Merge pull request #89 from LaurentMazare/extending_bert

Enabling `roberta` for the example (it's the same model as Bert, with just different naming.)

candle-examples/examples/bert/main.rs

@@ -9,7 +9,7 @@ use candle_hub::{api::Api, Cache, Repo, RepoType};
 use clap::Parser;
 use serde::Deserialize;
 use std::collections::HashMap;
-use tokenizers::Tokenizer;
+use tokenizers::{PaddingParams, Tokenizer};
 
 const DTYPE: DType = DType::F32;
 
@@ -89,9 +89,10 @@ impl HiddenAct {
     }
 }
 
-#[derive(Debug, Clone, Copy, PartialEq, Eq, Deserialize)]
+#[derive(Debug, Clone, Copy, PartialEq, Eq, Deserialize, Default)]
 #[serde(rename_all = "lowercase")]
 enum PositionEmbeddingType {
+    #[default]
     Absolute,
 }
 
@@ -110,9 +111,12 @@ struct Config {
     initializer_range: f64,
     layer_norm_eps: f64,
     pad_token_id: usize,
+    #[serde(default)]
     position_embedding_type: PositionEmbeddingType,
+    #[serde(default)]
     use_cache: bool,
     classifier_dropout: Option<f64>,
+    model_type: Option<String>,
 }
 
 impl Default for Config {
@@ -133,6 +137,7 @@ impl Default for Config {
             position_embedding_type: PositionEmbeddingType::Absolute,
             use_cache: true,
             classifier_dropout: None,
+            model_type: Some("bert".to_string()),
         }
     }
 }
@@ -156,6 +161,7 @@ impl Config {
             position_embedding_type: PositionEmbeddingType::Absolute,
             use_cache: true,
             classifier_dropout: None,
+            model_type: Some("bert".to_string()),
         }
     }
 }
@@ -594,12 +600,17 @@ impl BertModel {
         ) {
             (Ok(embeddings), Ok(encoder)) => (embeddings, encoder),
             (Err(err), _) | (_, Err(err)) => {
-                match (
-                    BertEmbeddings::load("bert.embeddings", vb, config),
-                    BertEncoder::load("bert.encoder", vb, config),
-                ) {
-                    (Ok(embeddings), Ok(encoder)) => (embeddings, encoder),
-                    _ => return Err(err),
+                if let Some(model_type) = &config.model_type {
+                    if let (Ok(embeddings), Ok(encoder)) = (
+                        BertEmbeddings::load(&format!("{model_type}.embeddings"), vb, config),
+                        BertEncoder::load(&format!("{model_type}.encoder"), vb, config),
+                    ) {
+                        (embeddings, encoder)
+                    } else {
+                        return Err(err);
+                    }
+                } else {
+                    return Err(err);
                 }
             }
         };
@@ -731,6 +742,12 @@ async fn main() -> Result<()> {
         let n_sentences = sentences.len();
         if let Some(pp) = tokenizer.get_padding_mut() {
             pp.strategy = tokenizers::PaddingStrategy::BatchLongest
+        } else {
+            let pp = PaddingParams {
+                strategy: tokenizers::PaddingStrategy::BatchLongest,
+                ..Default::default()
+            };
+            tokenizer.with_padding(Some(pp));
         }
         let tokens = tokenizer
             .encode_batch(sentences.to_vec(), true)
@@ -742,6 +759,7 @@ async fn main() -> Result<()> {
                 Ok(Tensor::new(tokens.as_slice(), device)?)
             })
             .collect::<Result<Vec<_>>>()?;
+
         let token_ids = Tensor::stack(&token_ids, 0)?;
         let token_type_ids = token_ids.zeros_like()?;
         println!("running inference on batch {:?}", token_ids.shape());