apps/FastDeploy
1
0
Fork 0
mirror of https://github.com/PaddlePaddle/FastDeploy.git synced 2025-10-23 16:44:22 +08:00
Code Issues Actions 5 Packages Projects Releases Wiki Activity

[BugFix]Fix load kv cache quant scale (#4077)

Browse Source 
* fix kv cache

* fix kv_cache

* fix kv cache
This commit is contained in:
YuanRisheng
2025-09-12 17:44:03 +08:00
committed by GitHub
parent c86b3357ce
commit 88ea565aba
2 changed files with 10 additions and 20 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
tests/quantization/test_kv_cache.py
View File

@@ -37,6 +37,7 @@ class MockLayer(nn.Layer):
self.fd_config = get_default_test_fd_config()
self.fd_config.model_config.num_key_value_heads = 1
self.head_dim = 1
self.kv_num_heads = 1
self.prefix = "mock_layer"
self.cache_k_scale = None
self.cache_v_scale = None
@@ -77,7 +78,7 @@ class TestKVCacheMethodBase(unittest.TestCase):
method = KVCacheMethodBase(config)
method.create_weights(self.layer)
self.assertEqual(self.layer.cache_k_scale.shape, [1, 1])
self.assertEqual(self.layer.cache_k_scale.shape, [1])
def test_create_weights_int4_zp(self):
# Test INT4 with zero point