update the modules

ferritin-esm/src/esm2/models/esm2.rs

@@ -6,30 +6,30 @@ use tokenizers::Tokenizer;
 
 #[derive(Deserialize, Clone)]
 pub struct ESM2Config {
-    num_attention_heads: i32,
-    attention_probs_dropout_prob: f32,
-    classifier_dropout: Option<f32>,
-    emb_layer_norm_before: bool,
-    esmfold_config: Option<String>,
-    hidden_act: String,
-    hidden_dropout_prob: f32,
-    hidden_size: i32,
-    initializer_range: f32,
-    intermediate_size: i32,
-    is_folding_model: bool,
-    layer_norm_eps: f32,
-    mask_token_id: i32,
-    max_position_embeddings: i32,
-    model_type: String,
-    num_hidden_layers: i32,
-    pad_token_id: i32,
-    position_embedding_type: String,
-    token_dropout: bool,
-    torch_dtype: String,
-    transformers_version: String,
-    use_cache: bool,
-    vocab_list: Option<Vec<String>>,
-    vocab_size: i32,
+    pub(crate) num_attention_heads: i32,
+    pub(crate) attention_probs_dropout_prob: f32,
+    pub(crate) classifier_dropout: Option<f32>,
+    pub(crate) emb_layer_norm_before: bool,
+    pub(crate) esmfold_config: Option<String>,
+    pub(crate) hidden_act: String,
+    pub(crate) hidden_dropout_prob: f32,
+    pub(crate) hidden_size: i32,
+    pub(crate) initializer_range: f32,
+    pub(crate) intermediate_size: i32,
+    pub(crate) is_folding_model: bool,
+    pub(crate) layer_norm_eps: f32,
+    pub(crate) mask_token_id: i32,
+    pub(crate) max_position_embeddings: i32,
+    pub(crate) model_type: String,
+    pub(crate) num_hidden_layers: i32,
+    pub(crate) pad_token_id: i32,
+    pub(crate) position_embedding_type: String,
+    pub(crate) token_dropout: bool,
+    pub(crate) torch_dtype: String,
+    pub(crate) transformers_version: String,
+    pub(crate) use_cache: bool,
+    pub(crate) vocab_list: Option<Vec<String>>,
+    pub(crate) vocab_size: i32,
 }
 
 impl ESM2Config {

ferritin-esm/src/esm2/models/modules.rs

@@ -376,10 +376,8 @@ pub struct RobertaLMHead {
 
 impl RobertaLMHead {
     pub fn load(vb: VarBuilder, config: &ESM2Config) -> Result<Self> {
-        // Todo: Fix this!
-        let embed_dim = 100;
-        let output_dim = 100;
-        let dense = candle_nn::linear(embed_dim, output_dim, vb.pp("dense"))?;
+        let ESM2Config { hidden_size, .. } = config;
+        let dense = nn::linear(*hidden_size as usize, *hidden_size as usize, vb.pp("dense"))?;
         let layer_norm = ESM1bLayerNorm::load(vb.pp("LayerNorm"), config)?;
 
         Ok(Self { dense, layer_norm })