Skip to content

refactor: rename quant policy to kv cache dtype #4718

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

Draft
CUHKSZzxy wants to merge 2 commits into
base: main
Choose a base branch
from
Draft

refactor: rename quant policy to kv cache dtype #4718

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
5dfe256
refactor: rename quant policy to kv cache dtype
CUHKSZzxy Jun 29, 2026
2d0baf3
Merge remote-tracking branch 'upstream/main' into refactor/kv-cache-d…
CUHKSZzxy Jul 8, 2026
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
4 changes: 2 additions & 2 deletions .github/scripts/eval_chat_config.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -316,11 +316,11 @@
model['path'] = model['path'] + '-inner-w8a8'

for model in [v for k, v in locals().items() if k.endswith('_kvint4')]:
model['engine_config']['quant_policy'] = 4
model['engine_config']['kv_cache_dtype'] = 4
model['abbr'] = model['abbr'] + '_kvint4'

for model in [v for k, v in locals().items() if k.endswith('_kvint8')]:
model['engine_config']['quant_policy'] = 8
model['engine_config']['kv_cache_dtype'] = 8
model['abbr'] = model['abbr'] + '_kvint8'

for model in [v for k, v in locals().items() if k.startswith('pytorch_')]:
Expand Down
8 changes: 4 additions & 4 deletions autotest/benchmark/test_apiserver_performance.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -113,7 +113,7 @@ def test_pytorch_apiserver_tp16(config, run_config, worker_id):
'model': 'Qwen/Qwen3-30B-A3B',
'backend': 'pytorch',
'communicator': 'nccl',
'quant_policy': 0,
'kv_cache_dtype': 0,
'parallel_config': {
'tp': 2
},
Expand All @@ -122,7 +122,7 @@ def test_pytorch_apiserver_tp16(config, run_config, worker_id):
'model': 'Qwen/Qwen3-30B-A3B',
'backend': 'turbomind',
'communicator': 'nccl',
'quant_policy': 4,
'kv_cache_dtype': 4,
'parallel_config': {
'tp': 2
},
Expand All @@ -131,7 +131,7 @@ def test_pytorch_apiserver_tp16(config, run_config, worker_id):
'model': 'Qwen/Qwen3-30B-A3B',
'backend': 'turbomind',
'communicator': 'cuda-ipc',
'quant_policy': 8,
'kv_cache_dtype': 8,
'parallel_config': {
'tp': 2
},
Expand All @@ -140,7 +140,7 @@ def test_pytorch_apiserver_tp16(config, run_config, worker_id):
'model': 'Qwen/Qwen3-VL-30B-A3B-Instruct',
'backend': 'pytorch',
'communicator': 'nccl',
'quant_policy': 8,
'kv_cache_dtype': 8,
'parallel_config': {
'tp': 2
},
Expand Down
8 changes: 4 additions & 4 deletions autotest/benchmark/test_prefixcache_performance.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -96,7 +96,7 @@ def test_pytorch_prefix_tp16(config, run_config, worker_id):
'model': 'Qwen/Qwen3-30B-A3B',
'backend': 'turbomind',
'communicator': 'cuda-ipc',
'quant_policy': 0,
'kv_cache_dtype': 0,
'parallel_config': {
'tp': 2
},
Expand All @@ -105,7 +105,7 @@ def test_pytorch_prefix_tp16(config, run_config, worker_id):
'model': 'Qwen/Qwen3-VL-30B-A3B-Instruct',
'backend': 'pytorch',
'communicator': 'nccl',
'quant_policy': 8,
'kv_cache_dtype': 8,
'parallel_config': {
'tp': 2
},
Expand All @@ -114,7 +114,7 @@ def test_pytorch_prefix_tp16(config, run_config, worker_id):
'model': 'Qwen/Qwen3-30B-A3B',
'backend': 'turbomind',
'communicator': 'cuda-ipc',
'quant_policy': 0,
'kv_cache_dtype': 0,
'parallel_config': {
'tp': 2
},
Expand All @@ -123,7 +123,7 @@ def test_pytorch_prefix_tp16(config, run_config, worker_id):
'model': 'Qwen/Qwen3-VL-30B-A3B-Instruct',
'backend': 'pytorch',
'communicator': 'nccl',
'quant_policy': 8,
'kv_cache_dtype': 8,
'parallel_config': {
'tp': 2
},
Expand Down
8 changes: 4 additions & 4 deletions autotest/benchmark/test_throughput_performance.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -96,7 +96,7 @@ def test_pytorch_throughput_tp16(config, run_config, worker_id):
'model': 'Qwen/Qwen3-30B-A3B',
'backend': 'turbomind',
'communicator': 'cuda-ipc',
'quant_policy': 0,
'kv_cache_dtype': 0,
'parallel_config': {
'tp': 2
},
Expand All @@ -105,7 +105,7 @@ def test_pytorch_throughput_tp16(config, run_config, worker_id):
'model': 'Qwen/Qwen3-VL-30B-A3B-Instruct',
'backend': 'pytorch',
'communicator': 'nccl',
'quant_policy': 8,
'kv_cache_dtype': 8,
'parallel_config': {
'tp': 2
},
Expand All @@ -123,7 +123,7 @@ def test_throughput_func_tp2(config, run_config, worker_id):
'model': 'meta-llama/Meta-Llama-3-1-8B-Instruct',
'backend': 'turbomind',
'communicator': 'nccl',
'quant_policy': 0,
'kv_cache_dtype': 0,
'parallel_config': {
'tp': 1
},
Expand All @@ -132,7 +132,7 @@ def test_throughput_func_tp2(config, run_config, worker_id):
'model': 'Qwen/Qwen3-VL-8B-Instruct',
'backend': 'pytorch',
'communicator': 'nccl',
'quant_policy': 8,
'kv_cache_dtype': 8,
'parallel_config': {
'tp': 1
},
Expand Down
2 changes: 1 addition & 1 deletion autotest/interface/pipeline/test_pipeline_func.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -618,7 +618,7 @@ def test_backend_config_validate_turbomind(config, model, backend, worker_id):
pipeline(model_path, backend_config=backend_config)

with pytest.raises(pydantic.ValidationError):
backend_config = backend(quant_policy=1)
backend_config = backend(kv_cache_dtype=1)
pipeline(model_path, backend_config=backend_config)

with pytest.raises(pydantic.ValidationError):
Expand Down
Loading
Loading