Skip to content

[LLM Inference] Qwen2_Moe Support wint4 #9129

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
qingqing01 merged 2 commits into from
Sep 14, 2024
Merged

[LLM Inference] Qwen2_Moe Support wint4 #9129

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
21a2a31
support wint4
CJ77Qi Sep 12, 2024
d778c5d
fix codestyle
CJ77Qi Sep 12, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
11 changes: 9 additions & 2 deletions paddlenlp/experimental/transformers/fused_transformer_layers.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1299,8 +1299,12 @@ def init_weight_shape(self, config):
self.moe_ffn2_weight_shape = [self.config.moe_config.num_experts, self.dim_feedforward, self.embed_dim]

if config.quant_type == "weight_only_int4":
self.moe_ffn1_weight_shape[2] //= 2
self.moe_ffn2_weight_shape[2] //= 2
if config.moe_config.has_shared_expert():
self.moe_ffn1_weight_shape[2] //= 2
self.moe_ffn2_weight_shape[1] //= 2
else:
self.moe_ffn1_weight_shape[2] //= 2
self.moe_ffn2_weight_shape[2] //= 2

if self.config.moe_config.has_shared_expert():
self.shared_expert_ffn1_weight_shape = [
Expand All @@ -1315,6 +1319,9 @@ def init_weight_shape(self, config):
self.embed_dim,
1,
]
if config.quant_type == "weight_only_int4":
self.shared_expert_ffn1_weight_shape[0] //= 2
self.shared_expert_ffn2_weight_shape[0] //= 2

def compute_qkv_linear(self, ln_out, i):
return weight_only_linear(
Expand Down