Remove some usages of 'pub' in models (#708)

Author: EricLBuehler

mistralrs-core/src/models/gemma.rs

@@ -423,9 +423,9 @@ pub struct Model {
     norm: RmsNorm,
     lm_head: Arc<dyn QuantMethod>,
     hidden_size: usize,
-    pub device: Device,
-    pub cache: Cache,
-    pub max_seq_len: usize,
+    device: Device,
+    cache: Cache,
+    max_seq_len: usize,
     mapper: Box<dyn DeviceMapper + Send + Sync>,
     cfg: ModelConfigMetadata,
 }

mistralrs-core/src/models/gemma2.rs

@@ -470,9 +470,9 @@ pub struct Model {
     norm: RmsNorm,
     lm_head: Arc<dyn QuantMethod>,
     hidden_size: usize,
-    pub device: Device,
-    pub cache: Cache,
-    pub max_seq_len: usize,
+    device: Device,
+    cache: Cache,
+    max_seq_len: usize,
     mapper: Box<dyn DeviceMapper + Send + Sync>,
     sliding_window: usize,
     final_logit_softcapping: Option<f64>,

mistralrs-core/src/models/llama.rs

@@ -388,8 +388,8 @@ pub struct Llama {
     blocks: Vec<Block>,
     ln_f: RmsNorm,
     lm_head: Arc<dyn QuantMethod>,
-    pub kv_cache: crate::pipeline::Cache,
-    pub device: Device,
+    kv_cache: crate::pipeline::Cache,
+    device: Device,
     mapper: Box<dyn DeviceMapper + Send + Sync>,
     cfg: ModelConfigMetadata,
 }

mistralrs-core/src/models/mistral.rs

@@ -411,9 +411,9 @@ pub struct Model {
     norm: RmsNorm,
     lm_head: Arc<dyn QuantMethod>,
     sliding_window: Option<usize>,
-    pub device: Device,
-    pub cache: Cache,
-    pub max_seq_len: usize,
+    device: Device,
+    pub(crate) cache: Cache,
+    max_seq_len: usize,
     mapper: Box<dyn DeviceMapper + Send + Sync>,
     cfg: ModelConfigMetadata,
 }

mistralrs-core/src/models/mixtral.rs

@@ -460,9 +460,9 @@ pub struct Model {
     norm: RmsNorm,
     lm_head: Arc<dyn QuantMethod>,
     sliding_window: Option<usize>,
-    pub device: Device,
-    pub cache: Cache,
-    pub max_seq_len: usize,
+    device: Device,
+    cache: Cache,
+    max_seq_len: usize,
     mapper: Box<dyn DeviceMapper + Send + Sync>,
     cfg: ModelConfigMetadata,
 }

mistralrs-core/src/models/phi2.rs

@@ -405,9 +405,9 @@ pub struct Model {
     layers: Vec<DecoderLayer>,
     final_layernorm: LayerNorm,
     lm_head: Arc<dyn QuantMethod>,
-    pub cache: Cache,
-    pub device: Device,
-    pub max_seq_len: usize,
+    cache: Cache,
+    device: Device,
+    max_seq_len: usize,
     mapper: Box<dyn DeviceMapper + Send + Sync>,
     cfg: ModelConfigMetadata,
 }

mistralrs-core/src/models/phi3.rs

@@ -398,9 +398,9 @@ pub struct Model {
     layers: Vec<DecoderLayer>,
     norm: RmsNorm,
     lm_head: Arc<dyn QuantMethod>,
-    pub device: Device,
-    pub cache: Cache,
-    pub max_seq_len: usize,
+    device: Device,
+    cache: Cache,
+    max_seq_len: usize,
     mapper: Box<dyn DeviceMapper + Send + Sync>,
     sliding_window: Option<usize>,
     cfg: ModelConfigMetadata,

mistralrs-core/src/models/qwen2.rs

@@ -393,9 +393,9 @@ pub struct Model {
     norm: RmsNorm,
     lm_head: Arc<dyn QuantMethod>,
     sliding_window: usize,
-    pub device: Device,
-    pub cache: Cache,
-    pub max_seq_len: usize,
+    device: Device,
+    cache: Cache,
+    max_seq_len: usize,
     mapper: Box<dyn DeviceMapper + Send + Sync>,
     cfg: ModelConfigMetadata,
 }

mistralrs-core/src/models/starcoder2.rs

@@ -392,9 +392,9 @@ pub struct Model {
     norm: LayerNorm,
     lm_head: Arc<dyn QuantMethod>,
     sliding_window: Option<usize>,
-    pub device: Device,
-    pub cache: Cache,
-    pub max_seq_len: usize,
+    device: Device,
+    cache: Cache,
+    max_seq_len: usize,
     mapper: Box<dyn DeviceMapper + Send + Sync>,
     cfg: ModelConfigMetadata,
 }

mistralrs-core/src/utils/model_config.rs

@@ -21,7 +21,7 @@ pub struct FileGGML {
 
 #[derive(derive_more::From)]
 pub struct Device<'a> {
-    pub device: &'a candle_core::Device,
+    device: &'a candle_core::Device,
     pub mapper: DeviceMapMetadata,
 }