drbh commited on 17 days ago

Commit

a602080

unverified ·

0 Parent(s):

Migrated from kernels-community/yoso

Browse files

This view is limited to 50 files because it contains too many changes. See raw diff

Files changed (50) hide show

.gitattributes +107 -0
README.md +15 -0
benchmarks/benchmark.py +239 -0
build.toml +32 -0
build/torch210-cu128-x86_64-windows/__init__.py +11 -0
build/torch210-cu128-x86_64-windows/_ops.py +9 -0
build/torch210-cu128-x86_64-windows/_yoso_cuda_400d834.pyd +3 -0
build/torch210-cu128-x86_64-windows/metadata.json +14 -0
build/torch210-cu128-x86_64-windows/yoso/__init__.py +26 -0
build/torch210-cxx11-cu126-aarch64-linux/__init__.py +11 -0
build/torch210-cxx11-cu126-aarch64-linux/_ops.py +9 -0
build/torch210-cxx11-cu126-aarch64-linux/_yoso_cuda_4218a9a.abi3.so +3 -0
build/torch210-cxx11-cu126-aarch64-linux/metadata.json +12 -0
build/torch210-cxx11-cu126-aarch64-linux/yoso/__init__.py +26 -0
build/torch210-cxx11-cu126-x86_64-linux/__init__.py +11 -0
build/torch210-cxx11-cu126-x86_64-linux/_ops.py +9 -0
build/torch210-cxx11-cu126-x86_64-linux/_yoso_cuda_4218a9a.abi3.so +3 -0
build/torch210-cxx11-cu126-x86_64-linux/metadata.json +12 -0
build/torch210-cxx11-cu126-x86_64-linux/yoso/__init__.py +26 -0
build/torch210-cxx11-cu128-aarch64-linux/__init__.py +11 -0
build/torch210-cxx11-cu128-aarch64-linux/_ops.py +9 -0
build/torch210-cxx11-cu128-aarch64-linux/_yoso_cuda_4218a9a.abi3.so +3 -0
build/torch210-cxx11-cu128-aarch64-linux/metadata.json +14 -0
build/torch210-cxx11-cu128-aarch64-linux/yoso/__init__.py +26 -0
build/torch210-cxx11-cu128-x86_64-linux/__init__.py +11 -0
build/torch210-cxx11-cu128-x86_64-linux/_ops.py +9 -0
build/torch210-cxx11-cu128-x86_64-linux/_yoso_cuda_4218a9a.abi3.so +3 -0
build/torch210-cxx11-cu128-x86_64-linux/metadata.json +14 -0
build/torch210-cxx11-cu128-x86_64-linux/yoso/__init__.py +26 -0
build/torch210-cxx11-cu130-aarch64-linux/__init__.py +11 -0
build/torch210-cxx11-cu130-aarch64-linux/_ops.py +9 -0
build/torch210-cxx11-cu130-aarch64-linux/_yoso_cuda_4218a9a.abi3.so +3 -0
build/torch210-cxx11-cu130-aarch64-linux/metadata.json +14 -0
build/torch210-cxx11-cu130-aarch64-linux/yoso/__init__.py +26 -0
build/torch210-cxx11-cu130-x86_64-linux/__init__.py +11 -0
build/torch210-cxx11-cu130-x86_64-linux/_ops.py +9 -0
build/torch210-cxx11-cu130-x86_64-linux/_yoso_cuda_4218a9a.abi3.so +3 -0
build/torch210-cxx11-cu130-x86_64-linux/metadata.json +14 -0
build/torch210-cxx11-cu130-x86_64-linux/yoso/__init__.py +26 -0
build/torch211-cxx11-cu126-aarch64-linux/__init__.py +11 -0
build/torch211-cxx11-cu126-aarch64-linux/_ops.py +9 -0
build/torch211-cxx11-cu126-aarch64-linux/_yoso_cuda_4218a9a.abi3.so +3 -0
build/torch211-cxx11-cu126-aarch64-linux/metadata.json +12 -0
build/torch211-cxx11-cu126-aarch64-linux/yoso/__init__.py +26 -0
build/torch211-cxx11-cu126-x86_64-linux/__init__.py +11 -0
build/torch211-cxx11-cu126-x86_64-linux/_ops.py +9 -0
build/torch211-cxx11-cu126-x86_64-linux/_yoso_cuda_4218a9a.abi3.so +3 -0
build/torch211-cxx11-cu126-x86_64-linux/metadata.json +12 -0
build/torch211-cxx11-cu126-x86_64-linux/yoso/__init__.py +26 -0
build/torch211-cxx11-cu128-aarch64-linux/__init__.py +11 -0

.gitattributes ADDED Viewed

	@@ -0,0 +1,107 @@

+*.7z filter=lfs diff=lfs merge=lfs -text
+*.arrow filter=lfs diff=lfs merge=lfs -text
+*.bin filter=lfs diff=lfs merge=lfs -text
+*.bz2 filter=lfs diff=lfs merge=lfs -text
+*.ckpt filter=lfs diff=lfs merge=lfs -text
+*.ftz filter=lfs diff=lfs merge=lfs -text
+*.gz filter=lfs diff=lfs merge=lfs -text
+*.h5 filter=lfs diff=lfs merge=lfs -text
+*.joblib filter=lfs diff=lfs merge=lfs -text
+*.lfs.* filter=lfs diff=lfs merge=lfs -text
+*.mlmodel filter=lfs diff=lfs merge=lfs -text
+*.model filter=lfs diff=lfs merge=lfs -text
+*.msgpack filter=lfs diff=lfs merge=lfs -text
+*.npy filter=lfs diff=lfs merge=lfs -text
+*.npz filter=lfs diff=lfs merge=lfs -text
+*.onnx filter=lfs diff=lfs merge=lfs -text
+*.ot filter=lfs diff=lfs merge=lfs -text
+*.parquet filter=lfs diff=lfs merge=lfs -text
+*.pb filter=lfs diff=lfs merge=lfs -text
+*.pickle filter=lfs diff=lfs merge=lfs -text
+*.pkl filter=lfs diff=lfs merge=lfs -text
+*.pt filter=lfs diff=lfs merge=lfs -text
+*.pth filter=lfs diff=lfs merge=lfs -text
+*.rar filter=lfs diff=lfs merge=lfs -text
+*.safetensors filter=lfs diff=lfs merge=lfs -text
+saved_model/**/* filter=lfs diff=lfs merge=lfs -text
+*.tar.* filter=lfs diff=lfs merge=lfs -text
+*.tar filter=lfs diff=lfs merge=lfs -text
+*.tflite filter=lfs diff=lfs merge=lfs -text
+*.tgz filter=lfs diff=lfs merge=lfs -text
+*.wasm filter=lfs diff=lfs merge=lfs -text
+*.xz filter=lfs diff=lfs merge=lfs -text
+*.zip filter=lfs diff=lfs merge=lfs -text
+*.zst filter=lfs diff=lfs merge=lfs -text
+*tfevents* filter=lfs diff=lfs merge=lfs -text
+build/torch27-cxx11-cu118-x86_64-linux/yoso/_yoso_3173620_dirty.abi3.so filter=lfs diff=lfs merge=lfs -text
+build/torch27-cxx11-cu126-x86_64-linux/yoso/_yoso_3173620_dirty.abi3.so filter=lfs diff=lfs merge=lfs -text
+build/torch27-cxx11-cu128-x86_64-linux/yoso/_yoso_3173620_dirty.abi3.so filter=lfs diff=lfs merge=lfs -text
+build/torch28-cxx11-cu126-x86_64-linux/yoso/_yoso_3173620_dirty.abi3.so filter=lfs diff=lfs merge=lfs -text
+build/torch28-cxx11-cu128-x86_64-linux/yoso/_yoso_3173620_dirty.abi3.so filter=lfs diff=lfs merge=lfs -text
+build/torch28-cxx11-cu129-x86_64-linux/yoso/_yoso_3173620_dirty.abi3.so filter=lfs diff=lfs merge=lfs -text
+build/torch29-cxx11-cu126-x86_64-linux/yoso/_yoso_3173620_dirty.abi3.so filter=lfs diff=lfs merge=lfs -text
+build/torch29-cxx11-cu128-x86_64-linux/yoso/_yoso_3173620_dirty.abi3.so filter=lfs diff=lfs merge=lfs -text
+build/torch29-cxx11-cu130-x86_64-linux/yoso/_yoso_3173620_dirty.abi3.so filter=lfs diff=lfs merge=lfs -text
+build/torch210-cxx11-cu126-x86_64-linux/_yoso_ea085fb.abi3.so filter=lfs diff=lfs merge=lfs -text
+build/torch210-cxx11-cu128-x86_64-linux/_yoso_ea085fb.abi3.so filter=lfs diff=lfs merge=lfs -text
+build/torch210-cxx11-cu130-x86_64-linux/_yoso_ea085fb.abi3.so filter=lfs diff=lfs merge=lfs -text
+build/torch28-cxx11-cu126-x86_64-linux/_yoso_ea085fb.abi3.so filter=lfs diff=lfs merge=lfs -text
+build/torch28-cxx11-cu128-x86_64-linux/_yoso_ea085fb.abi3.so filter=lfs diff=lfs merge=lfs -text
+build/torch28-cxx11-cu129-x86_64-linux/_yoso_ea085fb.abi3.so filter=lfs diff=lfs merge=lfs -text
+build/torch29-cxx11-cu126-x86_64-linux/_yoso_ea085fb.abi3.so filter=lfs diff=lfs merge=lfs -text
+build/torch29-cxx11-cu128-x86_64-linux/_yoso_ea085fb.abi3.so filter=lfs diff=lfs merge=lfs -text
+build/torch29-cxx11-cu130-x86_64-linux/_yoso_ea085fb.abi3.so filter=lfs diff=lfs merge=lfs -text
+build/torch210-cxx11-cu126-x86_64-linux/_yoso_fc1f972.abi3.so filter=lfs diff=lfs merge=lfs -text
+build/torch210-cxx11-cu128-x86_64-linux/_yoso_fc1f972.abi3.so filter=lfs diff=lfs merge=lfs -text
+build/torch210-cxx11-cu130-x86_64-linux/_yoso_fc1f972.abi3.so filter=lfs diff=lfs merge=lfs -text
+build/torch28-cxx11-cu126-x86_64-linux/_yoso_fc1f972.abi3.so filter=lfs diff=lfs merge=lfs -text
+build/torch28-cxx11-cu128-x86_64-linux/_yoso_fc1f972.abi3.so filter=lfs diff=lfs merge=lfs -text
+build/torch28-cxx11-cu129-x86_64-linux/_yoso_fc1f972.abi3.so filter=lfs diff=lfs merge=lfs -text
+build/torch29-cxx11-cu126-x86_64-linux/_yoso_fc1f972.abi3.so filter=lfs diff=lfs merge=lfs -text
+build/torch29-cxx11-cu128-x86_64-linux/_yoso_fc1f972.abi3.so filter=lfs diff=lfs merge=lfs -text
+build/torch29-cxx11-cu130-x86_64-linux/_yoso_fc1f972.abi3.so filter=lfs diff=lfs merge=lfs -text
+build/torch210-cxx11-cu126-x86_64-linux/_yoso_c11288f.abi3.so filter=lfs diff=lfs merge=lfs -text
+build/torch210-cxx11-cu128-x86_64-linux/_yoso_c11288f.abi3.so filter=lfs diff=lfs merge=lfs -text
+build/torch210-cxx11-cu130-x86_64-linux/_yoso_c11288f.abi3.so filter=lfs diff=lfs merge=lfs -text
+build/torch28-cxx11-cu126-x86_64-linux/_yoso_c11288f.abi3.so filter=lfs diff=lfs merge=lfs -text
+build/torch28-cxx11-cu128-x86_64-linux/_yoso_c11288f.abi3.so filter=lfs diff=lfs merge=lfs -text
+build/torch28-cxx11-cu129-x86_64-linux/_yoso_c11288f.abi3.so filter=lfs diff=lfs merge=lfs -text
+build/torch29-cxx11-cu126-x86_64-linux/_yoso_c11288f.abi3.so filter=lfs diff=lfs merge=lfs -text
+build/torch29-cxx11-cu128-x86_64-linux/_yoso_c11288f.abi3.so filter=lfs diff=lfs merge=lfs -text
+build/torch29-cxx11-cu130-x86_64-linux/_yoso_c11288f.abi3.so filter=lfs diff=lfs merge=lfs -text
+build/torch210-cxx11-cu126-x86_64-linux/_yoso_dd38bd3.abi3.so filter=lfs diff=lfs merge=lfs -text
+build/torch210-cxx11-cu128-x86_64-linux/_yoso_dd38bd3.abi3.so filter=lfs diff=lfs merge=lfs -text
+build/torch210-cxx11-cu130-x86_64-linux/_yoso_dd38bd3.abi3.so filter=lfs diff=lfs merge=lfs -text
+build/torch28-cxx11-cu126-x86_64-linux/_yoso_dd38bd3.abi3.so filter=lfs diff=lfs merge=lfs -text
+build/torch28-cxx11-cu128-x86_64-linux/_yoso_dd38bd3.abi3.so filter=lfs diff=lfs merge=lfs -text
+build/torch28-cxx11-cu129-x86_64-linux/_yoso_dd38bd3.abi3.so filter=lfs diff=lfs merge=lfs -text
+build/torch29-cxx11-cu126-x86_64-linux/_yoso_dd38bd3.abi3.so filter=lfs diff=lfs merge=lfs -text
+build/torch29-cxx11-cu128-x86_64-linux/_yoso_dd38bd3.abi3.so filter=lfs diff=lfs merge=lfs -text
+build/torch29-cxx11-cu130-x86_64-linux/_yoso_dd38bd3.abi3.so filter=lfs diff=lfs merge=lfs -text
+build/torch210-cxx11-cu126-x86_64-linux/_yoso_cuda_c42e4a3.abi3.so filter=lfs diff=lfs merge=lfs -text
+build/torch210-cxx11-cu128-x86_64-linux/_yoso_cuda_c42e4a3.abi3.so filter=lfs diff=lfs merge=lfs -text
+build/torch210-cxx11-cu130-x86_64-linux/_yoso_cuda_c42e4a3.abi3.so filter=lfs diff=lfs merge=lfs -text
+build/torch29-cxx11-cu126-x86_64-linux/_yoso_cuda_c42e4a3.abi3.so filter=lfs diff=lfs merge=lfs -text
+build/torch29-cxx11-cu128-x86_64-linux/_yoso_cuda_c42e4a3.abi3.so filter=lfs diff=lfs merge=lfs -text
+build/torch29-cxx11-cu130-x86_64-linux/_yoso_cuda_c42e4a3.abi3.so filter=lfs diff=lfs merge=lfs -text
+build/torch210-cxx11-cu126-aarch64-linux/_yoso_cuda_c42e4a3.abi3.so filter=lfs diff=lfs merge=lfs -text
+build/torch210-cxx11-cu128-aarch64-linux/_yoso_cuda_c42e4a3.abi3.so filter=lfs diff=lfs merge=lfs -text
+build/torch210-cxx11-cu130-aarch64-linux/_yoso_cuda_c42e4a3.abi3.so filter=lfs diff=lfs merge=lfs -text
+build/torch29-cxx11-cu126-aarch64-linux/_yoso_cuda_c42e4a3.abi3.so filter=lfs diff=lfs merge=lfs -text
+build/torch29-cxx11-cu128-aarch64-linux/_yoso_cuda_c42e4a3.abi3.so filter=lfs diff=lfs merge=lfs -text
+build/torch29-cxx11-cu130-aarch64-linux/_yoso_cuda_c42e4a3.abi3.so filter=lfs diff=lfs merge=lfs -text
+build/torch210-cu128-x86_64-windows/_yoso_cuda_400d834.pyd filter=lfs diff=lfs merge=lfs -text
+build/torch210-cxx11-cu126-aarch64-linux/_yoso_cuda_4218a9a.abi3.so filter=lfs diff=lfs merge=lfs -text
+build/torch210-cxx11-cu128-aarch64-linux/_yoso_cuda_4218a9a.abi3.so filter=lfs diff=lfs merge=lfs -text
+build/torch210-cxx11-cu130-aarch64-linux/_yoso_cuda_4218a9a.abi3.so filter=lfs diff=lfs merge=lfs -text
+build/torch211-cxx11-cu126-aarch64-linux/_yoso_cuda_4218a9a.abi3.so filter=lfs diff=lfs merge=lfs -text
+build/torch211-cxx11-cu128-aarch64-linux/_yoso_cuda_4218a9a.abi3.so filter=lfs diff=lfs merge=lfs -text
+build/torch211-cxx11-cu130-aarch64-linux/_yoso_cuda_4218a9a.abi3.so filter=lfs diff=lfs merge=lfs -text
+build/torch29-cxx11-cu129-aarch64-linux/_yoso_cuda_4218a9a.abi3.so filter=lfs diff=lfs merge=lfs -text
+build/torch210-cxx11-cu126-x86_64-linux/_yoso_cuda_4218a9a.abi3.so filter=lfs diff=lfs merge=lfs -text
+build/torch210-cxx11-cu128-x86_64-linux/_yoso_cuda_4218a9a.abi3.so filter=lfs diff=lfs merge=lfs -text
+build/torch210-cxx11-cu130-x86_64-linux/_yoso_cuda_4218a9a.abi3.so filter=lfs diff=lfs merge=lfs -text
+build/torch211-cxx11-cu126-x86_64-linux/_yoso_cuda_4218a9a.abi3.so filter=lfs diff=lfs merge=lfs -text
+build/torch211-cxx11-cu128-x86_64-linux/_yoso_cuda_4218a9a.abi3.so filter=lfs diff=lfs merge=lfs -text
+build/torch211-cxx11-cu130-x86_64-linux/_yoso_cuda_4218a9a.abi3.so filter=lfs diff=lfs merge=lfs -text
+build/torch29-cxx11-cu129-x86_64-linux/_yoso_cuda_4218a9a.abi3.so filter=lfs diff=lfs merge=lfs -text

README.md ADDED Viewed

	@@ -0,0 +1,15 @@

+---
+tags:
+- kernels
+- cuda
+---
+### Performance
+<img class="dark:hidden border border-gray-200 dark:border-gray-700 rounded-lg" src="media/benches_light_animation.svg" />
+<img class="hidden dark:block border border-gray-200 dark:border-gray-700 rounded-lg" src="media/benches_dark_animation.svg" />
+<img class="dark:hidden border border-gray-200 dark:border-gray-700 rounded-lg" src="media/benches_light_latency.svg" />
+<img class="hidden dark:block border border-gray-200 dark:border-gray-700 rounded-lg" src="media/benches_dark_latency.svg" />
+<img class="dark:hidden border border-gray-200 dark:border-gray-700 rounded-lg" src="media/benches_light_throughput.svg" />
+<img class="hidden dark:block border border-gray-200 dark:border-gray-700 rounded-lg" src="media/benches_dark_throughput.svg" />

benchmarks/benchmark.py ADDED Viewed

	@@ -0,0 +1,239 @@

+import torch
+from kernels.benchmark import Benchmark
+def lsh_weighted_cumulation_reference(
+    query_mask: torch.Tensor,
+    query_hash_code: torch.Tensor,
+    query_weight: torch.Tensor,
+    key_mask: torch.Tensor,
+    key_hash_code: torch.Tensor,
+    key_weight: torch.Tensor,
+    value: torch.Tensor,
+    hashtable_capacity: int,
+) -> torch.Tensor:
+    batch_size, num_query, num_hash_f = query_hash_code.shape
+    _, num_key, value_dim = value.shape
+    weight_dim = query_weight.shape[2]
+    device = value.device
+    dtype = value.dtype
+    output = torch.zeros(batch_size, num_query, value_dim, device=device, dtype=dtype)
+    for b in range(batch_size):
+        for weight_idx in range(weight_dim):
+            # Build hashtables for all hash functions
+            hashtables = torch.zeros(
+                num_hash_f, hashtable_capacity, value_dim, device=device, dtype=dtype
+            )
+            k_mask = key_mask[b, :].float()  # [num_key]
+            k_weight_val = key_weight[b, :, weight_idx]  # [num_key]
+            for h in range(num_hash_f):
+                k_hash = key_hash_code[b, :, h].long()  # [num_key]
+                # Weighted values: [num_key, value_dim]
+                weighted_values = (
+                    k_mask.unsqueeze(-1) * k_weight_val.unsqueeze(-1) * value[b]
+                )
+                k_hash_expanded = k_hash.unsqueeze(-1).expand(-1, value_dim)
+                hashtables[h].scatter_add_(0, k_hash_expanded, weighted_values)
+            # Query: sum over all hash functions
+            q_mask = query_mask[b, :].float()  # [num_query]
+            q_weight_val = query_weight[b, :, weight_idx]  # [num_query]
+            sum_val = torch.zeros(num_query, value_dim, device=device, dtype=dtype)
+            for h in range(num_hash_f):
+                q_hash = query_hash_code[b, :, h].long()  # [num_query]
+                gathered = hashtables[h][q_hash]  # [num_query, value_dim]
+                sum_val += gathered
+            # Apply query weight and divide by num_hash_f
+            output[b] += (
+                q_mask.unsqueeze(-1) * q_weight_val.unsqueeze(-1) * sum_val / num_hash_f
+            )
+    return output
+class YosoBenchmark(Benchmark):
+    seed: int = 42
+    def setup(self):
+        batch_size = 2
+        num_query = 128
+        num_key = 128
+        dim = 64
+        self.num_hash_f = 32
+        self.hash_code_len = 9
+        self.weight_dim = self.num_hash_f
+        self.value_dim = dim
+        self.hashtable_capacity = 1 << self.hash_code_len
+        self.query_mask = torch.ones(
+            batch_size, num_query, device=self.device, dtype=torch.int32
+        )
+        self.query_vector = torch.randn(
+            batch_size, num_query, dim, device=self.device, dtype=torch.float32
+        )
+        self.key_mask = torch.ones(
+            batch_size, num_key, device=self.device, dtype=torch.int32
+        )
+        self.key_vector = torch.randn(
+            batch_size, num_key, dim, device=self.device, dtype=torch.float32
+        )
+        self.value = torch.randn(
+            batch_size, num_key, self.value_dim, device=self.device, dtype=torch.float32
+        )
+        self.query_weight = torch.randn(
+            batch_size,
+            num_query,
+            self.weight_dim,
+            device=self.device,
+            dtype=torch.float32,
+        )
+        self.key_weight = torch.randn(
+            batch_size,
+            num_key,
+            self.weight_dim,
+            device=self.device,
+            dtype=torch.float32,
+        )
+        # Pre-compute hash codes for cumulation benchmarks
+        hash_result = self.kernel.fast_hash(
+            self.query_mask,
+            self.query_vector,
+            self.key_mask,
+            self.key_vector,
+            self.num_hash_f,
+            self.hash_code_len,
+            True,
+            1,
+        )
+        self.query_hash_code = hash_result[0]
+        self.key_hash_code = hash_result[1]
+        self.out = torch.empty(
+            batch_size,
+            num_query,
+            self.value_dim,
+            device=self.device,
+            dtype=torch.float32,
+        )
+    def benchmark_base(self):
+        self.out = self.kernel.lsh_weighted_cumulation(
+            self.query_mask,
+            self.query_hash_code,
+            self.query_weight,
+            self.key_mask,
+            self.key_hash_code,
+            self.key_weight,
+            self.value,
+            self.hashtable_capacity,
+            True,
+            1,
+        )
+    def verify_base(self) -> torch.Tensor:
+        return lsh_weighted_cumulation_reference(
+            self.query_mask,
+            self.query_hash_code,
+            self.query_weight,
+            self.key_mask,
+            self.key_hash_code,
+            self.key_weight,
+            self.value,
+            self.hashtable_capacity,
+        )
+    def setup_large(self):
+        batch_size = 4
+        num_query = 512
+        num_key = 512
+        dim = 128
+        self.num_hash_f = 32
+        self.hash_code_len = 9
+        self.weight_dim = self.num_hash_f
+        self.value_dim = dim
+        self.hashtable_capacity = 1 << self.hash_code_len
+        self.query_mask = torch.ones(
+            batch_size, num_query, device=self.device, dtype=torch.int32
+        )
+        self.query_vector = torch.randn(
+            batch_size, num_query, dim, device=self.device, dtype=torch.float32
+        )
+        self.key_mask = torch.ones(
+            batch_size, num_key, device=self.device, dtype=torch.int32
+        )
+        self.key_vector = torch.randn(
+            batch_size, num_key, dim, device=self.device, dtype=torch.float32
+        )
+        self.value = torch.randn(
+            batch_size, num_key, self.value_dim, device=self.device, dtype=torch.float32
+        )
+        self.query_weight = torch.randn(
+            batch_size,
+            num_query,
+            self.weight_dim,
+            device=self.device,
+            dtype=torch.float32,
+        )
+        self.key_weight = torch.randn(
+            batch_size,
+            num_key,
+            self.weight_dim,
+            device=self.device,
+            dtype=torch.float32,
+        )
+        hash_result = self.kernel.fast_hash(
+            self.query_mask,
+            self.query_vector,
+            self.key_mask,
+            self.key_vector,
+            self.num_hash_f,
+            self.hash_code_len,
+            True,
+            1,
+        )
+        self.query_hash_code = hash_result[0]
+        self.key_hash_code = hash_result[1]
+        self.out = torch.empty(
+            batch_size,
+            num_query,
+            self.value_dim,
+            device=self.device,
+            dtype=torch.float32,
+        )
+    def benchmark_large(self):
+        self.out = self.kernel.lsh_weighted_cumulation(
+            self.query_mask,
+            self.query_hash_code,
+            self.query_weight,
+            self.key_mask,
+            self.key_hash_code,
+            self.key_weight,
+            self.value,
+            self.hashtable_capacity,
+            True,
+            1,
+        )
+    def verify_large(self) -> torch.Tensor:
+        return lsh_weighted_cumulation_reference(
+            self.query_mask,
+            self.query_hash_code,
+            self.query_weight,
+            self.key_mask,
+            self.key_hash_code,
+            self.key_weight,
+            self.value,
+            self.hashtable_capacity,
+        )

build.toml ADDED Viewed

	@@ -0,0 +1,32 @@

+[general]
+name = "yoso"
+universal = false
+[torch]
+src = [
+  "torch-ext/torch_binding.cpp",
+  "torch-ext/torch_binding.h",
+]
+[kernel.yoso]
+depends = ["torch"]
+backend = "cuda"
+cuda-capabilities = [
+    "8.0",
+    "8.9",
+    "9.0",
+    "10.0",
+    "12.0",
+]
+include = ["."]
+src = [
+    "yoso/fast_lsh_cumulation_cuda.cu",
+    "yoso/fast_lsh_cumulation_cuda.h",
+    "yoso/fast_lsh_cumulation_torch.cpp",
+    "yoso/fast_lsh_cumulation.h",
+    "yoso/fast_lsh_cumulation.cu",
+    "yoso/common_cuda.h",
+    "yoso/common.h",
+    "yoso/common_cuda_device.h",
+]

build/torch210-cu128-x86_64-windows/__init__.py ADDED Viewed

	@@ -0,0 +1,11 @@

+from ._ops import ops
+fast_hash = ops.fast_hash
+lsh_cumulation = ops.lsh_cumulation
+lsh_weighted_cumulation = ops.lsh_weighted_cumulation
+__all__ = [
+    "fast_hash",
+    "lsh_cumulation",
+    "lsh_weighted_cumulation",
+]

build/torch210-cu128-x86_64-windows/_ops.py ADDED Viewed

	@@ -0,0 +1,9 @@

+import torch
+from . import _yoso_cuda_400d834
+ops = torch.ops._yoso_cuda_400d834
+def add_op_namespace_prefix(op_name: str):
+    """
+    Prefix op by namespace.
+    """
+    return f"_yoso_cuda_400d834::{op_name}"

build/torch210-cu128-x86_64-windows/_yoso_cuda_400d834.pyd ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:698d4c7581229298f15ebcdcbf056ef8331d7e50fc3d510a3b1a9a6a9fdf22ca
+size 1112576

build/torch210-cu128-x86_64-windows/metadata.json ADDED Viewed

	@@ -0,0 +1,14 @@

+{
+  "version": 1,
+  "python-depends": [],
+  "backend": {
+    "type": "cuda",
+    "archs": [
+      "10.0",
+      "12.0",
+      "8.0",
+      "8.9",
+      "9.0"
+    ]
+  }
+}

build/torch210-cu128-x86_64-windows/yoso/__init__.py ADDED Viewed

	@@ -0,0 +1,26 @@

+import ctypes
+import sys
+import importlib
+from pathlib import Path
+from types import ModuleType
+def _import_from_path(file_path: Path) -> ModuleType:
+    # We cannot use the module name as-is, after adding it to `sys.modules`,
+    # it would also be used for other imports. So, we make a module name that
+    # depends on the path for it to be unique using the hex-encoded hash of
+    # the path.
+    path_hash = "{:x}".format(ctypes.c_size_t(hash(file_path.absolute())).value)
+    module_name = path_hash
+    spec = importlib.util.spec_from_file_location(module_name, file_path)
+    if spec is None:
+        raise ImportError(f"Cannot load spec for {module_name} from {file_path}")
+    module = importlib.util.module_from_spec(spec)
+    if module is None:
+        raise ImportError(f"Cannot load module {module_name} from spec")
+    sys.modules[module_name] = module
+    spec.loader.exec_module(module)  # type: ignore
+    return module
+globals().update(vars(_import_from_path(Path(__file__).parent.parent / "__init__.py")))

build/torch210-cxx11-cu126-aarch64-linux/__init__.py ADDED Viewed

	@@ -0,0 +1,11 @@

+from ._ops import ops
+fast_hash = ops.fast_hash
+lsh_cumulation = ops.lsh_cumulation
+lsh_weighted_cumulation = ops.lsh_weighted_cumulation
+__all__ = [
+    "fast_hash",
+    "lsh_cumulation",
+    "lsh_weighted_cumulation",
+]

build/torch210-cxx11-cu126-aarch64-linux/_ops.py ADDED Viewed

	@@ -0,0 +1,9 @@

+import torch
+from . import _yoso_cuda_4218a9a
+ops = torch.ops._yoso_cuda_4218a9a
+def add_op_namespace_prefix(op_name: str):
+    """
+    Prefix op by namespace.
+    """
+    return f"_yoso_cuda_4218a9a::{op_name}"

build/torch210-cxx11-cu126-aarch64-linux/_yoso_cuda_4218a9a.abi3.so ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:2bb09bdc0593bb01cea62105e034943212b28cee89082a44c7412adc87943215
+size 2569064

build/torch210-cxx11-cu126-aarch64-linux/metadata.json ADDED Viewed

	@@ -0,0 +1,12 @@

+{
+  "version": 1,
+  "python-depends": [],
+  "backend": {
+    "type": "cuda",
+    "archs": [
+      "8.0",
+      "8.9",
+      "9.0"
+    ]
+  }
+}

build/torch210-cxx11-cu126-aarch64-linux/yoso/__init__.py ADDED Viewed

	@@ -0,0 +1,26 @@

+import ctypes
+import importlib.util
+import sys
+from pathlib import Path
+from types import ModuleType
+def _import_from_path(file_path: Path) -> ModuleType:
+    # We cannot use the module name as-is, after adding it to `sys.modules`,
+    # it would also be used for other imports. So, we make a module name that
+    # depends on the path for it to be unique using the hex-encoded hash of
+    # the path.
+    path_hash = "{:x}".format(ctypes.c_size_t(hash(file_path.absolute())).value)
+    module_name = path_hash
+    spec = importlib.util.spec_from_file_location(module_name, file_path)
+    if spec is None:
+        raise ImportError(f"Cannot load spec for {module_name} from {file_path}")
+    module = importlib.util.module_from_spec(spec)
+    if module is None:
+        raise ImportError(f"Cannot load module {module_name} from spec")
+    sys.modules[module_name] = module
+    spec.loader.exec_module(module)  # type: ignore
+    return module
+globals().update(vars(_import_from_path(Path(__file__).parent.parent / "__init__.py")))

build/torch210-cxx11-cu126-x86_64-linux/__init__.py ADDED Viewed

	@@ -0,0 +1,11 @@

+from ._ops import ops
+fast_hash = ops.fast_hash
+lsh_cumulation = ops.lsh_cumulation
+lsh_weighted_cumulation = ops.lsh_weighted_cumulation
+__all__ = [
+    "fast_hash",
+    "lsh_cumulation",
+    "lsh_weighted_cumulation",
+]

build/torch210-cxx11-cu126-x86_64-linux/_ops.py ADDED Viewed

	@@ -0,0 +1,9 @@

+import torch
+from . import _yoso_cuda_4218a9a
+ops = torch.ops._yoso_cuda_4218a9a
+def add_op_namespace_prefix(op_name: str):
+    """
+    Prefix op by namespace.
+    """
+    return f"_yoso_cuda_4218a9a::{op_name}"

build/torch210-cxx11-cu126-x86_64-linux/_yoso_cuda_4218a9a.abi3.so ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:02ce77a1a0e640f167c2542c3902706d15891b05c813b38d4435cf12088e4771
+size 2498248

build/torch210-cxx11-cu126-x86_64-linux/metadata.json ADDED Viewed

	@@ -0,0 +1,12 @@

+{
+  "version": 1,
+  "python-depends": [],
+  "backend": {
+    "type": "cuda",
+    "archs": [
+      "8.0",
+      "8.9",
+      "9.0"
+    ]
+  }
+}

build/torch210-cxx11-cu126-x86_64-linux/yoso/__init__.py ADDED Viewed

	@@ -0,0 +1,26 @@

+import ctypes
+import importlib.util
+import sys
+from pathlib import Path
+from types import ModuleType
+def _import_from_path(file_path: Path) -> ModuleType:
+    # We cannot use the module name as-is, after adding it to `sys.modules`,
+    # it would also be used for other imports. So, we make a module name that
+    # depends on the path for it to be unique using the hex-encoded hash of
+    # the path.
+    path_hash = "{:x}".format(ctypes.c_size_t(hash(file_path.absolute())).value)
+    module_name = path_hash
+    spec = importlib.util.spec_from_file_location(module_name, file_path)
+    if spec is None:
+        raise ImportError(f"Cannot load spec for {module_name} from {file_path}")
+    module = importlib.util.module_from_spec(spec)
+    if module is None:
+        raise ImportError(f"Cannot load module {module_name} from spec")
+    sys.modules[module_name] = module
+    spec.loader.exec_module(module)  # type: ignore
+    return module
+globals().update(vars(_import_from_path(Path(__file__).parent.parent / "__init__.py")))

build/torch210-cxx11-cu128-aarch64-linux/__init__.py ADDED Viewed

	@@ -0,0 +1,11 @@

+from ._ops import ops
+fast_hash = ops.fast_hash
+lsh_cumulation = ops.lsh_cumulation
+lsh_weighted_cumulation = ops.lsh_weighted_cumulation
+__all__ = [
+    "fast_hash",
+    "lsh_cumulation",
+    "lsh_weighted_cumulation",
+]

build/torch210-cxx11-cu128-aarch64-linux/_ops.py ADDED Viewed

	@@ -0,0 +1,9 @@

+import torch
+from . import _yoso_cuda_4218a9a
+ops = torch.ops._yoso_cuda_4218a9a
+def add_op_namespace_prefix(op_name: str):
+    """
+    Prefix op by namespace.
+    """
+    return f"_yoso_cuda_4218a9a::{op_name}"

build/torch210-cxx11-cu128-aarch64-linux/_yoso_cuda_4218a9a.abi3.so ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:91b0d7b3accfb6b76ab9dfbe11942158cf8648fd3ad6f7915b15eb3dc2add36f
+size 3159096

build/torch210-cxx11-cu128-aarch64-linux/metadata.json ADDED Viewed

	@@ -0,0 +1,14 @@

+{
+  "version": 1,
+  "python-depends": [],
+  "backend": {
+    "type": "cuda",
+    "archs": [
+      "10.0",
+      "12.0",
+      "8.0",
+      "8.9",
+      "9.0"
+    ]
+  }
+}

build/torch210-cxx11-cu128-aarch64-linux/yoso/__init__.py ADDED Viewed

	@@ -0,0 +1,26 @@

+import ctypes
+import importlib.util
+import sys
+from pathlib import Path
+from types import ModuleType
+def _import_from_path(file_path: Path) -> ModuleType:
+    # We cannot use the module name as-is, after adding it to `sys.modules`,
+    # it would also be used for other imports. So, we make a module name that
+    # depends on the path for it to be unique using the hex-encoded hash of
+    # the path.
+    path_hash = "{:x}".format(ctypes.c_size_t(hash(file_path.absolute())).value)
+    module_name = path_hash
+    spec = importlib.util.spec_from_file_location(module_name, file_path)
+    if spec is None:
+        raise ImportError(f"Cannot load spec for {module_name} from {file_path}")
+    module = importlib.util.module_from_spec(spec)
+    if module is None:
+        raise ImportError(f"Cannot load module {module_name} from spec")
+    sys.modules[module_name] = module
+    spec.loader.exec_module(module)  # type: ignore
+    return module
+globals().update(vars(_import_from_path(Path(__file__).parent.parent / "__init__.py")))

build/torch210-cxx11-cu128-x86_64-linux/__init__.py ADDED Viewed

	@@ -0,0 +1,11 @@

+from ._ops import ops
+fast_hash = ops.fast_hash
+lsh_cumulation = ops.lsh_cumulation
+lsh_weighted_cumulation = ops.lsh_weighted_cumulation
+__all__ = [
+    "fast_hash",
+    "lsh_cumulation",
+    "lsh_weighted_cumulation",
+]

build/torch210-cxx11-cu128-x86_64-linux/_ops.py ADDED Viewed

	@@ -0,0 +1,9 @@

+import torch
+from . import _yoso_cuda_4218a9a
+ops = torch.ops._yoso_cuda_4218a9a
+def add_op_namespace_prefix(op_name: str):
+    """
+    Prefix op by namespace.
+    """
+    return f"_yoso_cuda_4218a9a::{op_name}"

build/torch210-cxx11-cu128-x86_64-linux/_yoso_cuda_4218a9a.abi3.so ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:65e95d52be57053b7f09b0880484c5debf4e542ac257cb3683105305df39d839
+size 3040824

build/torch210-cxx11-cu128-x86_64-linux/metadata.json ADDED Viewed

	@@ -0,0 +1,14 @@

+{
+  "version": 1,
+  "python-depends": [],
+  "backend": {
+    "type": "cuda",
+    "archs": [
+      "10.0",
+      "12.0",
+      "8.0",
+      "8.9",
+      "9.0"
+    ]
+  }
+}

build/torch210-cxx11-cu128-x86_64-linux/yoso/__init__.py ADDED Viewed

	@@ -0,0 +1,26 @@

+import ctypes
+import importlib.util
+import sys
+from pathlib import Path
+from types import ModuleType
+def _import_from_path(file_path: Path) -> ModuleType:
+    # We cannot use the module name as-is, after adding it to `sys.modules`,
+    # it would also be used for other imports. So, we make a module name that
+    # depends on the path for it to be unique using the hex-encoded hash of
+    # the path.
+    path_hash = "{:x}".format(ctypes.c_size_t(hash(file_path.absolute())).value)
+    module_name = path_hash
+    spec = importlib.util.spec_from_file_location(module_name, file_path)
+    if spec is None:
+        raise ImportError(f"Cannot load spec for {module_name} from {file_path}")
+    module = importlib.util.module_from_spec(spec)
+    if module is None:
+        raise ImportError(f"Cannot load module {module_name} from spec")
+    sys.modules[module_name] = module
+    spec.loader.exec_module(module)  # type: ignore
+    return module
+globals().update(vars(_import_from_path(Path(__file__).parent.parent / "__init__.py")))

build/torch210-cxx11-cu130-aarch64-linux/__init__.py ADDED Viewed

	@@ -0,0 +1,11 @@

+from ._ops import ops
+fast_hash = ops.fast_hash
+lsh_cumulation = ops.lsh_cumulation
+lsh_weighted_cumulation = ops.lsh_weighted_cumulation
+__all__ = [
+    "fast_hash",
+    "lsh_cumulation",
+    "lsh_weighted_cumulation",
+]

build/torch210-cxx11-cu130-aarch64-linux/_ops.py ADDED Viewed

	@@ -0,0 +1,9 @@

+import torch
+from . import _yoso_cuda_4218a9a
+ops = torch.ops._yoso_cuda_4218a9a
+def add_op_namespace_prefix(op_name: str):
+    """
+    Prefix op by namespace.
+    """
+    return f"_yoso_cuda_4218a9a::{op_name}"

build/torch210-cxx11-cu130-aarch64-linux/_yoso_cuda_4218a9a.abi3.so ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:18a1cbf153beac65c36a3fcf2fdb16d6c22e8e962ebb3e4b91e7c3a927c98315
+size 3230640

build/torch210-cxx11-cu130-aarch64-linux/metadata.json ADDED Viewed

	@@ -0,0 +1,14 @@

+{
+  "version": 1,
+  "python-depends": [],
+  "backend": {
+    "type": "cuda",
+    "archs": [
+      "10.0",
+      "12.0",
+      "8.0",
+      "8.9",
+      "9.0"
+    ]
+  }
+}

build/torch210-cxx11-cu130-aarch64-linux/yoso/__init__.py ADDED Viewed

	@@ -0,0 +1,26 @@

+import ctypes
+import importlib.util
+import sys
+from pathlib import Path
+from types import ModuleType
+def _import_from_path(file_path: Path) -> ModuleType:
+    # We cannot use the module name as-is, after adding it to `sys.modules`,
+    # it would also be used for other imports. So, we make a module name that
+    # depends on the path for it to be unique using the hex-encoded hash of
+    # the path.
+    path_hash = "{:x}".format(ctypes.c_size_t(hash(file_path.absolute())).value)
+    module_name = path_hash
+    spec = importlib.util.spec_from_file_location(module_name, file_path)
+    if spec is None:
+        raise ImportError(f"Cannot load spec for {module_name} from {file_path}")
+    module = importlib.util.module_from_spec(spec)
+    if module is None:
+        raise ImportError(f"Cannot load module {module_name} from spec")
+    sys.modules[module_name] = module
+    spec.loader.exec_module(module)  # type: ignore
+    return module
+globals().update(vars(_import_from_path(Path(__file__).parent.parent / "__init__.py")))

build/torch210-cxx11-cu130-x86_64-linux/__init__.py ADDED Viewed

	@@ -0,0 +1,11 @@

+from ._ops import ops
+fast_hash = ops.fast_hash
+lsh_cumulation = ops.lsh_cumulation
+lsh_weighted_cumulation = ops.lsh_weighted_cumulation
+__all__ = [
+    "fast_hash",
+    "lsh_cumulation",
+    "lsh_weighted_cumulation",
+]

build/torch210-cxx11-cu130-x86_64-linux/_ops.py ADDED Viewed

	@@ -0,0 +1,9 @@

+import torch
+from . import _yoso_cuda_4218a9a
+ops = torch.ops._yoso_cuda_4218a9a
+def add_op_namespace_prefix(op_name: str):
+    """
+    Prefix op by namespace.
+    """
+    return f"_yoso_cuda_4218a9a::{op_name}"

build/torch210-cxx11-cu130-x86_64-linux/_yoso_cuda_4218a9a.abi3.so ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:7a4d07f13664724f51505edd45c4656f658f5d3ed8240f70d896214a8f15cac1
+size 3102168

build/torch210-cxx11-cu130-x86_64-linux/metadata.json ADDED Viewed

	@@ -0,0 +1,14 @@

+{
+  "version": 1,
+  "python-depends": [],
+  "backend": {
+    "type": "cuda",
+    "archs": [
+      "10.0",
+      "12.0",
+      "8.0",
+      "8.9",
+      "9.0"
+    ]
+  }
+}

build/torch210-cxx11-cu130-x86_64-linux/yoso/__init__.py ADDED Viewed

	@@ -0,0 +1,26 @@

+import ctypes
+import importlib.util
+import sys
+from pathlib import Path
+from types import ModuleType
+def _import_from_path(file_path: Path) -> ModuleType:
+    # We cannot use the module name as-is, after adding it to `sys.modules`,
+    # it would also be used for other imports. So, we make a module name that
+    # depends on the path for it to be unique using the hex-encoded hash of
+    # the path.
+    path_hash = "{:x}".format(ctypes.c_size_t(hash(file_path.absolute())).value)
+    module_name = path_hash
+    spec = importlib.util.spec_from_file_location(module_name, file_path)
+    if spec is None:
+        raise ImportError(f"Cannot load spec for {module_name} from {file_path}")
+    module = importlib.util.module_from_spec(spec)
+    if module is None:
+        raise ImportError(f"Cannot load module {module_name} from spec")
+    sys.modules[module_name] = module
+    spec.loader.exec_module(module)  # type: ignore
+    return module
+globals().update(vars(_import_from_path(Path(__file__).parent.parent / "__init__.py")))

build/torch211-cxx11-cu126-aarch64-linux/__init__.py ADDED Viewed

	@@ -0,0 +1,11 @@

+from ._ops import ops
+fast_hash = ops.fast_hash
+lsh_cumulation = ops.lsh_cumulation
+lsh_weighted_cumulation = ops.lsh_weighted_cumulation
+__all__ = [
+    "fast_hash",
+    "lsh_cumulation",
+    "lsh_weighted_cumulation",
+]

build/torch211-cxx11-cu126-aarch64-linux/_ops.py ADDED Viewed

	@@ -0,0 +1,9 @@

+import torch
+from . import _yoso_cuda_4218a9a
+ops = torch.ops._yoso_cuda_4218a9a
+def add_op_namespace_prefix(op_name: str):
+    """
+    Prefix op by namespace.
+    """
+    return f"_yoso_cuda_4218a9a::{op_name}"

build/torch211-cxx11-cu126-aarch64-linux/_yoso_cuda_4218a9a.abi3.so ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:17de742590f0b9c20b51ad23d0fdc7a8d2c6b4d8c5238aeadce31631ac1d8c12
+size 2569064

build/torch211-cxx11-cu126-aarch64-linux/metadata.json ADDED Viewed

	@@ -0,0 +1,12 @@

+{
+  "version": 1,
+  "python-depends": [],
+  "backend": {
+    "type": "cuda",
+    "archs": [
+      "8.0",
+      "8.9",
+      "9.0"
+    ]
+  }
+}

build/torch211-cxx11-cu126-aarch64-linux/yoso/__init__.py ADDED Viewed

	@@ -0,0 +1,26 @@

+import ctypes
+import importlib.util
+import sys
+from pathlib import Path
+from types import ModuleType
+def _import_from_path(file_path: Path) -> ModuleType:
+    # We cannot use the module name as-is, after adding it to `sys.modules`,
+    # it would also be used for other imports. So, we make a module name that
+    # depends on the path for it to be unique using the hex-encoded hash of
+    # the path.
+    path_hash = "{:x}".format(ctypes.c_size_t(hash(file_path.absolute())).value)
+    module_name = path_hash
+    spec = importlib.util.spec_from_file_location(module_name, file_path)
+    if spec is None:
+        raise ImportError(f"Cannot load spec for {module_name} from {file_path}")
+    module = importlib.util.module_from_spec(spec)
+    if module is None:
+        raise ImportError(f"Cannot load module {module_name} from spec")
+    sys.modules[module_name] = module
+    spec.loader.exec_module(module)  # type: ignore
+    return module
+globals().update(vars(_import_from_path(Path(__file__).parent.parent / "__init__.py")))

build/torch211-cxx11-cu126-x86_64-linux/__init__.py ADDED Viewed

	@@ -0,0 +1,11 @@

+from ._ops import ops
+fast_hash = ops.fast_hash
+lsh_cumulation = ops.lsh_cumulation
+lsh_weighted_cumulation = ops.lsh_weighted_cumulation
+__all__ = [
+    "fast_hash",
+    "lsh_cumulation",
+    "lsh_weighted_cumulation",
+]

build/torch211-cxx11-cu126-x86_64-linux/_ops.py ADDED Viewed

	@@ -0,0 +1,9 @@

+import torch
+from . import _yoso_cuda_4218a9a
+ops = torch.ops._yoso_cuda_4218a9a
+def add_op_namespace_prefix(op_name: str):
+    """
+    Prefix op by namespace.
+    """
+    return f"_yoso_cuda_4218a9a::{op_name}"

build/torch211-cxx11-cu126-x86_64-linux/_yoso_cuda_4218a9a.abi3.so ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:1567356a387001e1b7c487e2eb191443b9bb46f9e593455609215970c6d96f60
+size 2498248

build/torch211-cxx11-cu126-x86_64-linux/metadata.json ADDED Viewed

	@@ -0,0 +1,12 @@

+{
+  "version": 1,
+  "python-depends": [],
+  "backend": {
+    "type": "cuda",
+    "archs": [
+      "8.0",
+      "8.9",
+      "9.0"
+    ]
+  }
+}

build/torch211-cxx11-cu126-x86_64-linux/yoso/__init__.py ADDED Viewed

	@@ -0,0 +1,26 @@

+import ctypes
+import importlib.util
+import sys
+from pathlib import Path
+from types import ModuleType
+def _import_from_path(file_path: Path) -> ModuleType:
+    # We cannot use the module name as-is, after adding it to `sys.modules`,
+    # it would also be used for other imports. So, we make a module name that
+    # depends on the path for it to be unique using the hex-encoded hash of
+    # the path.
+    path_hash = "{:x}".format(ctypes.c_size_t(hash(file_path.absolute())).value)
+    module_name = path_hash
+    spec = importlib.util.spec_from_file_location(module_name, file_path)
+    if spec is None:
+        raise ImportError(f"Cannot load spec for {module_name} from {file_path}")
+    module = importlib.util.module_from_spec(spec)
+    if module is None:
+        raise ImportError(f"Cannot load module {module_name} from spec")
+    sys.modules[module_name] = module
+    spec.loader.exec_module(module)  # type: ignore
+    return module
+globals().update(vars(_import_from_path(Path(__file__).parent.parent / "__init__.py")))

build/torch211-cxx11-cu128-aarch64-linux/__init__.py ADDED Viewed

	@@ -0,0 +1,11 @@

+from ._ops import ops
+fast_hash = ops.fast_hash
+lsh_cumulation = ops.lsh_cumulation
+lsh_weighted_cumulation = ops.lsh_weighted_cumulation
+__all__ = [
+    "fast_hash",
+    "lsh_cumulation",
+    "lsh_weighted_cumulation",
+]