#13536: testing without 0-dim and 1D restrictions

tests/sweep_framework/sweeps/eltwise/binary/add/add_all_pytorch2.py

@@ -14,20 +14,16 @@
 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
-
-random.seed(0)
 
 # 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 = {
-    "nightly": {
+    "test_bcast_1d_13": {
         "input_shape": [
-            {"self": [0, 1], "other": [0, 1]},
-            {"self": [0], "other": [0]},
+            # {"self": [0, 1], "other": [0, 1]}, #0 is not a valid shape
+            # {"self": [0], "other": [0]}, #0 is not a valid shape
             {"self": [1, 1, 1024], "other": [1, 1, 1024]},
             {"self": [1, 1, 16, 32], "other": [1, 1, 16, 32]},
             {"self": [1, 1, 3072], "other": [1, 1, 3072]},
@@ -403,7 +399,7 @@
             {"self": [8732, 1], "other": [8732, 1]},
             {"self": [8732, 2], "other": [8732, 2]},
             {"self": [8732], "other": [8732]},
-            {"self": [], "other": []},
+            # {"self": [], "other": []}, #without empty tensor
             {"self": [920, 1, 256], "other": [256]},
             {"self": [920, 1, 256], "other": [920, 1, 256]},
             {"self": [1, 1, 1, 42], "other": -6.0},
@@ -504,7 +500,7 @@
             {"self": [7], "other": 0.0},
             {"self": [800], "other": 0.5},
             {"self": [80], "other": 0.5},
-            {"self": [], "other": 1},
+            # {"self": [], "other": 1}, #without empty tensor
         ],
         # {"self": [s0 + 1, s0 + 1], "other": 16},
         # {"self": [s0 + 1, s0 + 1], "other": 0},
@@ -533,6 +529,24 @@
 }
 
 
+# Invalidate vector is called during the generation phase where each vector will be passed in.
+# If invalidated, the vector will still be stored but will be skipped.
+# Returns False, None if the vector is valid, and True, str with a reason for invalidation if it is invalid. len(test_vector["input_shape"]["other"]) >= 4
+# def invalidate_vector(test_vector) -> Tuple[bool, Optional[str]]:
+#     if isinstance(test_vector["input_shape"]["other"], list) and len(test_vector["input_shape"]["self"]) >= 4:
+#         is_less_than_3d = len(test_vector["input_shape"]["other"]) < 3
+#         c_index = test_vector["input_shape"]["self"][-3]
+#         if is_less_than_3d or (
+#             len(test_vector["input_shape"]["other"]) >= 3 and test_vector["input_shape"]["other"][-3] < c_index
+#         ):
+#             print("checking channel bcast")
+#             print("input ", test_vector["input_shape"]["self"])
+#             print("other ", test_vector["input_shape"]["other"])
+#             return True, "channel dim bcast not supported"
+
+#     return False, None
+
+
 # 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.
@@ -548,19 +562,23 @@ def run(
     *,
     device,
 ) -> list:
-    data_seed = random.randint(0, 20000000)
-    torch.manual_seed(data_seed)
+    torch.manual_seed(0)
 
     torch_input_tensor_a = gen_func_with_cast_tt(
-        partial(torch_random, low=-100, high=100, dtype=torch.float32), input_a_dtype
+        partial(torch_random, low=-100, high=100, dtype=torch.bfloat16), input_a_dtype
     )(input_shape["self"])
 
     if isinstance(input_shape["other"], list):
-        torch_input_tensor_b = gen_func_with_cast_tt(
-            partial(torch_random, low=-100, high=100, dtype=torch.float32), input_b_dtype
-        )(input_shape["other"])
+        if len(input_shape["other"]):
+            torch_input_tensor_b = gen_func_with_cast_tt(
+                partial(torch_random, low=-100, high=100, dtype=torch.bfloat16), input_b_dtype
+            )(input_shape["other"])
+        else:
+            print("input shape", input_shape)
+            torch_input_tensor_b = torch.tensor(0, dtype=torch.bfloat16)
+            print("torch_input_tensor_b shape", torch_input_tensor_b)
     else:
-        torch_input_tensor_b = torch.tensor(input_shape["other"], dtype=torch.float32)
+        torch_input_tensor_b = torch.tensor(input_shape["other"], dtype=torch.bfloat16)
         # torch_input_tensor_b = input_shape["other"]
 
     golden_function = ttnn.get_golden_function(ttnn.add)
@@ -587,7 +605,13 @@ def run(
 
     start_time = start_measuring_time()
     result = ttnn.add(input_tensor_a, input_tensor_b)
-    output_tensor = ttnn.to_torch(result)
+
+    # handles 1 D input_a and scalar or empty [] input_b
+    if len(input_shape["self"]) == 1 and (not isinstance(input_shape["other"], list) or not input_shape["other"]):
+        output_tensor = ttnn.to_torch(result, original_shape=input_shape["self"])
+    else:
+        output_tensor = ttnn.to_torch(result)
+
     e2e_perf = stop_measuring_time(start_time)
 
-    return [check_with_pcc(torch_output_tensor, output_tensor, pcc=0.9999), e2e_perf]
+    return [check_with_pcc(torch_output_tensor, output_tensor, pcc=0.999), e2e_perf]

tests/sweep_framework/sweeps/eltwise/binary/div/div_tensor_pytorch2.py

@@ -6,22 +6,16 @@
 from functools import partial
 
 import torch
-import random
 import ttnn
 from tests.sweep_framework.sweep_utils.utils import gen_shapes
 from tests.tt_eager.python_api_testing.sweep_tests.generation_funcs import gen_func_with_cast_tt
 
 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
-
-random.seed(0)
-
 
 parameters = {
-    "nightly": {
+    "nightly_14": {
         "input_specs": [
             {"shape": [0, 1], "other": 1.0},
             {"shape": [1, 1, 16384, 256], "other": 5.656854249492381},
@@ -143,8 +137,7 @@ def run(
     *,
     device,
 ) -> list:
-    data_seed = random.randint(0, 20000000)
-    torch.manual_seed(data_seed)
+    torch.manual_seed(0)
 
     input_shape = input_specs["shape"]
     if len(input_shape) == 0:
@@ -186,7 +179,14 @@ def run(
     start_time = start_measuring_time()
 
     output_tensor = ttnn.divide(input_tensor_a, input_tensor_b, memory_config=output_memory_config)
-    output_tensor = ttnn.to_torch(output_tensor)
+
+    # handles 1 D input_a and scalar or empty [] input_b
+    if len(input_specs["shape"]) == 1 and (
+        not isinstance(input_specs["other"], list) or not input_specs["other"] or len(input_specs["other"]) == 1
+    ):
+        output_tensor = ttnn.to_torch(output_tensor, original_shape=input_specs["shape"])
+    else:
+        output_tensor = ttnn.to_torch(output_tensor)
 
     e2e_perf = stop_measuring_time(start_time)
 

tests/sweep_framework/sweeps/eltwise/binary/eq/eq_scalar_pytorch2.py

@@ -25,7 +25,7 @@
 # Each suite has a key name (in this case "suite_1" and "suite_2") 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 = {
-    "nightly": {
+    "pass_1": {
         "input_shape": [
             [1, 1, 256],
             [1, 16],
@@ -83,7 +83,11 @@ def run(
 
     start_time = start_measuring_time()
     output_tensor = ttnn.eq(input_tensor_a, scalar, memory_config=output_memory_config)
-    output_tensor = ttnn.to_torch(output_tensor)
+    # to handle 1D inputs giving 2D outputs
+    if len(input_shape) == 1:
+        output_tensor = ttnn.to_torch(output_tensor, original_shape=input_shape)
+    else:
+        output_tensor = ttnn.to_torch(output_tensor)
     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/binary/floor_divide/floor_divide_pytorch2.py

@@ -25,7 +25,7 @@
 # Each suite has a key name (in this case "suite_1" and "suite_2") 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 = {
-    "nightly": {
+    "pass_1": {
         "input_shape": [
             [128],
         ],
@@ -91,7 +91,11 @@ def run(
 
     start_time = start_measuring_time()
     output_tensor = ttnn.floor_div(input_tensor_a, input_tensor_b, memory_config=output_memory_config)
-    output_tensor = ttnn.to_torch(output_tensor)
+    # to handle 1D inputs giving 2D outputs
+    if len(input_shape) == 1:
+        output_tensor = ttnn.to_torch(output_tensor, original_shape=input_shape)
+    else:
+        output_tensor = ttnn.to_torch(output_tensor)
     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/binary/gt/gt_scalar_pytorch2.py

@@ -25,11 +25,11 @@
 # Each suite has a key name (in this case "suite_1" and "suite_2") 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 = {
-    "nightly": {
+    "pass_1": {
         "input_shape": [
             [10, 10],
             [15, 15],
-            [],
+            [],  # how should this be tested ?
         ],
         "scalar": [0, 0],
         "input_a_dtype": [ttnn.bfloat16],
@@ -74,7 +74,11 @@ def run(
 
     start_time = start_measuring_time()
     output_tensor = ttnn.gt(input_tensor_a, scalar, memory_config=output_memory_config)
-    output_tensor = ttnn.to_torch(output_tensor)
+    # to handle 1D inputs giving 2D outputs
+    if len(input_shape) == 1:
+        output_tensor = ttnn.to_torch(output_tensor, original_shape=input_shape)
+    else:
+        output_tensor = ttnn.to_torch(output_tensor)
     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/binary/multiply/mul_tensor_pytorch2.py

@@ -6,27 +6,22 @@
 from functools import partial
 
 import torch
-import random
 import ttnn
 from tests.sweep_framework.sweep_utils.utils import gen_shapes
 from tests.tt_eager.python_api_testing.sweep_tests.generation_funcs import gen_func_with_cast_tt
 
 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
-
-random.seed(0)
 
 # 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 = {
-    "nightly": {
+    "test_bcast_1d_11": {
         "input_shape": [
-            {"self": [0], "other": 0.5},
+            {"self": [0], "other": 0.5},  # is [0] a valid shape ?
             {"self": [1, 1, 1, 10], "other": -3.4028234663852886e38},
             {"self": [1, 1, 1, 12], "other": -3.4028234663852886e38},
             {"self": [1, 1, 1, 14], "other": -3.4028234663852886e38},
@@ -191,8 +186,8 @@
             {"self": [8732, 2], "other": 0.5},
             {"self": [8732], "other": 0.5},
             # vec other
-            {"self": [0, 1], "other": [0, 1]},
-            {"self": [0], "other": []},
+            # {"self": [0, 1], "other": [0, 1]},
+            {"self": [0], "other": []},  # invalid shape [0]
             {"self": [1, 1, 1, 17], "other": [1, 1, 1, 17]},
             {"self": [1, 1, 1, 1], "other": [1, 1, 1, 1]},
             {"self": [1, 1, 1, 2], "other": [1, 1, 1, 2]},
@@ -412,6 +407,28 @@
 }
 
 
+# Invalidate vector is called during the generation phase where each vector will be passed in.
+# If invalidated, the vector will still be stored but will be skipped.
+# Returns False, None if the vector is valid, and True, str with a reason for invalidation if it is invalid. len(test_vector["input_shape"]["other"]) >= 4
+# def invalidate_vector(test_vector) -> Tuple[bool, Optional[str]]:
+#     if len(test_vector["input_shape"]["self"]) == 0 or (
+#         isinstance(test_vector["input_shape"]["other"], list) and len(test_vector["input_shape"]["other"]) == 0
+#     ):
+#         return True, "empty shape not supported"
+# if isinstance(test_vector["input_shape"]["other"], list) and len(test_vector["input_shape"]["self"]) >= 4:
+#     is_less_than_3d = len(test_vector["input_shape"]["other"]) < 3
+#     c_index = test_vector["input_shape"]["self"][-3]
+#     if is_less_than_3d or (
+#         len(test_vector["input_shape"]["other"]) >= 3 and test_vector["input_shape"]["other"][-3] < c_index
+#     ):
+#         print("checking channel bcast")
+#         print("input ", test_vector["input_shape"]["self"])
+#         print("other ", test_vector["input_shape"]["other"])
+#         return True, "channel dim bcast not supported"
+
+# return False, None
+
+
 # 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.
@@ -427,19 +444,21 @@ def run(
     *,
     device,
 ) -> list:
-    data_seed = random.randint(0, 20000000)
-    torch.manual_seed(data_seed)
+    torch.manual_seed(0)
 
     torch_input_tensor_a = gen_func_with_cast_tt(
-        partial(torch_random, low=-100, high=100, dtype=torch.float32), input_a_dtype
+        partial(torch_random, low=-100, high=100, dtype=torch.bfloat16), input_a_dtype
     )(input_shape["self"])
 
     if isinstance(input_shape["other"], list):
-        torch_input_tensor_b = gen_func_with_cast_tt(
-            partial(torch_random, low=-100, high=100, dtype=torch.float32), input_b_dtype
-        )(input_shape["other"])
+        if len(input_shape["other"]):
+            torch_input_tensor_b = gen_func_with_cast_tt(
+                partial(torch_random, low=-100, high=100, dtype=torch.bfloat16), input_b_dtype
+            )(input_shape["other"])
+        else:
+            torch_input_tensor_b = torch.tensor(0, dtype=torch.bfloat16)
     else:
-        torch_input_tensor_b = torch.tensor(input_shape["other"], dtype=torch.float32)
+        torch_input_tensor_b = torch.tensor(input_shape["other"], dtype=torch.bfloat16)
         # torch_input_tensor_b = input_shape["other"]
 
     golden_function = ttnn.get_golden_function(ttnn.mul)
@@ -466,7 +485,11 @@ def run(
 
     start_time = start_measuring_time()
     result = ttnn.mul(input_tensor_a, input_tensor_b)
-    output_tensor = ttnn.to_torch(result)
+    # handles 1 D input_a and scalar or empty [] input_b
+    if len(input_shape["self"]) == 1 and (not isinstance(input_shape["other"], list) or not input_shape["other"]):
+        output_tensor = ttnn.to_torch(result, original_shape=input_shape["self"])
+    else:
+        output_tensor = ttnn.to_torch(result)
     e2e_perf = stop_measuring_time(start_time)
 
     return [check_with_pcc(torch_output_tensor, output_tensor, pcc=0.99), e2e_perf]

tests/sweep_framework/sweeps/eltwise/binary/remainder/remainder_scalar_pytorch2.py

@@ -25,7 +25,7 @@
 # Each suite has a key name (in this case "suite_1" and "suite_2") 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 = {
-    "nightly": {
+    "pass_1": {
         "input_shape": [
             [1],
         ],
@@ -72,7 +72,11 @@ def run(
 
     start_time = start_measuring_time()
     output_tensor = ttnn.remainder(input_tensor_a, scalar, memory_config=output_memory_config)
-    output_tensor = ttnn.to_torch(output_tensor)
+    # to handle 1D inputs giving 2D outputs
+    if len(input_shape) == 1:
+        output_tensor = ttnn.to_torch(output_tensor, original_shape=input_shape)
+    else:
+        output_tensor = ttnn.to_torch(output_tensor)
     e2e_perf = stop_measuring_time(start_time)
 
     return [check_with_pcc(torch_output_tensor, output_tensor, 0.999), e2e_perf]