v2: remove remaining logit scaling hacks from batch/optimized paths

2026-07-05 22:41:49 +08:00
parent 239474bef0
commit 7a8edf77ee
3 changed files with 0 additions and 32 deletions
@@ -110,12 +110,6 @@ extension E4BModel {
        try quantizedMatmulOptimized(input: lmInput, weights: embedWeight, 
                                     output: logitsBuffer, cmdBuf: cmdBuf3)
        
-        // Logits scaling (if needed)
-        if embedWeight.groupSize == 32 && embedWeight.inDim == hiddenSize {
-            let logitsScale = Float(30.0 / 116.23 / sqrt(Float(hiddenSize)))
-            try scaleBufferOptimized(logitsBuffer, scale: logitsScale, count: vocabSize, cmdBuf: cmdBuf3)
-        }
-        
        // Logit softcapping
        if let cap = finalLogitSoftcapping {
            try applyLogitSoftcappingOptimized(buffer: logitsBuffer, cap: cap,