Skip to content

Commit

Permalink
#8142: TTNN log-alike ops incomplete documentation (#13969)
Browse files Browse the repository at this point in the history 
* #8142: Add documentation for log alike ops

* #8142: Update sweep tests
  • Loading branch information
@mouliraj-mcw
mouliraj-mcw authored Oct 24, 2024
1 parent c93572e commit af3d010
Show file tree
Hide file tree
Showing 5 changed files with 157 additions and 99 deletions.
50 changes: 26 additions & 24 deletions tests/sweep_framework/sweeps/eltwise/unary/log10/log10.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -6,19 +6,13 @@
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.
Expand All @@ -28,45 +22,53 @@
"input_shape": gen_shapes([1, 1, 32, 32], [6, 12, 256, 256], [1, 1, 32, 32], 16)
+ gen_shapes([1, 32, 32], [12, 256, 256], [1, 32, 32], 16)
+ gen_shapes([32, 32], [256, 256], [32, 32], 32),
"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],
"input_dtype": [ttnn.bfloat16, ttnn.bfloat8_b],
"input_layout": [ttnn.TILE_LAYOUT, ttnn.ROW_MAJOR_LAYOUT],
"input_memory_config": [ttnn.DRAM_MEMORY_CONFIG, ttnn.L1_MEMORY_CONFIG],
"output_memory_config": [ttnn.DRAM_MEMORY_CONFIG, ttnn.L1_MEMORY_CONFIG],
},
}


# 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.
def invalidate_vector(test_vector) -> Tuple[bool, Optional[str]]:
if test_vector["input_layout"] == ttnn.ROW_MAJOR_LAYOUT:
return True, "Row Major layout is 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.
# 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,
input_dtype,
input_layout,
input_memory_config,
output_memory_config,
*,
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=1, high=100, dtype=torch.float32), input_a_dtype
torch_input_tensor = gen_func_with_cast_tt(
partial(torch_random, low=1, high=100, dtype=torch.float32), input_dtype
)(input_shape)
torch_output_tensor = torch.log10(torch_input_tensor_a)

input_tensor_a = ttnn.from_torch(
torch_input_tensor_a,
dtype=input_a_dtype,
layout=input_a_layout,
golden_function = ttnn.get_golden_function(ttnn.log10)
torch_output_tensor = golden_function(torch_input_tensor)
input_tensor = ttnn.from_torch(
torch_input_tensor,
dtype=input_dtype,
layout=input_layout,
device=device,
memory_config=input_a_memory_config,
memory_config=input_memory_config,
)

start_time = start_measuring_time()
result = ttnn.log10(input_tensor_a, memory_config=output_memory_config)
result = ttnn.log10(input_tensor, memory_config=output_memory_config)
output_tensor = ttnn.to_torch(result)
e2e_perf = stop_measuring_time(start_time)

Expand Down
50 changes: 27 additions & 23 deletions tests/sweep_framework/sweeps/eltwise/unary/log1p/log1p.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -6,19 +6,13 @@
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.
Expand All @@ -28,45 +22,55 @@
"input_shape": gen_shapes([1, 1, 32, 32], [6, 12, 256, 256], [1, 1, 32, 32], 16)
+ gen_shapes([1, 32, 32], [12, 256, 256], [1, 32, 32], 16)
+ gen_shapes([32, 32], [256, 256], [32, 32], 32),
"input_a_dtype": [ttnn.bfloat16],
"input_a_layout": [ttnn.TILE_LAYOUT],
"input_a_memory_config": [ttnn.DRAM_MEMORY_CONFIG, ttnn.L1_MEMORY_CONFIG],
"input_dtype": [ttnn.bfloat16, ttnn.bfloat8_b],
"input_layout": [ttnn.TILE_LAYOUT, ttnn.ROW_MAJOR_LAYOUT],
"input_memory_config": [ttnn.DRAM_MEMORY_CONFIG, ttnn.L1_MEMORY_CONFIG],
"output_memory_config": [ttnn.DRAM_MEMORY_CONFIG, ttnn.L1_MEMORY_CONFIG],
},
}


# 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.
def invalidate_vector(test_vector) -> Tuple[bool, Optional[str]]:
if test_vector["input_layout"] == ttnn.ROW_MAJOR_LAYOUT or test_vector["input_dtype"] == ttnn.bfloat8_b:
return True, "Row Major layout and bfloat8_b are 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.
# 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,
input_dtype,
input_layout,
input_memory_config,
output_memory_config,
*,
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=1, high=100, dtype=torch.float32), input_a_dtype
torch_input_tensor = gen_func_with_cast_tt(
partial(torch_random, low=1, high=100, dtype=torch.float32), input_dtype
)(input_shape)
torch_output_tensor = torch.log1p(torch_input_tensor_a)

input_tensor_a = ttnn.from_torch(
torch_input_tensor_a,
dtype=input_a_dtype,
layout=input_a_layout,
golden_function = ttnn.get_golden_function(ttnn.log1p)
torch_output_tensor = golden_function(torch_input_tensor)

input_tensor = ttnn.from_torch(
torch_input_tensor,
dtype=input_dtype,
layout=input_layout,
device=device,
memory_config=input_a_memory_config,
memory_config=input_memory_config,
)

start_time = start_measuring_time()
result = ttnn.log1p(input_tensor_a, memory_config=output_memory_config)
result = ttnn.log1p(input_tensor, memory_config=output_memory_config)
output_tensor = ttnn.to_torch(result)
e2e_perf = stop_measuring_time(start_time)

Expand Down
49 changes: 26 additions & 23 deletions tests/sweep_framework/sweeps/eltwise/unary/log2/log2.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -6,19 +6,13 @@
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.
Expand All @@ -28,45 +22,54 @@
"input_shape": gen_shapes([1, 1, 32, 32], [6, 12, 256, 256], [1, 1, 32, 32], 16)
+ gen_shapes([1, 32, 32], [12, 256, 256], [1, 32, 32], 16)
+ gen_shapes([32, 32], [256, 256], [32, 32], 32),
"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],
"input_dtype": [ttnn.bfloat16, ttnn.bfloat8_b],
"input_layout": [ttnn.TILE_LAYOUT, ttnn.ROW_MAJOR_LAYOUT],
"input_memory_config": [ttnn.DRAM_MEMORY_CONFIG, ttnn.L1_MEMORY_CONFIG],
"output_memory_config": [ttnn.DRAM_MEMORY_CONFIG, ttnn.L1_MEMORY_CONFIG],
},
}


# 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.
def invalidate_vector(test_vector) -> Tuple[bool, Optional[str]]:
if test_vector["input_layout"] == ttnn.ROW_MAJOR_LAYOUT:
return True, "Row Major layout is 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.
# 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,
input_dtype,
input_layout,
input_memory_config,
output_memory_config,
*,
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=1, high=100, dtype=torch.float32), input_a_dtype
torch_input_tensor = gen_func_with_cast_tt(
partial(torch_random, low=1, high=100, dtype=torch.float32), input_dtype
)(input_shape)
torch_output_tensor = torch.log2(torch_input_tensor_a)
golden_function = ttnn.get_golden_function(ttnn.log2)
torch_output_tensor = golden_function(torch_input_tensor)

input_tensor_a = ttnn.from_torch(
torch_input_tensor_a,
dtype=input_a_dtype,
layout=input_a_layout,
input_tensor = ttnn.from_torch(
torch_input_tensor,
dtype=input_dtype,
layout=input_layout,
device=device,
memory_config=input_a_memory_config,
memory_config=input_memory_config,
)

start_time = start_measuring_time()
result = ttnn.log2(input_tensor_a, memory_config=output_memory_config)
result = ttnn.log2(input_tensor, memory_config=output_memory_config)
output_tensor = ttnn.to_torch(result)
e2e_perf = stop_measuring_time(start_time)

Expand Down
49 changes: 26 additions & 23 deletions tests/sweep_framework/sweeps/eltwise/unary/log_sigmoid/log_sigmoid.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -6,19 +6,13 @@
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.
Expand All @@ -28,45 +22,54 @@
"input_shape": gen_shapes([1, 1, 32, 32], [6, 12, 256, 256], [1, 1, 32, 32], 16)
+ gen_shapes([1, 32, 32], [12, 256, 256], [1, 32, 32], 16)
+ gen_shapes([32, 32], [256, 256], [32, 32], 32),
"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],
"input_dtype": [ttnn.bfloat16, ttnn.bfloat8_b],
"input_layout": [ttnn.TILE_LAYOUT, ttnn.ROW_MAJOR_LAYOUT],
"input_memory_config": [ttnn.DRAM_MEMORY_CONFIG, ttnn.L1_MEMORY_CONFIG],
"output_memory_config": [ttnn.DRAM_MEMORY_CONFIG, ttnn.L1_MEMORY_CONFIG],
},
}


# 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.
def invalidate_vector(test_vector) -> Tuple[bool, Optional[str]]:
if test_vector["input_layout"] == ttnn.ROW_MAJOR_LAYOUT:
return True, "Row Major layout is 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.
# 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,
input_dtype,
input_layout,
input_memory_config,
output_memory_config,
*,
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=-4, high=10, dtype=torch.float32), input_a_dtype
torch_input_tensor = gen_func_with_cast_tt(
partial(torch_random, low=-4, high=10, dtype=torch.float32), input_dtype
)(input_shape)
torch_output_tensor = torch.nn.functional.logsigmoid(torch_input_tensor_a)
golden_function = ttnn.get_golden_function(ttnn.log_sigmoid)
torch_output_tensor = golden_function(torch_input_tensor)

input_tensor_a = ttnn.from_torch(
torch_input_tensor_a,
dtype=input_a_dtype,
layout=input_a_layout,
input_tensor = ttnn.from_torch(
torch_input_tensor,
dtype=input_dtype,
layout=input_layout,
device=device,
memory_config=input_a_memory_config,
memory_config=input_memory_config,
)

start_time = start_measuring_time()
result = ttnn.log_sigmoid(input_tensor_a, memory_config=output_memory_config)
result = ttnn.log_sigmoid(input_tensor, memory_config=output_memory_config)
output_tensor = ttnn.to_torch(result)
e2e_perf = stop_measuring_time(start_time)

Expand Down
Loading

0 comments on commit af3d010

Please sign in to comment.