#11512: Add frac, ceil and trunc sweeps

.github/workflows/ttnn-run-sweeps.yaml

@@ -13,6 +13,9 @@ on:
           - add
           - line_all_gather
           - logical_and_
+          - eltwise.frac
+          - eltwise.ceil
+          - eltwise.trunc
           - matmul.full.matmul_default_block_sharded
           - matmul.full.matmul_default_height_sharded
           - matmul.full.matmul_default_interleaved

tests/sweep_framework/sweeps/eltwise/ceil.py

@@ -0,0 +1,77 @@
+# SPDX-FileCopyrightText: © 2024 Tenstorrent Inc.
+
+# SPDX-License-Identifier: Apache-2.0
+
+import torch
+
+import ttnn
+
+from tests.ttnn.utils_for_testing import check_with_pcc, start_measuring_time, stop_measuring_time
+from models.utility_functions import torch_random
+
+# Override the default timeout in seconds for hang detection.
+TIMEOUT = 30
+
+# Parameters provided to the test vector generator are defined here.
+# They are defined as dict-type suites that contain the arguments to the run function as keys, and lists of possible inputs as values.
+# Each suite has a key name (in this case "suite_1") which will associate the test vectors to this specific suite of inputs.
+# Developers can create their own generator functions and pass them to the parameters as inputs.
+parameters = {
+    "suite_1": {
+        "input_shape": [
+            [8, 1, 33, 256],
+            [8, 1, 256, 32],
+            [8, 8, 256, 384],
+            [8, 5, 13, 512],
+            [8, 5, 32, 512],
+            [1, 1, 32, 16384],
+        ],
+        "input_a_dtype": [ttnn.bfloat16, ttnn.bfloat8_b],
+        "input_a_layout": [ttnn.TILE_LAYOUT],
+        "input_a_memory_config": [ttnn.DRAM_MEMORY_CONFIG, ttnn.L1_MEMORY_CONFIG],
+        "output_memory_config": [ttnn.DRAM_MEMORY_CONFIG, ttnn.L1_MEMORY_CONFIG],
+    },
+}
+
+
+# This is the run instructions for the test, defined by the developer.
+# The run function must take the above-defined parameters as inputs.
+# The runner will call this run function with each test vector, and the returned results from this function will be stored.
+# If you defined a device_mesh_fixture above, the object you yielded will be passed into this function as 'device'. Otherwise, it will be the default ttnn device opened by the infra.
+def run(
+    input_shape,
+    input_a_dtype,
+    input_a_layout,
+    input_a_memory_config,
+    output_memory_config,
+    *,
+    device,
+) -> list:
+    torch_input_tensor_a = torch_random(input_shape, -100, 100, dtype=torch.bfloat16)
+
+    if input_a_dtype == ttnn.bfloat16:
+        torch_input_tensor_a = torch_input_tensor_a.to(torch.bfloat16)
+
+    elif input_a_dtype == ttnn.bfloat8_b:
+        tt_tensor = ttnn.from_torch(
+            torch_input_tensor_a, dtype=ttnn.bfloat8_b, layout=ttnn.TILE_LAYOUT, device=None, memory_config=None
+        )
+
+        torch_input_tensor_a = ttnn.to_torch(tt_tensor)
+
+    torch_output_tensor = torch.ceil(torch_input_tensor_a)
+
+    input_tensor_a = ttnn.from_torch(
+        torch_input_tensor_a,
+        dtype=input_a_dtype,
+        layout=input_a_layout,
+        device=device,
+        memory_config=input_a_memory_config,
+    )
+
+    start_time = start_measuring_time()
+    result = ttnn.ceil(input_tensor_a, memory_config=output_memory_config)
+    output_tensor = ttnn.to_torch(result)
+    e2e_perf = stop_measuring_time(start_time)
+
+    return [check_with_pcc(torch_output_tensor, output_tensor, 0.999), e2e_perf]

tests/sweep_framework/sweeps/eltwise/frac.py

@@ -0,0 +1,77 @@
+# SPDX-FileCopyrightText: © 2024 Tenstorrent Inc.
+
+# SPDX-License-Identifier: Apache-2.0
+
+import torch
+
+import ttnn
+
+from tests.ttnn.utils_for_testing import check_with_pcc, start_measuring_time, stop_measuring_time
+from models.utility_functions import torch_random
+
+# Override the default timeout in seconds for hang detection.
+TIMEOUT = 30
+
+# Parameters provided to the test vector generator are defined here.
+# They are defined as dict-type suites that contain the arguments to the run function as keys, and lists of possible inputs as values.
+# Each suite has a key name (in this case "suite_1") which will associate the test vectors to this specific suite of inputs.
+# Developers can create their own generator functions and pass them to the parameters as inputs.
+parameters = {
+    "suite_1": {
+        "input_shape": [
+            [8, 1, 33, 256],
+            [8, 1, 256, 32],
+            [8, 8, 256, 384],
+            [8, 5, 13, 512],
+            [8, 5, 32, 512],
+            [1, 1, 32, 16384],
+        ],
+        "input_a_dtype": [ttnn.bfloat16, ttnn.bfloat8_b],
+        "input_a_layout": [ttnn.TILE_LAYOUT],
+        "input_a_memory_config": [ttnn.DRAM_MEMORY_CONFIG, ttnn.L1_MEMORY_CONFIG],
+        "output_memory_config": [ttnn.DRAM_MEMORY_CONFIG, ttnn.L1_MEMORY_CONFIG],
+    },
+}
+
+
+# This is the run instructions for the test, defined by the developer.
+# The run function must take the above-defined parameters as inputs.
+# The runner will call this run function with each test vector, and the returned results from this function will be stored.
+# If you defined a device_mesh_fixture above, the object you yielded will be passed into this function as 'device'. Otherwise, it will be the default ttnn device opened by the infra.
+def run(
+    input_shape,
+    input_a_dtype,
+    input_a_layout,
+    input_a_memory_config,
+    output_memory_config,
+    *,
+    device,
+) -> list:
+    torch_input_tensor_a = torch_random(input_shape, -100, 100, dtype=torch.bfloat16)
+
+    if input_a_dtype == ttnn.bfloat16:
+        torch_input_tensor_a = torch_input_tensor_a.to(torch.bfloat16)
+
+    elif input_a_dtype == ttnn.bfloat8_b:
+        tt_tensor = ttnn.from_torch(
+            torch_input_tensor_a, dtype=ttnn.bfloat8_b, layout=ttnn.TILE_LAYOUT, device=None, memory_config=None
+        )
+
+        torch_input_tensor_a = ttnn.to_torch(tt_tensor)
+
+    torch_output_tensor = torch.frac(torch_input_tensor_a)
+
+    input_tensor_a = ttnn.from_torch(
+        torch_input_tensor_a,
+        dtype=input_a_dtype,
+        layout=input_a_layout,
+        device=device,
+        memory_config=input_a_memory_config,
+    )
+
+    start_time = start_measuring_time()
+    result = ttnn.frac(input_tensor_a, memory_config=output_memory_config)
+    output_tensor = ttnn.to_torch(result)
+    e2e_perf = stop_measuring_time(start_time)
+
+    return [check_with_pcc(torch_output_tensor, output_tensor, 0.999), e2e_perf]

tests/sweep_framework/sweeps/eltwise/trunc.py

@@ -0,0 +1,77 @@
+# SPDX-FileCopyrightText: © 2024 Tenstorrent Inc.
+
+# SPDX-License-Identifier: Apache-2.0
+
+import torch
+
+import ttnn
+
+from tests.ttnn.utils_for_testing import check_with_pcc, start_measuring_time, stop_measuring_time
+from models.utility_functions import torch_random
+
+# Override the default timeout in seconds for hang detection.
+TIMEOUT = 30
+
+# Parameters provided to the test vector generator are defined here.
+# They are defined as dict-type suites that contain the arguments to the run function as keys, and lists of possible inputs as values.
+# Each suite has a key name (in this case "suite_1") which will associate the test vectors to this specific suite of inputs.
+# Developers can create their own generator functions and pass them to the parameters as inputs.
+parameters = {
+    "suite_1": {
+        "input_shape": [
+            [8, 1, 33, 256],
+            [8, 1, 256, 32],
+            [8, 8, 256, 384],
+            [8, 5, 13, 512],
+            [8, 5, 32, 512],
+            [1, 1, 32, 16384],
+        ],
+        "input_a_dtype": [ttnn.bfloat16, ttnn.bfloat8_b],
+        "input_a_layout": [ttnn.TILE_LAYOUT],
+        "input_a_memory_config": [ttnn.DRAM_MEMORY_CONFIG, ttnn.L1_MEMORY_CONFIG],
+        "output_memory_config": [ttnn.DRAM_MEMORY_CONFIG, ttnn.L1_MEMORY_CONFIG],
+    },
+}
+
+
+# This is the run instructions for the test, defined by the developer.
+# The run function must take the above-defined parameters as inputs.
+# The runner will call this run function with each test vector, and the returned results from this function will be stored.
+# If you defined a device_mesh_fixture above, the object you yielded will be passed into this function as 'device'. Otherwise, it will be the default ttnn device opened by the infra.
+def run(
+    input_shape,
+    input_a_dtype,
+    input_a_layout,
+    input_a_memory_config,
+    output_memory_config,
+    *,
+    device,
+) -> list:
+    torch_input_tensor_a = torch_random(input_shape, -100, 100, dtype=torch.bfloat16)
+
+    if input_a_dtype == ttnn.bfloat16:
+        torch_input_tensor_a = torch_input_tensor_a.to(torch.bfloat16)
+
+    elif input_a_dtype == ttnn.bfloat8_b:
+        tt_tensor = ttnn.from_torch(
+            torch_input_tensor_a, dtype=ttnn.bfloat8_b, layout=ttnn.TILE_LAYOUT, device=None, memory_config=None
+        )
+
+        torch_input_tensor_a = ttnn.to_torch(tt_tensor)
+
+    torch_output_tensor = torch.trunc(torch_input_tensor_a)
+
+    input_tensor_a = ttnn.from_torch(
+        torch_input_tensor_a,
+        dtype=input_a_dtype,
+        layout=input_a_layout,
+        device=device,
+        memory_config=input_a_memory_config,
+    )
+
+    start_time = start_measuring_time()
+    result = ttnn.trunc(input_tensor_a, memory_config=output_memory_config)
+    output_tensor = ttnn.to_torch(result)
+    e2e_perf = stop_measuring_time(start_time)
+
+    return [check_with_pcc(torch_output_tensor, output_tensor, 0.999), e2e_perf]