[Kernel][Misc] register ops to prevent graph breaks (vllm-project#6917)

Co-authored-by: Sage Moore <sage@neuralmagic.com>

.github/PULL_REQUEST_TEMPLATE.md

@@ -39,6 +39,16 @@ FIX #xxxx (*link existing issues this PR will resolve*)
     <li>Please add documentation to <code>docs/source/</code> if the PR modifies the user-facing behaviors of vLLM. It helps vLLM user understand and utilize the new features or changes.</li>
 </ul>
 
+<h3>Adding or changing kernels</h3>
+<p>Each custom kernel needs a schema and one or more implementations to be registered with PyTorch.</p>
+<ul>
+    <li>Make sure custom ops are registered following PyTorch guidelines: <a href="https://pytorch.org/tutorials/advanced/cpp_custom_ops.html#cpp-custom-ops-tutorial">Custom C++ and CUDA Operators</a> and <a href="https://docs.google.com/document/d/1_W62p8WJOQQUzPsJYa7s701JXt0qf2OfLub2sbkHOaU">The Custom Operators Manual</a></li>
+    <li>Custom operations that return <code>Tensors</code> require meta-functions. Meta-functions should be implemented and registered in python so that dynamic dims can be handled automatically. See above documents for a description of meta-functions.</li>
+    <li>Use <a href="https://pytorch.org/docs/stable/library.html#torch.library.opcheck"><code>torch.libary.opcheck()</code></a> to test the function registration and meta-function for any registered ops.  See <code>tests/kernels</code> for examples.</li>
+    <li>When changing the C++ signature of an existing op, the schema must be updated to reflect the changes.</li>
+    <li>If a new custom type is needed, see the following document: <a href="https://docs.google.com/document/d/18fBMPuOJ0fY5ZQ6YyrHUppw9FA332CpNtgB6SOIgyuA">Custom Class Support in PT2</a>.
+</ul>
+
 <h3>Notes for Large Changes</h3>
 <p>Please keep the changes as concise as possible. For major architectural changes (>500 LOC excluding kernel/data/config/test), we would expect a GitHub issue (RFC) discussing the technical design and justification. Otherwise, we will tag it with <code>rfc-required</code> and might not go through the PR.</p>
 

cmake/utils.cmake

@@ -350,6 +350,7 @@ function (define_gpu_extension_target GPU_MOD_NAME)
   target_include_directories(${GPU_MOD_NAME} PRIVATE csrc
     ${GPU_INCLUDE_DIRECTORIES})
 
+  # TODO: is torch_python_LIBRARY needed?
   target_link_libraries(${GPU_MOD_NAME} PRIVATE torch ${torch_python_LIBRARY}
     ${GPU_LIBRARIES})
 

csrc/cpu/torch_bindings.cpp

@@ -32,8 +32,8 @@ TORCH_LIBRARY_EXPAND(TORCH_EXTENSION_NAME, ops) {
   // PagedAttention V2.
   ops.def(
       "paged_attention_v2("
-      "    Tensor! out, Tensor exp_sums, Tensor max_logits,"
-      "    Tensor tmp_out, Tensor query, Tensor key_cache,"
+      "    Tensor! out, Tensor! exp_sums, Tensor! max_logits,"
+      "    Tensor! tmp_out, Tensor query, Tensor key_cache,"
       "    Tensor value_cache, int num_kv_heads, float scale,"
       "    Tensor block_tables, Tensor seq_lens, int block_size,"
       "    int max_seq_len, Tensor? alibi_slopes,"
@@ -122,8 +122,8 @@ TORCH_LIBRARY_EXPAND(CONCAT(TORCH_EXTENSION_NAME, _cache_ops), cache_ops) {
 
   // Copy the cache blocks from src to dst.
   cache_ops.def(
-      "copy_blocks(Tensor[]! key_caches, Tensor[]! value_caches, Tensor "
-      "block_mapping) -> ()");
+      "copy_blocks(Tensor(a!)[] key_caches, Tensor[](b!) value_caches, "
+      "Tensor block_mapping) -> ()");
   cache_ops.impl("copy_blocks", torch::kCPU, &copy_blocks);
 
   // Reshape the key and value tensors and cache them.

csrc/ops.h

@@ -123,9 +123,17 @@ torch::Tensor gptq_marlin_repack(torch::Tensor& b_q_weight, torch::Tensor& perm,
                                  int64_t size_k, int64_t size_n,
                                  int64_t num_bits);
 
+torch::Tensor gptq_marlin_repack_meta(torch::Tensor& b_q_weight,
+                                      torch::Tensor& perm, c10::SymInt size_k,
+                                      c10::SymInt size_n, int64_t num_bits);
+
 torch::Tensor awq_marlin_repack(torch::Tensor& b_q_weight, int64_t size_k,
                                 int64_t size_n, int64_t num_bits);
 
+torch::Tensor awq_marlin_repack_meta(torch::Tensor& b_q_weight,
+                                     c10::SymInt size_k, c10::SymInt size_n,
+                                     int64_t num_bits);
+
 torch::Tensor ggml_dequantize(torch::Tensor W, int64_t type, int64_t m,
                               int64_t n);
 

csrc/quantization/gptq_marlin/awq_marlin_repack.cu

@@ -267,3 +267,15 @@ torch::Tensor awq_marlin_repack(torch::Tensor& b_q_weight, int64_t size_k,
 }
 
 #endif
+
+torch::Tensor awq_marlin_repack_meta(torch::Tensor& b_q_weight,
+                                     c10::SymInt size_k, c10::SymInt size_n,
+                                     int64_t num_bits) {
+  int const pack_factor = 32 / num_bits;
+  auto options = torch::TensorOptions()
+                     .dtype(b_q_weight.dtype())
+                     .device(b_q_weight.device());
+  return torch::empty_symint(
+      {size_k / marlin::tile_size, size_n * marlin::tile_size / pack_factor},
+      options);
+}

csrc/quantization/gptq_marlin/gptq_marlin_repack.cu

@@ -342,3 +342,15 @@ torch::Tensor gptq_marlin_repack(torch::Tensor& b_q_weight, torch::Tensor& perm,
 }
 
 #endif
+
+torch::Tensor gptq_marlin_repack_meta(torch::Tensor& b_q_weight,
+                                      torch::Tensor& perm, c10::SymInt size_k,
+                                      c10::SymInt size_n, int64_t num_bits) {
+  int const pack_factor = 32 / num_bits;
+  auto options = torch::TensorOptions()
+                     .dtype(b_q_weight.dtype())
+                     .device(b_q_weight.device());
+  return torch::empty_symint(
+      {size_k / marlin::tile_size, size_n * marlin::tile_size / pack_factor},
+      options);
+}

csrc/torch_bindings.cpp

@@ -36,8 +36,8 @@ TORCH_LIBRARY_EXPAND(TORCH_EXTENSION_NAME, ops) {
   // PagedAttention V2.
   ops.def(
       "paged_attention_v2("
-      "    Tensor! out, Tensor exp_sums, Tensor max_logits,"
-      "    Tensor tmp_out, Tensor query, Tensor key_cache,"
+      "    Tensor! out, Tensor! exp_sums, Tensor! max_logits,"
+      "    Tensor! tmp_out, Tensor query, Tensor key_cache,"
       "    Tensor value_cache, int num_kv_heads, float scale,"
       "    Tensor block_tables, Tensor seq_lens, int block_size,"
       "    int max_seq_len, Tensor? alibi_slopes,"
@@ -73,7 +73,11 @@ TORCH_LIBRARY_EXPAND(TORCH_EXTENSION_NAME, ops) {
   ops.impl("gelu_quick", torch::kCUDA, &gelu_quick);
 
   // prepare_inputs advance_step
-  ops.def("advance_step", &advance_step);
+  ops.def(
+      "advance_step(int num_seqs, int num_queries, int block_size, "
+      "Tensor! input_tokens, Tensor sampled_token_ids, "
+      "Tensor! input_positions, Tensor! seq_lens, Tensor! slot_mapping, "
+      "Tensor block_tables) -> ()");
   ops.impl("advance_step", torch::kCUDA, &advance_step);
 
   // Layernorm
@@ -110,27 +114,56 @@ TORCH_LIBRARY_EXPAND(TORCH_EXTENSION_NAME, ops) {
   // Quantization ops
 #ifndef USE_ROCM
   // Quantized GEMM for AQLM.
-  ops.def("aqlm_gemm", &aqlm_gemm);
+  ops.def(
+      "aqlm_gemm(Tensor input, Tensor codes, Tensor codebooks, "
+      "Tensor scales, int[] codebook_partition_sizes, Tensor? bias) "
+      "-> Tensor");
   ops.impl("aqlm_gemm", torch::kCUDA, &aqlm_gemm);
 
   // Decompression method for AQLM.
-  ops.def("aqlm_dequant", &aqlm_dequant);
+  ops.def(
+      "aqlm_dequant(Tensor codes, Tensor codebooks, "
+      "int[] codebook_partition_sizes) -> Tensor");
   ops.impl("aqlm_dequant", torch::kCUDA, &aqlm_dequant);
 
   // Quantized GEMM for AWQ.
-  ops.def("awq_gemm", &awq_gemm);
+  ops.def(
+      "awq_gemm(Tensor _in_feats, Tensor _kernel, Tensor _scaling_factors, "
+      "Tensor _zeros, int split_k_iters) -> Tensor");
   ops.impl("awq_gemm", torch::kCUDA, &awq_gemm);
 
   // Dequantization for AWQ.
-  ops.def("awq_dequantize", &awq_dequantize);
+  ops.def(
+      "awq_dequantize(Tensor _kernel, Tensor _scaling_factors, "
+      "Tensor _zeros, int split_k_iters, int thx, int thy) -> Tensor");
   ops.impl("awq_dequantize", torch::kCUDA, &awq_dequantize);
 
+  // Note about marlin kernel 'workspace' arguments:
+  // Technically these should be mutable since they are modified by the kernel.
+  // But since they are set back to zero once the kernel is finished we can
+  // hand wave and say that they have no net effect.
+  //
+  // The reason to mark 'workspace' as immutable is so that they don't interfere
+  // with using ScalarType arguments in the ops. If they are marked as mutable,
+  // pytorch throws an assert in
+  // 'torch._higher_order_ops._register_effectful_op' that prevents these
+  // kernels from being torch.compile'd.
+  // See the following document for more info on custom types and ops that use
+  // custom types:
+  // https://docs.google.com/document/d/18fBMPuOJ0fY5ZQ6YyrHUppw9FA332CpNtgB6SOIgyuA
+
   // Marlin (Dense) Optimized Quantized GEMM for GPTQ.
-  ops.def("marlin_gemm", &marlin_gemm);
+  ops.def(
+      "marlin_gemm(Tensor a, Tensor b_q_weight, Tensor b_scales, "
+      "Tensor! workspace, int size_m, int size_n, int size_k) -> Tensor");
   ops.impl("marlin_gemm", torch::kCUDA, &marlin_gemm);
 
   // Marlin_24 (Sparse) Optimized Quantized GEMM for GPTQ.
-  ops.def("gptq_marlin_24_gemm", &gptq_marlin_24_gemm);
+  ops.def(
+      "gptq_marlin_24_gemm(Tensor a, Tensor b_q_weight, Tensor b_meta, "
+      "Tensor b_scales, Tensor workspace, "
+      "__torch__.torch.classes._core_C.ScalarType b_q_type, "
+      "int size_m, int size_n, int size_k) -> Tensor");
   ops.impl("gptq_marlin_24_gemm", torch::kCUDA, &gptq_marlin_24_gemm);
 
   // Machete (Dense) Optimized Mixed Precision GEMM for Hopper.
@@ -149,35 +182,55 @@ TORCH_LIBRARY_EXPAND(TORCH_EXTENSION_NAME, ops) {
   ops.impl("machete_prepack_B", torch::kCUDA, &machete::prepack_B);
 
   // gptq_marlin Optimized Quantized GEMM for GPTQ.
-  ops.def("gptq_marlin_gemm", &gptq_marlin_gemm);
+  ops.def(
+      "gptq_marlin_gemm(Tensor a, Tensor b_q_weight, Tensor b_scales, "
+      "Tensor b_zeros, Tensor g_idx, Tensor perm, Tensor workspace, "
+      "__torch__.torch.classes._core_C.ScalarType b_q_type, "
+      "int size_m, int size_n, int size_k, bool is_k_full, "
+      "bool has_zp, bool use_fp32_reduce) -> Tensor");
   ops.impl("gptq_marlin_gemm", torch::kCUDA, &gptq_marlin_gemm);
 
   // gptq_marlin repack from GPTQ.
-  ops.def("gptq_marlin_repack", &gptq_marlin_repack);
+  ops.def(
+      "gptq_marlin_repack(Tensor b_q_weight, Tensor perm, "
+      "SymInt size_k, SymInt size_n, int num_bits) -> Tensor");
   ops.impl("gptq_marlin_repack", torch::kCUDA, &gptq_marlin_repack);
+  ops.impl("gptq_marlin_repack", torch::kMeta, &gptq_marlin_repack_meta);
 
   // awq_marlin repack from AWQ.
-  ops.def("awq_marlin_repack", &awq_marlin_repack);
+  ops.def(
+      "awq_marlin_repack(Tensor b_q_weight, SymInt size_k, "
+      "SymInt size_n, int num_bits) -> Tensor");
   ops.impl("awq_marlin_repack", torch::kCUDA, &awq_marlin_repack);
+  ops.impl("awq_marlin_repack", torch::kMeta, &awq_marlin_repack_meta);
 
   // Dequantization for GGML.
-  ops.def("ggml_dequantize", &ggml_dequantize);
+  ops.def("ggml_dequantize(Tensor W, int type, int m, int n) -> Tensor");
   ops.impl("ggml_dequantize", torch::kCUDA, &ggml_dequantize);
 
   // mmvq kernel for GGML.
-  ops.def("ggml_mul_mat_vec_a8", &ggml_mul_mat_vec_a8);
+  ops.def(
+      "ggml_mul_mat_vec_a8(Tensor W, Tensor X, int type, int row) "
+      "-> Tensor");
   ops.impl("ggml_mul_mat_vec_a8", torch::kCUDA, &ggml_mul_mat_vec_a8);
 
   // mmq kernel for GGML.
-  ops.def("ggml_mul_mat_a8", &ggml_mul_mat_a8);
+  ops.def("ggml_mul_mat_a8(Tensor W, Tensor X, int type, int row) -> Tensor");
   ops.impl("ggml_mul_mat_a8", torch::kCUDA, &ggml_mul_mat_a8);
 
   // fp8_marlin Optimized Quantized GEMM for FP8 weight-only.
-  ops.def("fp8_marlin_gemm", &fp8_marlin_gemm);
+  ops.def(
+      "fp8_marlin_gemm(Tensor a, Tensor b_q_weight, Tensor b_scales, "
+      "Tensor! workspace, int num_bits, int size_m, int size_n, "
+      "int size_k) -> Tensor");
   ops.impl("fp8_marlin_gemm", torch::kCUDA, &fp8_marlin_gemm);
 
   // marlin_qqq_gemm for QQQ.
-  ops.def("marlin_qqq_gemm", &marlin_qqq_gemm);
+  ops.def(
+      "marlin_qqq_gemm(Tensor a, Tensor b_q_weight, "
+      "Tensor s_tok, Tensor s_ch, Tensor s_group, "
+      "Tensor! workspace, int size_m, int size_n, "
+      "int size_k) -> Tensor");
   ops.impl("marlin_qqq_gemm", torch::kCUDA, &marlin_qqq_gemm);
 
   // CUTLASS w8a8 GEMM, supporting symmetric per-tensor or per-row/column
@@ -199,16 +252,16 @@ TORCH_LIBRARY_EXPAND(TORCH_EXTENSION_NAME, ops) {
 
   // Check if cutlass scaled_mm is supported for CUDA devices of the given
   // capability
-  ops.def("cutlass_scaled_mm_supports_fp8", &cutlass_scaled_mm_supports_fp8);
-  ops.impl("cutlass_scaled_mm_supports_fp8", torch::kCUDA,
-           &cutlass_scaled_mm_supports_fp8);
+  ops.def("cutlass_scaled_mm_supports_fp8(int cuda_device_capability) -> bool");
+  ops.impl("cutlass_scaled_mm_supports_fp8", &cutlass_scaled_mm_supports_fp8);
+
   // Mamba selective scan kernel
   ops.def(
       "selective_scan_fwd(Tensor! u, Tensor! delta,"
       "Tensor! A, Tensor! B, Tensor! C,"
       "Tensor? D_, Tensor? z_, Tensor? delta_bias_,"
       "bool delta_softplus,"
-      "Tensor? index_, Tensor? x) -> Tensor[]");
+      "Tensor? index_, Tensor(a! -> *)? x) -> Tensor(a)[]");
   ops.impl("selective_scan_fwd", torch::kCUDA, &selective_scan_fwd);
 
   ops.def(
@@ -230,7 +283,12 @@ TORCH_LIBRARY_EXPAND(TORCH_EXTENSION_NAME, ops) {
 #endif
 
   // Quantized GEMM for GPTQ.
-  ops.def("gptq_gemm", &gptq_gemm);
+  // Note: even though the C++ inferred schema is correct for this op, it seems
+  // to prevent the meta function registry.
+  ops.def(
+      "gptq_gemm(Tensor a, Tensor b_q_weight, Tensor b_gptq_qzeros, "
+      "Tensor b_gptq_scales, Tensor b_g_idx, bool use_exllama, int bit) "
+      "-> Tensor");
   ops.impl("gptq_gemm", torch::kCUDA, &gptq_gemm);
 
   // Post processing for GPTQ.
@@ -250,8 +308,8 @@ TORCH_LIBRARY_EXPAND(TORCH_EXTENSION_NAME, ops) {
 
   // Compute dynamic-per-token FP8 quantized tensor and scaling factor.
   ops.def(
-      "dynamic_per_token_scaled_fp8_quant(Tensor! out, Tensor input, Tensor! "
-      "scale, Tensor? scale_ub) -> "
+      "dynamic_per_token_scaled_fp8_quant(Tensor! out, Tensor input, "
+      "Tensor! scale, Tensor? scale_ub) -> "
       "()");
   ops.impl("dynamic_per_token_scaled_fp8_quant", torch::kCUDA,
            &dynamic_per_token_scaled_fp8_quant);
@@ -288,8 +346,8 @@ TORCH_LIBRARY_EXPAND(CONCAT(TORCH_EXTENSION_NAME, _cache_ops), cache_ops) {
 
   // Copy the cache blocks from src to dst.
   cache_ops.def(
-      "copy_blocks(Tensor[]! key_caches, Tensor[]! value_caches, Tensor "
-      "block_mapping) -> ()");
+      "copy_blocks(Tensor(a!)[] key_caches, Tensor[](b!) value_caches, "
+      "Tensor block_mapping) -> ()");
   cache_ops.impl("copy_blocks", torch::kCUDA, &copy_blocks);
 
   // Reshape the key and value tensors and cache them.
@@ -314,33 +372,37 @@ TORCH_LIBRARY_EXPAND(CONCAT(TORCH_EXTENSION_NAME, _cache_ops), cache_ops) {
 
   // Convert the key and value cache to fp8 data type.
   cache_ops.def(
-      "convert_fp8(Tensor! dst_cache, Tensor src_cache, float scale, str "
-      "kv_cache_dtype) -> ()");
+      "convert_fp8(Tensor! dst_cache, Tensor src_cache, float scale, "
+      "str kv_cache_dtype) -> ()");
   cache_ops.impl("convert_fp8", torch::kCUDA, &convert_fp8);
 }
 
 TORCH_LIBRARY_EXPAND(CONCAT(TORCH_EXTENSION_NAME, _cuda_utils), cuda_utils) {
   // Cuda utils
 
   // Gets the specified device attribute.
-  cuda_utils.def("get_device_attribute", &get_device_attribute);
-  cuda_utils.impl("get_device_attribute", torch::kCUDA, &get_device_attribute);
+  cuda_utils.def("get_device_attribute(int attribute, int device_id) -> int");
+  cuda_utils.impl("get_device_attribute", &get_device_attribute);
 
   // Gets the maximum shared memory per block device attribute.
-  cuda_utils.def("get_max_shared_memory_per_block_device_attribute",
-                 &get_max_shared_memory_per_block_device_attribute);
+  cuda_utils.def(
+      "get_max_shared_memory_per_block_device_attribute(int device_id) -> int");
   cuda_utils.impl("get_max_shared_memory_per_block_device_attribute",
-                  torch::kCUDA,
                   &get_max_shared_memory_per_block_device_attribute);
 }
 
 #ifndef USE_ROCM
 TORCH_LIBRARY_EXPAND(CONCAT(TORCH_EXTENSION_NAME, _custom_ar), custom_ar) {
   // Custom all-reduce kernels
-  custom_ar.def("init_custom_ar", &init_custom_ar);
+  custom_ar.def(
+      "init_custom_ar(Tensor meta, Tensor rank_data, "
+      "str[] handles, int[] offsets, int rank, "
+      "bool full_nvlink) -> int");
   custom_ar.impl("init_custom_ar", torch::kCUDA, &init_custom_ar);
 
-  custom_ar.def("should_custom_ar", &should_custom_ar);
+  custom_ar.def(
+      "should_custom_ar(Tensor inp, int max_size, int world_size, "
+      "bool full_nvlink) -> bool");
   custom_ar.impl("should_custom_ar", torch::kCUDA, &should_custom_ar);
 
   custom_ar.def("all_reduce_reg(int fa, Tensor inp, Tensor! out) -> ()");
@@ -352,21 +414,15 @@ TORCH_LIBRARY_EXPAND(CONCAT(TORCH_EXTENSION_NAME, _custom_ar), custom_ar) {
   custom_ar.impl("all_reduce_unreg", torch::kCUDA, &all_reduce_unreg);
 
   custom_ar.def("dispose", &dispose);
-  custom_ar.impl("dispose", torch::kCPU, &dispose);
-
   custom_ar.def("meta_size", &meta_size);
-  custom_ar.impl("meta_size", torch::kCPU, &meta_size);
 
-  custom_ar.def("register_buffer", &register_buffer);
+  custom_ar.def(
+      "register_buffer(int fa, Tensor t, str[] handles, "
+      "int[] offsets) -> ()");
   custom_ar.impl("register_buffer", torch::kCUDA, &register_buffer);
 
   custom_ar.def("get_graph_buffer_ipc_meta", &get_graph_buffer_ipc_meta);
-  custom_ar.impl("get_graph_buffer_ipc_meta", torch::kCPU,
-                 &get_graph_buffer_ipc_meta);
-
   custom_ar.def("register_graph_buffers", &register_graph_buffers);
-  custom_ar.impl("register_graph_buffers", torch::kCPU,
-                 &register_graph_buffers);
 }
 #endif