Support SqueezeLLM (#1326)

Co-authored-by: squeeze-ai-lab <squeezeailab.bair@gmail.com>
Co-authored-by: Woosuk Kwon <woosuk.kwon@berkeley.edu>

benchmarks/benchmark_latency.py

@@ -70,7 +70,7 @@ if __name__ == '__main__':
     parser.add_argument('--tokenizer', type=str, default=None)
     parser.add_argument('--quantization',
                         '-q',
-                        choices=['awq', None],
+                        choices=['awq', 'squeezellm', None],
                         default=None)
     parser.add_argument('--tensor-parallel-size', '-tp', type=int, default=1)
     parser.add_argument('--input-len', type=int, default=32)

benchmarks/benchmark_throughput.py

@@ -201,7 +201,7 @@ if __name__ == "__main__":
     parser.add_argument("--tokenizer", type=str, default=None)
     parser.add_argument('--quantization',
                         '-q',
-                        choices=['awq', None],
+                        choices=['awq', 'squeezellm', None],
                         default=None)
     parser.add_argument("--tensor-parallel-size", "-tp", type=int, default=1)
     parser.add_argument("--n",

csrc/quantization.cpp

@@ -7,9 +7,13 @@ torch::Tensor awq_gemm(
   torch::Tensor _zeros,
   int split_k_iters);
 
+void squeezellm_gemm(
+  torch::Tensor vec,
+  torch::Tensor mat,
+  torch::Tensor mul,
+  torch::Tensor lookup_table);
+
 PYBIND11_MODULE(TORCH_EXTENSION_NAME, m) {
-  m.def(
+  m.def("awq_gemm", &awq_gemm, "Quantized GEMM for AWQ");
-    "awq_gemm",
+  m.def("squeezellm_gemm", &squeezellm_gemm, "Quantized GEMM for SqueezeLLM");
-    &awq_gemm,
-    "Quantized GEMM for AWQ");
 }

csrc/quantization/squeezellm/quant_cuda_kernel.cu

@@ -0,0 +1,148 @@
+#include <torch/all.h>
+#include <torch/python.h>
+#include <cuda.h>
+#include <cuda_runtime.h>
+#include <cuda_fp16.h>
+
+// half-tensor
+#include <c10/cuda/CUDAStream.h>
+#include <ATen/cuda/CUDATensorMethods.cuh>
+
+#define BLOCKWIDTH 128
+#define BLOCKHEIGHT4 16
+
+namespace vllm {
+namespace squeezellm {
+
+__device__ inline unsigned int as_unsigned(int i) {
+  return *reinterpret_cast<unsigned int*>(&i);
+}
+
+// 4-bit matvec kernel (LUT-based)
+__global__ void NUQ4MatMulKernel(
+    const  half2* __restrict__ vec,
+    const    int* __restrict__ mat,
+           half2* __restrict__ mul,
+    const  __half* __restrict__ lookup_table,
+    int height,
+    int width,
+    int batch,
+    int vec_height
+) {
+
+  const int blockwidth2 = BLOCKWIDTH / 2;
+
+  int row = BLOCKHEIGHT4 * blockIdx.x;
+  int col =  BLOCKWIDTH * blockIdx.y + threadIdx.x;
+
+  __shared__ half2 blockvec[blockwidth2];
+
+  __shared__ __half deq2[16][BLOCKWIDTH];
+  int off = threadIdx.x;
+  int column_offset = col * 16;
+  for (int val = 0; val < 16; val += 1) {
+    int lut_index = column_offset + val;
+    deq2[val][off] = lookup_table[lut_index];
+  }
+
+  __half res;
+  half2 res2;
+  half2 tmp2;
+
+  int i;
+  int k;
+
+  unsigned int tmp1;
+  unsigned int lut_index1, lut_index2;
+
+  for (int b = 0; b < batch; ++b){
+    i = width * row + col;
+    res = __int2half_rd(0);
+    k = 0;
+
+    __syncthreads();
+    if (threadIdx.x < blockwidth2)
+      blockvec[threadIdx.x] = vec[b * vec_height / 2 + (row / BLOCKHEIGHT4) * blockwidth2 + threadIdx.x];
+    __syncthreads();
+
+    while (k < blockwidth2) {
+      tmp1 = as_unsigned(mat[i]);
+
+      res2 = {};
+      tmp2 = {};
+
+      lut_index1 = tmp1 & 0xF;
+      lut_index2 = (tmp1 >> 4) & 0xF;
+      tmp2.x = deq2[lut_index1][off];
+      tmp2.y = deq2[lut_index2][off];
+      res2 = __hfma2(tmp2, blockvec[k + 0], res2);
+
+      lut_index1 = (tmp1 >> 8) & 0xF;
+      lut_index2 = (tmp1 >> 12) & 0xF;
+      tmp2.x = deq2[lut_index1][off];
+      tmp2.y = deq2[lut_index2][off];
+      res2 = __hfma2(tmp2, blockvec[k + 1], res2);
+
+      lut_index1 = (tmp1 >> 16) & 0xF;
+      lut_index2 = (tmp1 >> 20) & 0xF;
+      tmp2.x = deq2[lut_index1][off];
+      tmp2.y = deq2[lut_index2][off];
+      res2 = __hfma2(tmp2, blockvec[k + 2], res2);
+
+      lut_index1 = (tmp1 >> 24) & 0xF;
+      lut_index2 = (tmp1 >> 28) & 0xF;
+      tmp2.x = deq2[lut_index1][off];
+      tmp2.y = deq2[lut_index2][off];
+      res2 = __hfma2(tmp2, blockvec[k + 3], res2);
+
+      res = __hadd(__hadd(res2.x, res2.y), res);
+
+      i += width;
+      k += 4;
+    }
+
+    // col%2 -> only set one of the two values
+    half2 res3 = {};
+    if (col % 2 == 0) {
+      res3.x = res;
+    } else {
+      res3.y = res;
+    }
+
+    atomicAdd(&mul[b * width / 2 + col / 2], res3);
+  }
+}
+
+} // namespace squeezellm
+} // namespace vllm
+
+// 4-bit matvec kernel (LUT-based)
+void squeezellm_gemm(
+  torch::Tensor vec,
+  torch::Tensor mat,
+  torch::Tensor mul,
+  torch::Tensor lookup_table
+) {
+  int height = mat.size(0);
+  int width = mat.size(1);
+
+  int batch = vec.size(0);
+  int vec_height = vec.size(1);
+
+  dim3 blocks(
+    (height + BLOCKHEIGHT4 - 1) / BLOCKHEIGHT4,
+    (width + BLOCKWIDTH - 1) / BLOCKWIDTH
+  );
+  dim3 threads(BLOCKWIDTH);
+
+  vllm::squeezellm::NUQ4MatMulKernel<<<blocks, threads>>>(
+    (half2*) vec.data<at::Half>(),
+    mat.data_ptr<int>(),
+    (half2*) mul.data<at::Half>(),
+    (__half*) lookup_table.data<at::Half>(),
+    height, width, batch, vec_height
+  );
+}
+
+#undef BLOCKWIDTH
+#undef BLOCKHEIGHT4

setup.py

@@ -200,6 +200,7 @@ quantization_extension = CUDAExtension(
     sources=[
         "csrc/quantization.cpp",
         "csrc/quantization/awq/gemm_kernels.cu",
+        "csrc/quantization/squeezellm/quant_cuda_kernel.cu",
     ],
     extra_compile_args={
         "cxx": CXX_FLAGS,

vllm/config.py

@@ -103,7 +103,7 @@ class ModelConfig:
         self.tokenizer_mode = tokenizer_mode
 
     def _verify_quantization(self) -> None:
-        supported_quantization = ["awq"]
+        supported_quantization = ["awq", "squeezellm"]
         if self.quantization is None:
             return
         quantization = self.quantization.lower()

vllm/engine/arg_utils.py

@@ -168,7 +168,7 @@ class EngineArgs:
         parser.add_argument('--quantization',
                             '-q',
                             type=str,
-                            choices=['awq', None],
+                            choices=['awq', 'squeezellm', None],
                             default=None,
                             help='Method used to quantize the weights')
         return parser

vllm/model_executor/layers/quantized_linear/__init__.py

@@ -1,10 +1,14 @@
 from vllm.model_executor.layers.quantized_linear.awq import (
     AWQColumnParallelLinear, AWQRowParallelLinear)
+from vllm.model_executor.layers.quantized_linear.squeezellm import (
+    SqueezeLLMColumnParallelLinear, SqueezeLLMRowParallelLinear)
 from vllm.model_executor.parallel_utils.layers import (ColumnParallelLinear,
                                                        RowParallelLinear)
 
 _QUANTIZED_LINEAR_REGISTRY = {
     "awq": (AWQColumnParallelLinear, AWQRowParallelLinear),
+    "squeezellm":
+    (SqueezeLLMColumnParallelLinear, SqueezeLLMRowParallelLinear),
 }
 
 

vllm/model_executor/layers/quantized_linear/awq.py

@@ -11,9 +11,11 @@ from vllm.model_executor.parallel_utils.layers import (ColumnParallelLinear,
 class AWQColumnParallelLinear(ColumnParallelLinear):
 
     def create_weights(self, dtype: torch.dtype) -> None:
-        assert self.input_size % self.quant_config.weight_bits == 0
+        assert self.input_size % self.quant_config.group_size == 0
-        assert (self.output_size_per_partition %
+        if self.output_size_per_partition % self.quant_config.pack_factor != 0:
-                self.quant_config.pack_factor == 0)
+            raise ValueError(
+                "The tensor parallel size is not aligned with the quantized "
+                "weight shape. Please use a different tensor parallel size.")
         self.qweight = Parameter(
             torch.empty(
                 self.input_size,
@@ -62,9 +64,11 @@ class AWQColumnParallelLinear(ColumnParallelLinear):
 class AWQRowParallelLinear(RowParallelLinear):
 
     def create_weights(self, dtype: torch.dtype) -> None:
-        assert (self.input_size_per_partition %
-                self.quant_config.weight_bits == 0)
         assert self.output_size % self.quant_config.pack_factor == 0
+        if self.input_size_per_partition % self.quant_config.group_size != 0:
+            raise ValueError(
+                "The tensor parallel size is not aligned with the quantized "
+                "weight shape. Please use a different tensor parallel size.")
         self.qweight = Parameter(
             torch.empty(
                 self.input_size_per_partition,

vllm/model_executor/layers/quantized_linear/squeezellm.py

@@ -0,0 +1,84 @@
+from typing import Optional
+
+import torch
+from torch.nn.parameter import Parameter
+
+from vllm import quantization_ops
+from vllm.model_executor.parallel_utils.layers import (ColumnParallelLinear,
+                                                       RowParallelLinear)
+
+
+class SqueezeLLMColumnParallelLinear(ColumnParallelLinear):
+
+    def create_weights(self, dtype: torch.dtype) -> None:
+        assert self.input_size % self.quant_config.pack_factor == 0
+        self.qweight = Parameter(
+            torch.empty(
+                self.input_size // self.quant_config.pack_factor,
+                self.output_size_per_partition,
+                device="cuda",
+                dtype=torch.int32,
+            ),
+            requires_grad=False,
+        )
+        self.lookup_table = Parameter(
+            torch.empty(
+                self.output_size_per_partition,
+                self.quant_config.weight_bits**2,
+                device="cuda",
+                dtype=dtype,
+            ),
+            requires_grad=False,
+        )
+
+    def apply_weights(
+        self,
+        x: torch.Tensor,
+        bias: Optional[torch.Tensor],
+    ) -> torch.Tensor:
+        out_shape = x.shape[:-1] + (self.qweight.shape[-1], )
+        reshaped_x = x.reshape(-1, x.shape[-1])
+        # NOTE: The output tensor should be zero-initialized.
+        out = torch.zeros(out_shape, device="cuda", dtype=torch.float16)
+        quantization_ops.squeezellm_gemm(reshaped_x, self.qweight, out,
+                                         self.lookup_table)
+
+        if bias is not None:
+            out = out + bias
+        return out.reshape(out_shape)
+
+
+class SqueezeLLMRowParallelLinear(RowParallelLinear):
+
+    def create_weights(self, dtype: torch.dtype) -> None:
+        if self.input_size_per_partition % self.quant_config.pack_factor != 0:
+            raise ValueError(
+                "The tensor parallel size is not aligned with the quantized "
+                "weight shape. Please use a different tensor parallel size.")
+        self.qweight = Parameter(
+            torch.empty(
+                self.input_size_per_partition // self.quant_config.pack_factor,
+                self.output_size,
+                device="cuda",
+                dtype=torch.int32,
+            ),
+            requires_grad=False,
+        )
+        self.lookup_table = Parameter(
+            torch.empty(
+                self.output_size,
+                self.quant_config.weight_bits**2,
+                device="cuda",
+                dtype=dtype,
+            ),
+            requires_grad=False,
+        )
+
+    def apply_weights(self, x: torch.Tensor) -> torch.Tensor:
+        reshaped_x = x.reshape(-1, x.shape[-1])
+        out_shape = x.shape[:-1] + (self.qweight.shape[-1], )
+        # NOTE: The output tensor should be zero-initialized.
+        out = torch.zeros(out_shape, device="cuda", dtype=torch.float16)
+        quantization_ops.squeezellm_gemm(reshaped_x, self.qweight, out,
+                                         self.lookup_table)
+        return out.reshape(out_shape)

vllm/model_executor/models/llama.py

@@ -314,17 +314,21 @@ class LlamaForCausalLM(nn.Module):
                      load_format: str = "auto",
                      revision: Optional[str] = None):
         if self.quant_config is None:
-            weight_suffixes = ["weight"]
+            col_weight_suffixes = ["weight"]
+            row_weight_suffixes = ["weight"]
         else:
-            weight_suffixes = self.quant_config.get_tp_tensor_names()
+            col_weight_suffixes = (
+                self.quant_config.get_col_parallel_tensor_names())
+            row_weight_suffixes = (
+                self.quant_config.get_row_parallel_tensor_names())
 
         column_parallel_weights: List[str] = []
         for layer in self._column_parallel_layers:
-            for suffix in weight_suffixes:
+            for suffix in col_weight_suffixes:
                 column_parallel_weights.append(f"{layer}.{suffix}")
         row_parallel_weights: List[str] = []
         for layer in self._row_parallel_layers:
-            for suffix in weight_suffixes:
+            for suffix in row_weight_suffixes:
                 row_parallel_weights.append(f"{layer}.{suffix}")
 
         tp_size = get_tensor_model_parallel_world_size()
@@ -351,10 +355,10 @@ class LlamaForCausalLM(nn.Module):
             if "rotary_emb.inv_freq" in name:
                 continue
 
-            is_packed = False
+            packed_dim = None
             is_transposed = False
             if self.quant_config is not None:
-                is_packed = self.quant_config.is_packed(name)
+                packed_dim = self.quant_config.get_packed_dim(name)
                 is_transposed = self.quant_config.is_transposed(name)
             if is_transposed:
                 loaded_weight = convert_pyslice_to_tensor(loaded_weight)
@@ -368,7 +372,9 @@ class LlamaForCausalLM(nn.Module):
                 if is_transposed:
                     param = param.T
 
-                if is_packed:
+                if packed_dim is not None:
+                    shard_dim = 0 if not is_transposed else 1
+                    if packed_dim == shard_dim:
                         shard_size //= self.quant_config.pack_factor
                         offset //= self.quant_config.pack_factor
 

vllm/model_executor/models/mistral.py

@@ -298,17 +298,21 @@ class MistralForCausalLM(nn.Module):
                      load_format: str = "auto",
                      revision: Optional[str] = None):
         if self.quant_config is None:
-            weight_suffixes = ["weight"]
+            col_weight_suffixes = ["weight"]
+            row_weight_suffixes = ["weight"]
         else:
-            weight_suffixes = self.quant_config.get_tp_tensor_names()
+            col_weight_suffixes = (
+                self.quant_config.get_col_parallel_tensor_names())
+            row_weight_suffixes = (
+                self.quant_config.get_row_parallel_tensor_names())
 
         column_parallel_weights: List[str] = []
         for layer in self._column_parallel_layers:
-            for suffix in weight_suffixes:
+            for suffix in col_weight_suffixes:
                 column_parallel_weights.append(f"{layer}.{suffix}")
         row_parallel_weights: List[str] = []
         for layer in self._row_parallel_layers:
-            for suffix in weight_suffixes:
+            for suffix in row_weight_suffixes:
                 row_parallel_weights.append(f"{layer}.{suffix}")
 
         tp_size = get_tensor_model_parallel_world_size()
@@ -331,10 +335,10 @@ class MistralForCausalLM(nn.Module):
             if "rotary_emb.inv_freq" in name:
                 continue
 
-            is_packed = False
+            packed_dim = None
             is_transposed = False
             if self.quant_config is not None:
-                is_packed = self.quant_config.is_packed(name)
+                packed_dim = self.quant_config.get_packed_dim(name)
                 is_transposed = self.quant_config.is_transposed(name)
             if is_transposed:
                 loaded_weight = convert_pyslice_to_tensor(loaded_weight)
@@ -348,7 +352,9 @@ class MistralForCausalLM(nn.Module):
                 if is_transposed:
                     param = param.T
 
-                if is_packed:
+                if packed_dim is not None:
+                    shard_dim = 0 if not is_transposed else 1
+                    if packed_dim == shard_dim:
                         shard_size //= self.quant_config.pack_factor
                         offset //= self.quant_config.pack_factor
 

vllm/model_executor/quantization_utils/__init__.py

@@ -2,9 +2,11 @@ from typing import Type
 
 from vllm.model_executor.quantization_utils.awq import AWQConfig
 from vllm.model_executor.quantization_utils.base import QuantizationConfig
+from vllm.model_executor.quantization_utils.squeezellm import SqueezeLLMConfig
 
 _QUANTIZATION_REGISTRY = {
     "awq": AWQConfig,
+    "squeezellm": SqueezeLLMConfig,
 }
 
 

vllm/model_executor/quantization_utils/awq.py

@@ -60,13 +60,17 @@ class AWQConfig(QuantizationConfig):
         return cls(weight_bits, group_size, zero_point)
 
     @classmethod
-    def get_packed_tensor_names(cls) -> List[str]:
+    def get_packed_tensors(cls) -> Dict[str, int]:
-        return ["qweight", "qzeros"]
+        return {"qweight": 1, "qzeros": 1}
 
     @classmethod
     def get_transposed_tensor_names(cls) -> List[str]:
         return ["qweight", "qzeros", "scales"]
 
     @classmethod
-    def get_tp_tensor_names(cls) -> List[str]:
+    def get_col_parallel_tensor_names(cls) -> List[str]:
+        return ["qweight", "qzeros", "scales"]
+
+    @classmethod
+    def get_row_parallel_tensor_names(cls) -> List[str]:
         return ["qweight", "qzeros", "scales"]

vllm/model_executor/quantization_utils/base.py

@@ -1,4 +1,4 @@
-from typing import Any, Dict, List
+from typing import Any, Dict, List, Optional
 
 import torch
 
@@ -45,19 +45,25 @@ class QuantizationConfig:
                          "quantization config.")
 
     @classmethod
-    def get_packed_tensor_names(cls) -> List[str]:
+    def get_packed_tensors(cls) -> Dict[str, int]:
+        """Returns a dictionary of packed tensor names and their pack dims."""
         raise NotImplementedError
 
     @classmethod
-    def is_packed(cls, tensor_name: str) -> bool:
+    def get_packed_dim(cls, tensor_name: str) -> Optional[int]:
-        """Returns True if a tensor is packed.
+        """Returns the pack dim of a tensor if it is packed.
 
         A tensor is considered packed if each element in the tensor is a
         packed representation of multiple elements in the original tensor.
         For example, an INT32 element in the tensor may represent 8 INT4
         elements in the original tensor.
+        If the tensor is not packed, returns None.
         """
-        return any(tag in tensor_name for tag in cls.get_packed_tensor_names())
+        packed_tensors = cls.get_packed_tensors()
+        for packed_tensor_name, pack_dim in packed_tensors.items():
+            if packed_tensor_name in tensor_name:
+                return pack_dim
+        return None
 
     @classmethod
     def get_transposed_tensor_names(cls) -> List[str]:
@@ -71,5 +77,9 @@ class QuantizationConfig:
                    for tag in cls.get_transposed_tensor_names())
 
     @classmethod
-    def get_tp_tensor_names(cls) -> List[str]:
+    def get_col_parallel_tensor_names(cls) -> List[str]:
+        raise NotImplementedError
+
+    @classmethod
+    def get_row_parallel_tensor_names(cls) -> List[str]:
         raise NotImplementedError

vllm/model_executor/quantization_utils/squeezellm.py

@@ -0,0 +1,65 @@
+from typing import Any, Dict, List
+
+import torch
+
+from vllm.model_executor.quantization_utils.base import QuantizationConfig
+
+
+class SqueezeLLMConfig(QuantizationConfig):
+    """Config class for SqueezeLLM.
+
+    Reference: https://arxiv.org/pdf/2306.07629
+    """
+
+    def __init__(
+        self,
+        weight_bits: int,
+    ) -> None:
+        self.weight_bits = weight_bits
+
+        if self.weight_bits != 4:
+            raise ValueError(
+                "Currently, only 4-bit weight quantization is supported for "
+                f"SqueezeLLM, but got {self.weight_bits} bits.")
+
+        self.pack_factor = 32 // self.weight_bits
+
+    def __repr__(self) -> str:
+        return f"SqueezeLLMConfig(weight_bits={self.weight_bits})"
+
+    @classmethod
+    def get_name(cls) -> str:
+        return "squeezellm"
+
+    @classmethod
+    def get_supported_act_dtypes(cls) -> List[torch.dtype]:
+        return [torch.half]
+
+    @classmethod
+    def get_min_capability(cls) -> int:
+        return 70
+
+    @classmethod
+    def get_config_filenames(cls) -> List[str]:
+        return ["quant_config.json"]
+
+    @classmethod
+    def from_config(cls, config: Dict[str, Any]) -> "SqueezeLLMConfig":
+        weight_bits = cls.get_from_keys(config, ["wbits"])
+        return cls(weight_bits)
+
+    @classmethod
+    def get_packed_tensors(cls) -> Dict[str, int]:
+        return {"qweight": 0}
+
+    @classmethod
+    def get_transposed_tensor_names(cls) -> List[str]:
+        return ["qweight"]
+
+    @classmethod
+    def get_col_parallel_tensor_names(cls) -> List[str]:
+        return ["qweight", "lookup_table"]
+
+    @classmethod
+    def get_row_parallel_tensor_names(cls) -> List[str]:
+        return ["qweight"]