fix(embed): to_dtype F32 before to_vec1以适应 F16 模型
This commit is contained in:
@@ -49,7 +49,7 @@ impl<'a> MiniCPM4GenerateModel<'a> {
|
|||||||
pub fn embed_text(&mut self, text: &str) -> Result<Vec<f32>> {
|
pub fn embed_text(&mut self, text: &str) -> Result<Vec<f32>> {
|
||||||
let input_ids = self.tokenizer.text_encode(text.to_string(), &self.device)?;
|
let input_ids = self.tokenizer.text_encode(text.to_string(), &self.device)?;
|
||||||
let embedding = self.model.embed(&input_ids)?;
|
let embedding = self.model.embed(&input_ids)?;
|
||||||
let vec: Vec<f32> = embedding.flatten_all()?.to_vec1()?;
|
let vec: Vec<f32> = embedding.flatten_all()?.to_dtype(DType::F32)?.to_vec1()?;
|
||||||
Ok(vec)
|
Ok(vec)
|
||||||
}
|
}
|
||||||
|
|
||||||
|
|||||||
Reference in New Issue
Block a user