Skip to content

[HS-6944] Fix for deepseek distill models #359

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

Open
nazneenn wants to merge 2 commits into
base: v1.22.0
Choose a base branch
from
Open

[HS-6944] Fix for deepseek distill models #359

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
f04f5c2
deepseek-distill changes
Sep 10, 2025
92e494f
deepseek-distill changes
Sep 10, 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
6 changes: 3 additions & 3 deletions calibration/calibrate_model.sh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -43,7 +43,7 @@ create_measure_config() {

if [[ $model_name_lower =~ ^mixtral ]]; then
tmp_config="{\"method\": \"HOOKS\",\"mode\": \"MEASURE\",\"observer\": \"maxabs\",\"allowlist\": {\"types\": [], \"names\": []},\"blocklist\": {\"types\": [], \"names\": [\"self_attn\", \"lm_head\"]},\"quantize_weight\": false,\"dump_stats_path\": \"$1/$2/$3/inc_output\"}"
elif [[ $model_name_lower =~ ^deepseek ]]; then
elif [[ $model_name_lower =~ ^deepseek && ! $model_name_lower =~ distill ]]; then
tmp_config="{\"method\": \"HOOKS\",\"mode\": \"MEASURE\",\"observer\": \"maxabs\",\"allowlist\": {\"types\": [], \"names\": []},\"blocklist\": {\"types\": [], \"names\": [\"lm_head\", \"mlp\\\.gate\\\b\"]},\"quantize_weight\": false,\"dump_stats_path\": \"$1/$2/$3/inc_output\"}"
else
tmp_config="{\"method\": \"HOOKS\",\"mode\": \"MEASURE\",\"observer\": \"maxabs\",\"allowlist\": {\"types\": [], \"names\": []},\"blocklist\": {\"types\": [], \"names\": []},\"quantize_weight\": false,\"dump_stats_path\": \"$1/$2/$3/inc_output\"}"
Expand All @@ -63,7 +63,7 @@ create_quant_config() {
else
tmp_config="{\"mode\": \"QUANTIZE\",\"observer\": \"maxabs\",\"scale_method\": \"maxabs_hw\",\"allowlist\": {\"types\": [],\"names\": []},\"blocklist\": {\"types\": [],\"names\": [\"self_attn\", \"lm_head\"]},\"dump_stats_path\": \"$1/$2/$3/inc_output\"}"
fi
elif [[ $model_name_lower =~ ^deepseek ]]; then
elif [[ $model_name_lower =~ ^deepseek && ! $model_name_lower =~ distill ]]; then
tmp_config="{\"mode\": \"QUANTIZE\",\"observer\": \"maxabs\",\"scale_method\": \"maxabs_hw\", \"scale_format\": \"scalar\", \"allowlist\": {\"types\": [],\"names\": []},\"blocklist\": {\"types\": [],\"names\": [\"lm_head\", \"mlp\\\.gate\\\b\"]},\"dump_stats_path\": \"$1/$2/$3/inc_output\"}"
else
tmp_config="{\"mode\": \"QUANTIZE\",\"observer\": \"maxabs\",\"scale_method\": \"maxabs_hw\",\"allowlist\": {\"types\": [],\"names\": []},\"blocklist\": {\"types\": [],\"names\": []},\"dump_stats_path\": \"$1/$2/$3/inc_output\"}"
Expand Down Expand Up @@ -202,7 +202,7 @@ else
fi


if [[ "$model_name_lower" == *"deepseek"* ]]; then
if [[ "$model_name_lower" == *"deepseek"* && "$model_name_lower" != *distill* ]]; then
EXTRA_FLAGS_STEP_2+="--block-quant --expert-parallel "
EXTRA_ENVS_STEP_2="VLLM_HPU_FORCE_CHANNEL_FP8=0"
EXTRA_FLAGS_STEP_3+="--deepseek "
Expand Down