@@ -323,20 +323,20 @@ impl Qwen2Vl {
323
323
#[ derive( Clone , Debug , Serialize , Deserialize ) ]
324
324
#[ serde( rename_all = "snake_case" ) ]
325
325
pub struct Qwen2_5VlVisionConfig {
326
- pub ( crate ) depth : usize ,
327
- pub ( crate ) hidden_act : String ,
328
- pub ( crate ) hidden_size : usize ,
329
- pub ( crate ) intermediate_size : usize ,
330
- pub ( crate ) num_heads : usize ,
331
- pub ( crate ) in_chans : usize ,
332
- pub ( crate ) out_hidden_size : usize ,
333
- pub ( crate ) patch_size : usize ,
334
- pub ( crate ) spatial_merge_size : usize ,
326
+ // pub(crate) depth: usize,
327
+ // pub(crate) hidden_act: String,
328
+ // pub(crate) hidden_size: usize,
329
+ // pub(crate) intermediate_size: usize,
330
+ // pub(crate) num_heads: usize,
331
+ // pub(crate) in_chans: usize,
332
+ // pub(crate) out_hidden_size: usize,
333
+ // pub(crate) patch_size: usize,
334
+ // pub(crate) spatial_merge_size: usize,
335
335
pub ( crate ) spatial_patch_size : usize ,
336
- pub ( crate ) window_size : usize ,
337
- pub ( crate ) fullatt_block_indexes : Vec < usize > ,
338
- pub ( crate ) tokens_per_second : usize ,
339
- pub ( crate ) temporal_patch_size : usize ,
336
+ // pub(crate) window_size: usize,
337
+ // pub(crate) fullatt_block_indexes: Vec<usize>,
338
+ // pub(crate) tokens_per_second: usize,
339
+ // pub(crate) temporal_patch_size: usize,
340
340
}
341
341
342
342
#[ derive( Clone , Debug , Serialize , Deserialize ) ]
@@ -348,7 +348,7 @@ pub struct Qwen2_5Vl {
348
348
impl Qwen2_5Vl {
349
349
pub fn get_number_of_features ( & self , height : usize , width : usize ) -> usize {
350
350
let num_pixels = height * width;
351
- num_pixels / self . vision_config . patch_size . pow ( 2 )
351
+ num_pixels / self . vision_config . spatial_patch_size . pow ( 2 )
352
352
}
353
353
}
354
354
0 commit comments