From 99b40247ea7579bca2b318d7294f79d7ce69a242 Mon Sep 17 00:00:00 2001 From: Yuanle Liu Date: Mon, 15 Dec 2025 20:08:07 +0800 Subject: [PATCH] [Cherry-Pick][BugFix] fix dynamic c8 in v1 loader(#5562) (#5519) * fix dyname load bug * update * update --- .../model_executor/layers/quantization/kv_cache.py | 9 +++++---- 1 file changed, 5 insertions(+), 4 deletions(-) diff --git a/fastdeploy/model_executor/layers/quantization/kv_cache.py b/fastdeploy/model_executor/layers/quantization/kv_cache.py index cd461fde7..2ccc06f0e 100644 --- a/fastdeploy/model_executor/layers/quantization/kv_cache.py +++ b/fastdeploy/model_executor/layers/quantization/kv_cache.py @@ -263,10 +263,11 @@ class KVCacheMethodBase(QuantMethodBase): """ use for loader v1 """ - if layer.cache_k_scale._is_initialized(): - layer.cache_k_out_scale.set_value(1 / layer.cache_k_scale) - if layer.cache_v_scale._is_initialized(): - layer.cache_v_out_scale.set_value(1 / layer.cache_v_scale) + if "block_wise" not in layer.cache_quant_type_str: + if layer.cache_k_scale._is_initialized(): + layer.cache_k_out_scale.set_value(1 / layer.cache_k_scale) + if layer.cache_v_scale._is_initialized(): + layer.cache_v_out_scale.set_value(1 / layer.cache_v_scale) def apply(self, layer): """