Skip to content
This repository was archived by the owner on Oct 25, 2024. It is now read-only.

Commit 33defb7

Browse files
authored
[LLM Runtime] Remove the identical branch (#894)
1 parent 0b18d06 commit 33defb7

File tree

3 files changed

+2
-7
lines changed

3 files changed

+2
-7
lines changed

examples/huggingface/pytorch/text-classification/deployment/emotion/distilbert_base_uncased/executor_utils.py

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -37,7 +37,7 @@ def accuracy(self, batch_size, seq_len, dataset_name, task_name, data_dir, token
3737
# load metric
3838
log.info("Load metric ......")
3939
if dataset_name and task_name is not None:
40-
metric = load_metric("accuracy")
40+
metric = load_metric(dataset_name, task_name)
4141
else:
4242
metric = load_metric("accuracy")
4343
# execute

intel_extension_for_transformers/llm/runtime/deprecated/compile/sub_graph/qkv_merge.py

Lines changed: 0 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -102,7 +102,6 @@ def get_zero_ratio(matrix, block):
102102
if ((v_matmul.attr.__contains__("output_dtype")
103103
and q_matmul.attr.__contains__("output_dtype")
104104
and k_matmul.attr.__contains__("output_dtype")
105-
and q_matmul.attr["output_dtype"] == v_matmul.attr["output_dtype"]
106105
and q_matmul.attr["output_dtype"] == v_matmul.attr["output_dtype"])
107106
or (v_matmul.attr.__contains__("output_dtype") == False
108107
and q_matmul.attr.__contains__("output_dtype") == False

intel_extension_for_transformers/llm/runtime/graph/scripts/convert_gptq_bloom.py

Lines changed: 1 addition & 5 deletions
Original file line numberDiff line numberDiff line change
@@ -195,11 +195,7 @@ def bytes_to_unicode():
195195

196196
ftype_cur = 0
197197
if ".weight" in name and list_vars[name].dim() == 2:
198-
if name.replace(".weight",
199-
"") in weight_config and weight_config[name.replace(".weight", "")]["dtype"] != "fp32":
200-
ftype_cur = 2 # 13
201-
else:
202-
ftype_cur = 2 # 2
198+
ftype_cur = 2 # TODO(Zhenwei) support jblas
203199

204200
data = list_vars[src].squeeze().numpy()
205201
data = data.astype(np.float32)

0 commit comments

Comments
 (0)