We read every piece of feedback, and take your input very seriously.
To see all available qualifiers, see our documentation.
There was an error while loading. Please reload this page.
1 parent 1b290ac commit eb825c1Copy full SHA for eb825c1
vllm/model_executor/models/gpt_j.py
@@ -250,7 +250,7 @@ def load_weights(self,
250
if att_weight_name not in name:
251
continue
252
param = state_dict[name.replace(att_weight_name, "qkv_proj")]
253
- shard_size = param.shape[1]
+ shard_size = param.shape[0] // 3
254
loaded_weight = loaded_weight[shard_size * tp_rank:shard_size *
255
(tp_rank + 1)]
256
param_slice = param.data[shard_size * stride_id:shard_size *
0 commit comments