Skip to content

[0.10.2][Security] Fix/remove logically dead code (#448) #461

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
merged 2 commits into from
Oct 27, 2025
Merged

[0.10.2][Security] Fix/remove logically dead code (#448) #461

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
8df24f7
[Security] Fix/remove logically dead code (#448)
afierka-intel Oct 23, 2025
9982797
Merge branch 'v0.10.2_next' into port/afierka/fix-logic-dead-code
afierka-intel Oct 27, 2025
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
19 changes: 5 additions & 14 deletions vllm_gaudi/ops/hpu_gptq.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -168,7 +168,6 @@ def create_weights(

group_size = self.quant_config.group_size if self.quant_config.group_size != -1 else input_size
scale_and_zero_size = input_size // group_size
scale_and_zero_input_dim = None

qweight = PackedvLLMParameter(data=torch.empty(
input_size_per_partition // self.quant_config.pack_factor,
Expand Down Expand Up @@ -205,20 +204,12 @@ def create_weights(
),
"weight_loader": weight_loader
}
if scale_and_zero_input_dim is None:
scales = ChannelQuantScaleParameter(output_dim=1, **weight_scale_args)
qzeros = PackedColumnParameter(output_dim=1,
packed_dim=1,
packed_factor=self.quant_config.pack_factor,
**qzeros_args)

else:
scales = GroupQuantScaleParameter(output_dim=1, input_dim=0, **weight_scale_args)
qzeros = PackedvLLMParameter(input_dim=0,
output_dim=1,
packed_dim=1,
packed_factor=self.quant_config.pack_factor,
**qzeros_args)
scales = ChannelQuantScaleParameter(output_dim=1, **weight_scale_args)
qzeros = PackedColumnParameter(output_dim=1,
packed_dim=1,
packed_factor=self.quant_config.pack_factor,
**qzeros_args)

qzeros.pack_factor = self.quant_config.pack_factor

Expand Down
16 changes: 6 additions & 10 deletions vllm_gaudi/v1/worker/hpu_model_runner.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -4310,20 +4310,16 @@ def copy_kv_blocks(

i = 0
global hpu_buffer
use_hpu_buffer = False
for layer_name in src_kv_caches:
key_cache = src_kv_caches[layer_name][0]
value_cache = src_kv_caches[layer_name][1]

if direction == "d2h" and use_hpu_buffer:
hpu_buffer[i][0] = key_cache.index_select(0, src_slot_mapping)
hpu_buffer[i][1] = value_cache.index_select(0, src_slot_mapping)
else:
#import remote_pdb;remote_pdb.set_trace()
dst_kv_caches[layer_name][0].index_put_((dst_slot_mapping, ),
key_cache.index_select(0, src_slot_mapping).to(target_device))
dst_kv_caches[layer_name][1].index_put_((dst_slot_mapping, ),
value_cache.index_select(0, src_slot_mapping).to(target_device))
dst_kv_caches[layer_name][0].index_put_((dst_slot_mapping, ),
key_cache.index_select(0, src_slot_mapping).to(target_device))
dst_kv_caches[layer_name][1].index_put_((dst_slot_mapping, ),
value_cache.index_select(0, src_slot_mapping).to(target_device))
if direction == "d2h":
dst_kv_caches[layer_name] = dst_kv_caches[layer_name].unflatten(1, (-1, block_size))

i = i + 1

Expand Down
Loading