ayo/ipex-llm
1
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions 1

update q4k convert (#11179)

Browse source
This commit is contained in:
Ruonan Wang 2024-05-31 11:36:53 +08:00 committed by GitHub
parent f0aaa130a9
commit 50b5f4476f
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
1 changed files with 2 additions and 2 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

4
python/llm/src/ipex_llm/transformers/low_bit_linear.py
View file

@ -244,7 +244,7 @@ def ggml_q_format_convet_cpu2xpu(tensor: torch.Tensor, num_elem: int, qtype: int
src = ctypes.c_void_p(tensor.data.data_ptr())
if qtype in [SYM_INT4, ASYM_INT4, SYM_INT8, NF4, NF3, FP4, FP6, FP8E4, FP8E5]:
if qtype in [SYM_INT4, ASYM_INT4, SYM_INT8, NF4, NF3, FP4, FP6, FP8E4, FP8E5, Q4_K]:
dst_tensor = torch.empty_like(tensor)
elif qtype == ggml_tensor_qtype["sym_int5"]:
QK = ggml.ggml_qk_size(qtype)
@ -269,7 +269,7 @@ def ggml_q_format_convet_xpu2cpu(tensor: torch.Tensor, num_elem: int, qtype: int
src = ctypes.c_void_p(tensor.data.data_ptr())
if qtype in [SYM_INT4, ASYM_INT4, SYM_INT8, NF4, NF3, FP4, FP6, FP8E4, FP8E5]:
if qtype in [SYM_INT4, ASYM_INT4, SYM_INT8, NF4, NF3, FP4, FP6, FP8E4, FP8E5, Q4_K]:
dst_tensor = torch.empty_like(tensor)
elif qtype == ggml_tensor_qtype["sym_int5"]:
QK = ggml.ggml_qk_size(ggml_tensor_qtype["asym_int5"])