Pixtral polishing. (#2522)

* Pixtral polishing.

* Clippy fix.

candle-examples/examples/pixtral/main.rs

@@ -73,22 +73,18 @@ impl TextGeneration {
         let img_break = get_token("[IMG_BREAK]")?;
         let img_end = get_token("[IMG_END]")?;
         let start_gen = std::time::Instant::now();
-        let mut pos = 0;
         for index in 0..sample_len {
             let logits = if index > 0 {
                 let context_size = if index > 0 { 1 } else { tokens.len() };
                 let start_pos = tokens.len().saturating_sub(context_size);
                 let ctxt = &tokens[start_pos..];
                 let input = Tensor::new(ctxt, &self.device)?.unsqueeze(0)?;
-                let logits = self.model.language_model.forward(&input, pos)?;
+                self.model.lm_forward(&input)?
-                pos += context_size;
-                logits
             } else {
                 let (_b, _c, h, w) = self.image.dims4()?;
                 let h = h / self.model.patch_size;
                 let w = w / self.model.patch_size;
-                let image_embeds = self.model.vision_tower.forward(&self.image)?;
+                let image_embeds = self.model.encode_image(&self.image)?;
-                let image_embeds = self.model.multi_modal_projector.forward(&image_embeds)?;
                 println!("generated image embeddings {image_embeds:?}");
                 let image_embeds = image_embeds.to_dtype(self.model.dtype)?;
                 for &t in tokens.iter() {
@@ -124,12 +120,7 @@ impl TextGeneration {
                 input_embeds.push(end_embeds);
 
                 let input_embeds = Tensor::cat(&input_embeds, 1)?;
-                let logits = self
+                self.model.lm_forward_embeds(&input_embeds)?
-                    .model
-                    .language_model
-                    .forward_embeds(&input_embeds, None, pos)?;
-                pos += input_embeds.dim(1)?;
-                logits
             };
             let logits = logits.squeeze(0)?.squeeze(0)?.to_dtype(DType::F32)?;
             let logits = if self.repeat_penalty == 1. {

candle-transformers/src/models/pixtral/llava.rs

@@ -48,6 +48,7 @@ pub struct Model {
     pub vision_tower: vision_model::Model,
     pub patch_size: usize,
     pub dtype: candle::DType,
+    pub pos: usize,
 }
 
 impl Model {
@@ -67,6 +68,31 @@ impl Model {
             vision_tower,
             patch_size: cfg.vision_config.patch_size,
             dtype: vb.dtype(),
+            pos: 0,
         })
     }
+
+    pub fn clear_kv_cache(&mut self) {
+        self.language_model.clear_kv_cache();
+        self.pos = 0;
+    }
+
+    pub fn encode_image(&self, image: &Tensor) -> Result<Tensor> {
+        let image_embeds = self.vision_tower.forward(image)?;
+        self.multi_modal_projector.forward(&image_embeds)
+    }
+
+    pub fn lm_forward(&mut self, input_ids: &Tensor) -> Result<Tensor> {
+        let (_, seq_len) = input_ids.dims2()?;
+        let logits = self.language_model.forward(input_ids, self.pos)?;
+        self.pos += seq_len;
+        Ok(logits)
+    }
+
+    pub fn lm_forward_embeds(&mut self, xs: &Tensor) -> Result<Tensor> {
+        let (_, seq_len, _) = xs.dims3()?;
+        let logits = self.language_model.forward_embeds(xs, None, self.pos)?;
+        self.pos += seq_len;
+        Ok(logits)
+    }
 }