Commit 18b4e6f3 authored by zhuwenwen's avatar zhuwenwen
Browse files

set USE_FUSED_RMS_QUANT=1, USE_FUSED_SILU_MUL_QUANT=1 and ...

set USE_FUSED_RMS_QUANT=1, USE_FUSED_SILU_MUL_QUANT=1 and  VLLM_SCHED_ENABLE_MINIMAL_INJECTION=1 for dpsk-v3
parent 62420b27
...@@ -259,7 +259,12 @@ def get_model_architecture( ...@@ -259,7 +259,12 @@ def get_model_architecture(
os.environ['VLLM_REJECT_SAMPLE_OPT'] = '1' os.environ['VLLM_REJECT_SAMPLE_OPT'] = '1'
# if not envs.is_set("VLLM_USE_LIGHTOP_RMS_ROPE_CONCAT"): # if not envs.is_set("VLLM_USE_LIGHTOP_RMS_ROPE_CONCAT"):
# os.environ['VLLM_USE_LIGHTOP_RMS_ROPE_CONCAT'] = '1' # os.environ['VLLM_USE_LIGHTOP_RMS_ROPE_CONCAT'] = '1'
if not envs.is_set("USE_FUSED_RMS_QUANT"):
os.environ['USE_FUSED_RMS_QUANT'] = '1'
if not envs.is_set("USE_FUSED_SILU_MUL_QUANT"):
os.environ['USE_FUSED_SILU_MUL_QUANT'] = '1'
if not envs.is_set("VLLM_SCHED_ENABLE_MINIMAL_INJECTION"):
os.environ['VLLM_SCHED_ENABLE_MINIMAL_INJECTION'] = '1'
else: else:
if not envs.is_set("VLLM_USE_PD_SPLIT"): if not envs.is_set("VLLM_USE_PD_SPLIT"):
os.environ['VLLM_USE_PD_SPLIT'] = '1' os.environ['VLLM_USE_PD_SPLIT'] = '1'
...@@ -298,6 +303,12 @@ def get_model_architecture( ...@@ -298,6 +303,12 @@ def get_model_architecture(
os.environ['VLLM_REJECT_SAMPLE_OPT'] = '1' os.environ['VLLM_REJECT_SAMPLE_OPT'] = '1'
# if not envs.is_set("VLLM_USE_LIGHTOP_RMS_ROPE_CONCAT"): # if not envs.is_set("VLLM_USE_LIGHTOP_RMS_ROPE_CONCAT"):
# os.environ['VLLM_USE_LIGHTOP_RMS_ROPE_CONCAT'] = '1' # os.environ['VLLM_USE_LIGHTOP_RMS_ROPE_CONCAT'] = '1'
if not envs.is_set("USE_FUSED_RMS_QUANT"):
os.environ['USE_FUSED_RMS_QUANT'] = '1'
if not envs.is_set("USE_FUSED_SILU_MUL_QUANT"):
os.environ['USE_FUSED_SILU_MUL_QUANT'] = '1'
if not envs.is_set("VLLM_SCHED_ENABLE_MINIMAL_INJECTION"):
os.environ['VLLM_SCHED_ENABLE_MINIMAL_INJECTION'] = '1'
else: else:
if not envs.is_set("VLLM_USE_PD_SPLIT"): if not envs.is_set("VLLM_USE_PD_SPLIT"):
os.environ['VLLM_USE_PD_SPLIT'] = '1' os.environ['VLLM_USE_PD_SPLIT'] = '1'
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment