357 lines
16 KiB
Plaintext
357 lines
16 KiB
Plaintext
/*
|
|
* Copyright (c) 2024 by FlashInfer team.
|
|
*
|
|
* 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.
|
|
*/
|
|
#include <flashinfer/pos_enc.cuh>
|
|
|
|
#include "pytorch_extension_utils.h"
|
|
|
|
using namespace flashinfer;
|
|
|
|
void apply_rope(at::Tensor q, at::Tensor k, at::Tensor q_rope, at::Tensor k_rope, at::Tensor indptr,
|
|
at::Tensor offsets, int64_t rotary_dim, bool interleave, double rope_scale,
|
|
double rope_theta) {
|
|
CHECK_LAST_DIM_CONTIGUOUS(q);
|
|
CHECK_LAST_DIM_CONTIGUOUS(k);
|
|
CHECK_INPUT(indptr);
|
|
CHECK_INPUT(offsets);
|
|
|
|
auto device = q.device();
|
|
CHECK_EQ(k.device(), device);
|
|
CHECK_DIM(3, q); // q: (nnz, H_Q, D)
|
|
CHECK_DIM(3, k); // k: (nnz, H_K, D)
|
|
CHECK_DIM(1, indptr); // indptr: (B + 1)
|
|
CHECK_DIM(1, offsets); // offsets: (B)
|
|
CHECK_EQ(q.size(0), k.size(0));
|
|
CHECK_EQ(q.size(2), k.size(2));
|
|
unsigned int num_qo_heads = q.size(1);
|
|
unsigned int num_kv_heads = k.size(1);
|
|
unsigned int head_dim = q.size(2);
|
|
unsigned int batch_size = offsets.size(0);
|
|
CHECK_EQ(indptr.size(0), batch_size + 1);
|
|
size_t q_stride_n = q.stride(0);
|
|
size_t q_stride_h = q.stride(1);
|
|
size_t k_stride_n = k.stride(0);
|
|
size_t k_stride_h = k.stride(1);
|
|
size_t q_rope_stride_n = q_rope.stride(0);
|
|
size_t q_rope_stride_h = q_rope.stride(1);
|
|
size_t k_rope_stride_n = k_rope.stride(0);
|
|
size_t k_rope_stride_h = k_rope.stride(1);
|
|
CHECK_EQ(indptr.scalar_type(), offsets.scalar_type());
|
|
|
|
const c10::cuda::OptionalCUDAGuard device_guard(q.device());
|
|
auto stream = at::cuda::getCurrentCUDAStream();
|
|
DISPATCH_PYTORCH_DTYPE_TO_CTYPE_FP16(q.scalar_type(), c_type, [&] {
|
|
return DISPATCH_PYTORCH_IDTYPE_TO_CTYPE(indptr.scalar_type(), c_idtype, [&] {
|
|
cudaError_t status = BatchQKApplyRotary(
|
|
static_cast<c_type*>(q.data_ptr()), static_cast<c_type*>(k.data_ptr()),
|
|
static_cast<c_type*>(q_rope.data_ptr()), static_cast<c_type*>(k_rope.data_ptr()),
|
|
static_cast<c_idtype*>(indptr.data_ptr()), static_cast<c_idtype*>(offsets.data_ptr()),
|
|
batch_size, num_qo_heads, num_kv_heads, rotary_dim, head_dim, q_stride_n, q_stride_h,
|
|
k_stride_n, k_stride_h, q_rope_stride_n, q_rope_stride_h, k_rope_stride_n,
|
|
k_rope_stride_h, interleave, rope_scale, rope_theta, stream);
|
|
TORCH_CHECK(status == cudaSuccess, "BatchQKApplyRotary failed with error code " +
|
|
std::string(cudaGetErrorString(status)));
|
|
return true;
|
|
});
|
|
});
|
|
}
|
|
|
|
void apply_rope_pos_ids(at::Tensor q, at::Tensor k, at::Tensor q_rope, at::Tensor k_rope,
|
|
at::Tensor pos_ids, int64_t rotary_dim, bool interleave, double rope_scale,
|
|
double rope_theta) {
|
|
CHECK_LAST_DIM_CONTIGUOUS(q);
|
|
CHECK_LAST_DIM_CONTIGUOUS(k);
|
|
CHECK_INPUT(pos_ids);
|
|
|
|
auto device = q.device();
|
|
CHECK_EQ(k.device(), device);
|
|
CHECK_DIM(3, q); // q: (nnz, H_Q, D)
|
|
CHECK_DIM(3, k); // k: (nnz, H_K, D)
|
|
CHECK_EQ(q.size(0), k.size(0));
|
|
CHECK_EQ(q.size(2), k.size(2));
|
|
unsigned int num_qo_heads = q.size(1);
|
|
unsigned int num_kv_heads = k.size(1);
|
|
unsigned int head_dim = q.size(2);
|
|
unsigned int nnz = q.size(0);
|
|
size_t q_stride_n = q.stride(0);
|
|
size_t q_stride_h = q.stride(1);
|
|
size_t k_stride_n = k.stride(0);
|
|
size_t k_stride_h = k.stride(1);
|
|
size_t q_rope_stride_n = q_rope.stride(0);
|
|
size_t q_rope_stride_h = q_rope.stride(1);
|
|
size_t k_rope_stride_n = k_rope.stride(0);
|
|
size_t k_rope_stride_h = k_rope.stride(1);
|
|
|
|
const c10::cuda::OptionalCUDAGuard device_guard(q.device());
|
|
auto stream = at::cuda::getCurrentCUDAStream();
|
|
DISPATCH_PYTORCH_DTYPE_TO_CTYPE_FP16(q.scalar_type(), c_type, [&] {
|
|
return DISPATCH_PYTORCH_IDTYPE_TO_CTYPE(pos_ids.scalar_type(), c_idtype, [&] {
|
|
cudaError_t status = BatchQKApplyRotaryPosIds(
|
|
static_cast<c_type*>(q.data_ptr()), static_cast<c_type*>(k.data_ptr()),
|
|
static_cast<c_type*>(q_rope.data_ptr()), static_cast<c_type*>(k_rope.data_ptr()),
|
|
static_cast<c_idtype*>(pos_ids.data_ptr()), nnz, num_qo_heads, num_kv_heads, rotary_dim,
|
|
head_dim, q_stride_n, q_stride_h, k_stride_n, k_stride_h, q_rope_stride_n,
|
|
q_rope_stride_h, k_rope_stride_n, k_rope_stride_h, interleave, rope_scale, rope_theta,
|
|
stream);
|
|
TORCH_CHECK(status == cudaSuccess, "BatchQKApplyRotaryPosIds failed with error code " +
|
|
std::string(cudaGetErrorString(status)));
|
|
return true;
|
|
});
|
|
});
|
|
}
|
|
|
|
void apply_rope_pos_ids_cos_sin_cache(at::Tensor q, at::Tensor k, at::Tensor q_rope,
|
|
at::Tensor k_rope, at::Tensor cos_sin_cache,
|
|
at::Tensor pos_ids, bool interleave) {
|
|
CHECK_LAST_DIM_CONTIGUOUS(q);
|
|
CHECK_LAST_DIM_CONTIGUOUS(k);
|
|
CHECK_INPUT(cos_sin_cache);
|
|
CHECK_INPUT(pos_ids);
|
|
auto device = q.device();
|
|
CHECK_EQ(k.device(), device);
|
|
CHECK_EQ(cos_sin_cache.device(), device);
|
|
CHECK_EQ(pos_ids.device(), device);
|
|
CHECK_DIM(3, q); // q: (nnz, H_Q, D)
|
|
CHECK_DIM(3, k); // k: (nnz, H_K, D)
|
|
// cos_sin_cache: (max_seq_len, R)
|
|
// First half of R is cos, second half is sin
|
|
CHECK_DIM(2, cos_sin_cache);
|
|
CHECK_EQ(q.size(0), k.size(0));
|
|
CHECK_EQ(q.size(2), k.size(2));
|
|
unsigned int rotary_dim = cos_sin_cache.size(1);
|
|
unsigned int num_qo_heads = q.size(1);
|
|
unsigned int num_kv_heads = k.size(1);
|
|
unsigned int head_dim = q.size(2);
|
|
unsigned int nnz = q.size(0);
|
|
size_t q_stride_n = q.stride(0);
|
|
size_t q_stride_h = q.stride(1);
|
|
size_t k_stride_n = k.stride(0);
|
|
size_t k_stride_h = k.stride(1);
|
|
size_t q_rope_stride_n = q_rope.stride(0);
|
|
size_t q_rope_stride_h = q_rope.stride(1);
|
|
size_t k_rope_stride_n = k_rope.stride(0);
|
|
size_t k_rope_stride_h = k_rope.stride(1);
|
|
|
|
const c10::cuda::OptionalCUDAGuard device_guard(q.device());
|
|
auto stream = at::cuda::getCurrentCUDAStream();
|
|
DISPATCH_PYTORCH_DTYPE_TO_CTYPE_FP16(q.scalar_type(), c_type, [&] {
|
|
return DISPATCH_PYTORCH_IDTYPE_TO_CTYPE(pos_ids.scalar_type(), c_idtype, [&] {
|
|
cudaError_t status = BatchQKApplyRotaryPosIdsCosSinCache(
|
|
static_cast<c_type*>(q.data_ptr()), static_cast<c_type*>(k.data_ptr()),
|
|
static_cast<c_type*>(q_rope.data_ptr()), static_cast<c_type*>(k_rope.data_ptr()),
|
|
static_cast<float*>(cos_sin_cache.data_ptr()), static_cast<c_idtype*>(pos_ids.data_ptr()),
|
|
nnz, num_qo_heads, num_kv_heads, rotary_dim, head_dim, q_stride_n, q_stride_h, k_stride_n,
|
|
k_stride_h, q_rope_stride_n, q_rope_stride_h, k_rope_stride_n, k_rope_stride_h,
|
|
interleave, stream);
|
|
TORCH_CHECK(status == cudaSuccess,
|
|
"BatchQKApplyRotaryPosIdsCosSinCache failed with error code " +
|
|
std::string(cudaGetErrorString(status)));
|
|
return true;
|
|
});
|
|
});
|
|
}
|
|
|
|
void apply_llama31_rope(at::Tensor q, at::Tensor k, at::Tensor q_rope, at::Tensor k_rope,
|
|
at::Tensor indptr, at::Tensor offsets, int64_t rotary_dim, bool interleave,
|
|
double rope_scale, double rope_theta, double low_freq_factor,
|
|
double high_freq_factor, double old_context_length) {
|
|
CHECK_CUDA(q); // not necessarily contiguous
|
|
CHECK_CUDA(k); // not necessarily contiguous
|
|
CHECK_INPUT(indptr);
|
|
CHECK_INPUT(offsets);
|
|
|
|
auto device = q.device();
|
|
CHECK_EQ(k.device(), device);
|
|
CHECK_DIM(3, q); // q: (nnz, H_Q, D)
|
|
CHECK_DIM(3, k); // k: (nnz, H_K, D)
|
|
CHECK_DIM(1, indptr); // indptr: (B + 1)
|
|
CHECK_DIM(1, offsets); // offsets: (B)
|
|
CHECK_EQ(q.size(0), k.size(0));
|
|
CHECK_EQ(q.size(2), k.size(2));
|
|
unsigned int num_qo_heads = q.size(1);
|
|
unsigned int num_kv_heads = k.size(1);
|
|
unsigned int head_dim = q.size(2);
|
|
unsigned int batch_size = offsets.size(0);
|
|
CHECK_EQ(indptr.size(0), batch_size + 1);
|
|
CHECK_EQ(indptr.scalar_type(), offsets.scalar_type());
|
|
size_t q_stride_n = q.stride(0);
|
|
size_t q_stride_h = q.stride(1);
|
|
size_t k_stride_n = k.stride(0);
|
|
size_t k_stride_h = k.stride(1);
|
|
size_t q_rope_stride_n = q_rope.stride(0);
|
|
size_t q_rope_stride_h = q_rope.stride(1);
|
|
size_t k_rope_stride_n = k_rope.stride(0);
|
|
size_t k_rope_stride_h = k_rope.stride(1);
|
|
CHECK_EQ(indptr.scalar_type(), offsets.scalar_type());
|
|
|
|
const c10::cuda::OptionalCUDAGuard device_guard(q.device());
|
|
auto stream = at::cuda::getCurrentCUDAStream();
|
|
DISPATCH_PYTORCH_DTYPE_TO_CTYPE_FP16(q.scalar_type(), c_type, [&] {
|
|
return DISPATCH_PYTORCH_IDTYPE_TO_CTYPE(indptr.scalar_type(), c_idtype, [&] {
|
|
cudaError_t status = BatchQKApplyLlama31Rotary(
|
|
static_cast<c_type*>(q.data_ptr()), static_cast<c_type*>(k.data_ptr()),
|
|
static_cast<c_type*>(q_rope.data_ptr()), static_cast<c_type*>(k_rope.data_ptr()),
|
|
static_cast<c_idtype*>(indptr.data_ptr()), static_cast<c_idtype*>(offsets.data_ptr()),
|
|
batch_size, num_qo_heads, num_kv_heads, rotary_dim, head_dim, q_stride_n, q_stride_h,
|
|
k_stride_n, k_stride_h, q_rope_stride_n, q_rope_stride_h, k_rope_stride_n,
|
|
k_rope_stride_h, interleave, rope_scale, rope_theta, low_freq_factor, high_freq_factor,
|
|
old_context_length, stream);
|
|
TORCH_CHECK(status == cudaSuccess, "BatchQKApplyLlama31Rotary failed with error code " +
|
|
std::string(cudaGetErrorString(status)));
|
|
return true;
|
|
});
|
|
});
|
|
}
|
|
|
|
void apply_llama31_rope_pos_ids(at::Tensor q, at::Tensor k, at::Tensor q_rope, at::Tensor k_rope,
|
|
at::Tensor pos_ids, int64_t rotary_dim, bool interleave,
|
|
double rope_scale, double rope_theta, double low_freq_factor,
|
|
double high_freq_factor, double old_context_length) {
|
|
CHECK_CUDA(q); // not necessarily contiguous
|
|
CHECK_CUDA(k); // not necessarily contiguous
|
|
CHECK_INPUT(pos_ids);
|
|
|
|
auto device = q.device();
|
|
CHECK_EQ(k.device(), device);
|
|
CHECK_DIM(3, q); // q: (nnz, H_Q, D)
|
|
CHECK_DIM(3, k); // k: (nnz, H_K, D)
|
|
CHECK_EQ(q.size(0), k.size(0));
|
|
CHECK_EQ(q.size(2), k.size(2));
|
|
unsigned int num_qo_heads = q.size(1);
|
|
unsigned int num_kv_heads = k.size(1);
|
|
unsigned int head_dim = q.size(2);
|
|
unsigned int nnz = q.size(0);
|
|
size_t q_stride_n = q.stride(0);
|
|
size_t q_stride_h = q.stride(1);
|
|
size_t k_stride_n = k.stride(0);
|
|
size_t k_stride_h = k.stride(1);
|
|
size_t q_rope_stride_n = q_rope.stride(0);
|
|
size_t q_rope_stride_h = q_rope.stride(1);
|
|
size_t k_rope_stride_n = k_rope.stride(0);
|
|
size_t k_rope_stride_h = k_rope.stride(1);
|
|
|
|
const c10::cuda::OptionalCUDAGuard device_guard(q.device());
|
|
auto stream = at::cuda::getCurrentCUDAStream();
|
|
DISPATCH_PYTORCH_DTYPE_TO_CTYPE_FP16(q.scalar_type(), c_type, [&] {
|
|
return DISPATCH_PYTORCH_IDTYPE_TO_CTYPE(pos_ids.scalar_type(), c_idtype, [&] {
|
|
cudaError_t status = BatchQKApplyLlama31RotaryPosIds(
|
|
static_cast<c_type*>(q.data_ptr()), static_cast<c_type*>(k.data_ptr()),
|
|
static_cast<c_type*>(q_rope.data_ptr()), static_cast<c_type*>(k_rope.data_ptr()),
|
|
static_cast<c_idtype*>(pos_ids.data_ptr()), nnz, num_qo_heads, num_kv_heads, rotary_dim,
|
|
head_dim, q_stride_n, q_stride_h, k_stride_n, k_stride_h, q_rope_stride_n,
|
|
q_rope_stride_h, k_rope_stride_n, k_rope_stride_h, interleave, rope_scale, rope_theta,
|
|
low_freq_factor, high_freq_factor, old_context_length, stream);
|
|
TORCH_CHECK(status == cudaSuccess, "BatchQKApplyLlama31RotaryPosIds failed with error code " +
|
|
std::string(cudaGetErrorString(status)));
|
|
return true;
|
|
});
|
|
});
|
|
}
|
|
|
|
void mla_rope_quantize(at::Tensor q_rope_in, at::Tensor k_rope_in, at::Tensor q_nope_in,
|
|
at::Tensor k_nope_in, at::Tensor q_rope_out, at::Tensor k_rope_out,
|
|
at::Tensor q_nope_out, at::Tensor k_nope_out, at::Tensor cos_sin_cache,
|
|
at::Tensor pos_ids, double quant_scale_q, double quant_scale_kv,
|
|
bool interleave) {
|
|
CHECK_LAST_DIM_CONTIGUOUS(q_rope_in);
|
|
CHECK_LAST_DIM_CONTIGUOUS(k_rope_in);
|
|
CHECK_LAST_DIM_CONTIGUOUS(q_nope_in);
|
|
CHECK_LAST_DIM_CONTIGUOUS(k_nope_in);
|
|
CHECK_LAST_DIM_CONTIGUOUS(q_rope_out);
|
|
CHECK_LAST_DIM_CONTIGUOUS(k_rope_out);
|
|
CHECK_LAST_DIM_CONTIGUOUS(q_nope_out);
|
|
CHECK_LAST_DIM_CONTIGUOUS(k_nope_out);
|
|
CHECK_INPUT(cos_sin_cache);
|
|
CHECK_INPUT(pos_ids);
|
|
|
|
CHECK_EQ(q_rope_in.size(-1), 64);
|
|
CHECK_EQ(k_rope_in.size(-1), 64);
|
|
CHECK_EQ(q_nope_in.size(-1), 512);
|
|
CHECK_EQ(k_nope_in.size(-1), 512);
|
|
CHECK_EQ(q_rope_out.size(-1), 64);
|
|
CHECK_EQ(k_rope_out.size(-1), 64);
|
|
CHECK_EQ(q_nope_out.size(-1), 512);
|
|
CHECK_EQ(k_nope_out.size(-1), 512);
|
|
auto scalar_type_in = q_rope_in.scalar_type();
|
|
TORCH_CHECK(scalar_type_in == k_rope_in.scalar_type());
|
|
TORCH_CHECK(scalar_type_in == q_nope_in.scalar_type());
|
|
TORCH_CHECK(scalar_type_in == k_nope_in.scalar_type());
|
|
auto quant_type_out = q_rope_out.scalar_type();
|
|
TORCH_CHECK(quant_type_out == k_rope_out.scalar_type());
|
|
TORCH_CHECK(quant_type_out == q_nope_out.scalar_type());
|
|
TORCH_CHECK(quant_type_out == k_nope_out.scalar_type());
|
|
|
|
CHECK_DIM(3, q_rope_in); // q_rope_in: (nnz, H_Q, 64)
|
|
CHECK_DIM(3, q_nope_in); // q_nope_in: (nnz, H_Q, 512)
|
|
CHECK_DIM(2, k_rope_in); // k_rope_in: (nnz, 64)
|
|
CHECK_DIM(2, k_nope_in); // k_nope_in: (nnz, 512)
|
|
CHECK_DIM(3, q_rope_out); // q_rope_out: (nnz, H_Q, 64)
|
|
CHECK_DIM(3, q_nope_out); // q_nope_out: (nnz, H_Q, 512)
|
|
CHECK_DIM(2, k_rope_out); // k_rope_out: (nnz, 64)
|
|
CHECK_DIM(2, k_nope_out); // k_nope_out: (nnz, 512)
|
|
uint32_t nnz = q_rope_in.size(0);
|
|
CHECK_EQ(q_nope_in.size(0), nnz);
|
|
CHECK_EQ(k_nope_in.size(0), nnz);
|
|
CHECK_EQ(q_rope_out.size(0), nnz);
|
|
CHECK_EQ(k_rope_out.size(0), nnz);
|
|
CHECK_EQ(q_nope_out.size(0), nnz);
|
|
CHECK_EQ(k_nope_out.size(0), nnz);
|
|
uint32_t num_heads = q_rope_in.size(1);
|
|
CHECK_EQ(q_rope_in.size(1), num_heads);
|
|
CHECK_EQ(q_nope_in.size(1), num_heads);
|
|
CHECK_EQ(q_rope_out.size(1), num_heads);
|
|
CHECK_EQ(q_nope_out.size(1), num_heads);
|
|
|
|
const uint32_t q_rope_in_stride_n = q_rope_in.stride(0);
|
|
const uint32_t q_rope_in_stride_h = q_rope_in.stride(1);
|
|
const uint32_t q_nope_in_stride_n = q_nope_in.stride(0);
|
|
const uint32_t q_nope_in_stride_h = q_nope_in.stride(1);
|
|
const uint32_t q_rope_out_stride_n = q_rope_out.stride(0);
|
|
const uint32_t q_rope_out_stride_h = q_rope_out.stride(1);
|
|
const uint32_t q_nope_out_stride_n = q_nope_out.stride(0);
|
|
const uint32_t q_nope_out_stride_h = q_nope_out.stride(1);
|
|
const uint32_t k_rope_in_stride = k_rope_in.stride(0);
|
|
const uint32_t k_nope_in_stride = k_nope_in.stride(0);
|
|
const uint32_t k_rope_out_stride = k_rope_out.stride(0);
|
|
const uint32_t k_nope_out_stride = k_nope_out.stride(0);
|
|
|
|
const c10::cuda::OptionalCUDAGuard device_guard(q_rope_in.device());
|
|
auto stream = at::cuda::getCurrentCUDAStream();
|
|
|
|
DISPATCH_PYTORCH_DTYPE_TO_CTYPE_FP16(scalar_type_in, c_type, [&] {
|
|
return DISPATCH_PYTORCH_DTYPE_TO_CTYPE_FP8(quant_type_out, c_quant_type, [&] {
|
|
return DISPATCH_PYTORCH_IDTYPE_TO_CTYPE(pos_ids.scalar_type(), c_idtype, [&] {
|
|
cudaError_t status = MLARopeQuantize(
|
|
static_cast<c_type*>(q_rope_in.data_ptr()), static_cast<c_type*>(k_rope_in.data_ptr()),
|
|
static_cast<c_type*>(q_nope_in.data_ptr()), static_cast<c_type*>(k_nope_in.data_ptr()),
|
|
static_cast<c_quant_type*>(q_rope_out.data_ptr()),
|
|
static_cast<c_quant_type*>(k_rope_out.data_ptr()),
|
|
static_cast<c_quant_type*>(q_nope_out.data_ptr()),
|
|
static_cast<c_quant_type*>(k_nope_out.data_ptr()),
|
|
static_cast<float*>(cos_sin_cache.data_ptr()),
|
|
static_cast<c_idtype*>(pos_ids.data_ptr()), nnz, num_heads, q_rope_in_stride_n,
|
|
q_rope_in_stride_h, q_nope_in_stride_n, q_nope_in_stride_h, q_rope_out_stride_n,
|
|
q_rope_out_stride_h, q_nope_out_stride_n, q_nope_out_stride_h, k_rope_in_stride,
|
|
k_nope_in_stride, k_rope_out_stride, k_nope_out_stride, quant_scale_q, quant_scale_kv,
|
|
interleave, stream);
|
|
TORCH_CHECK(status == cudaSuccess,
|
|
"BatchQKApplyRotaryPosIdsCosSinCache failed with error code " +
|
|
std::string(cudaGetErrorString(status)));
|
|
return true;
|
|
});
|
|
});
|
|
});
|
|
}
|