse Source 
Co-authored-by: Yi Xu <xuyi@me.com>
This commit is contained in:
ilookee
2024-08-23 22:50:02 +08:00
committed by GitHub
parent ccdbe87639
commit fdc2622686
1 changed files with 1 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
candle-transformers/src/models/qwen2.rs
View File

@ -361,7 +361,7 @@ pub struct ModelForCausalLM {
impl ModelForCausalLM { impl ModelForCausalLM {
pub fn new(cfg: &Config, vb: VarBuilder) -> Result<Self> { pub fn new(cfg: &Config, vb: VarBuilder) -> Result<Self> {
let base_model = Model::new(cfg, vb.clone())?; let base_model = Model::new(cfg, vb.clone())?;
let lm_head = if vb.contains_tensor("lm_head") { let lm_head = if vb.contains_tensor("lm_head.weight") {
linear_no_bias(cfg.hidden_size, cfg.vocab_size, vb.pp("lm_head"))? linear_no_bias(cfg.hidden_size, cfg.vocab_size, vb.pp("lm_head"))?
} else { } else {
Linear::from_weights(base_model.embed_tokens.embeddings().clone(), None) Linear::from_weights(base_model.embed_tokens.embeddings().clone(), None)