Add WaterSIC ZSIC core algorithm with unit tests

Signed-off-by: Kai Xu <kaix@nvidia.com>

Author: kaix-nv

modelopt/torch/quantization/algorithms.py …orch/quantization/algorithms/__init__.pymodelopt/torch/quantization/algorithms.py renamed to modelopt/torch/quantization/algorithms/__init__.py modelopt/torch/quantization/algorithms.py renamed to modelopt/torch/quantization/algorithms/__init__.py

@@ -38,12 +38,12 @@
 from modelopt.torch.utils import create_param_grad_clear_hook, print_rank_0, report_memory
 from modelopt.torch.utils.distributed import DistributedProcessGroup, ParallelState, is_master
 
-from . import config as mtq_config
-from . import model_calib
-from .config import QuantizeConfig, QuantizerAttributeConfig
-from .conversion import set_quantizer_by_cfg
-from .nn import QuantLinearConvBase, QuantModule, SequentialQuantizer, TensorQuantizer
-from .utils import is_quantized_linear
+from .. import config as mtq_config
+from .. import model_calib
+from ..config import QuantizeConfig, QuantizerAttributeConfig
+from ..conversion import set_quantizer_by_cfg
+from ..nn import QuantLinearConvBase, QuantModule, SequentialQuantizer, TensorQuantizer
+from ..utils import is_quantized_linear
 
 
 def estimate_quant_compression(quant_cfg: QuantizeConfig) -> float:
@@ -615,8 +615,8 @@ def before_search(self):
         # Import here to avoid circular import
         from modelopt.torch.quantization.model_quant import calibrate
 
-        from .conversion import restore_quantizer_state, update_quantize_metadata
-        from .utils import get_quantizer_state_dict, set_quantizer_state_dict
+        from ..conversion import restore_quantizer_state, update_quantize_metadata
+        from ..utils import get_quantizer_state_dict, set_quantizer_state_dict
 
         super().before_search()
         restored_method = getattr(self, "method", None)

modelopt/torch/quantization/algorithms/watersic_kv/__init__.py

@@ -0,0 +1,18 @@
+# SPDX-FileCopyrightText: Copyright (c) 2024 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+# SPDX-License-Identifier: Apache-2.0
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+# http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+"""WaterSIC KV-cache quantization algorithm."""
+
+from __future__ import annotations