def _plugin_GPU_op_VS_tf_ops():
"""
Compare the result of converting to CSR between plugin GPU ops and tf ops.
"""
print("[INFO]: converting to CSR, plugin GPU vs tf ops..")
dataset_names = ['./performance_profile/train.tfrecord']
dataset_gpu = create_dataset(dataset_names=dataset_names,
feature_desc=feature_desc,
batch_size=batch_size,
n_epochs=1,
distribute_keys=True,
gpu_count=gpu_count,
embedding_type=embedding_type,
use_which_device='gpu')
dataset_tf = CreateDataset(dataset_names=dataset_names,
feature_desc=feature_desc,
batch_size=batch_size,
n_epochs=1,
slot_num=26,
max_nnz=1,
convert_to_csr=True,
gpu_count=gpu_count,
embedding_type=embedding_type)()
dataset_gpu = iter(dataset_gpu)
dataset_tf = iter(dataset_tf)
for iter_i in range(iterations):
row_indices, values, nnz_array_gpu = next(dataset_gpu)[2:5]
row_offsets_gpu, value_tensor_gpu, nnz_array_gpu = hugectr_tf_ops.distribute_keys_gpu(
row_indices=row_indices,
values=values,
embedding_name='hugectr_embedding',
embedding_type=embedding_type,
batch_size=batch_size,
slot_num=26,
gpu_count=gpu_count,
max_nnz=1)
row_offsets_tf, value_tensor_tf, nnz_array_tf = next(
dataset_tf)[2:5]
try:
tf.debugging.assert_equal(
row_offsets_gpu[:, 0:row_offsets_tf.shape[1]],
row_offsets_tf)
tf.debugging.assert_equal(
value_tensor_gpu[:, 0:value_tensor_tf.shape[1]],
value_tensor_tf)
tf.debugging.assert_equal(nnz_array_gpu, nnz_array_tf)
except tf.errors.InvalidArgumentError as error:
raise RuntimeError(
"Error in %s, gpu_count %d, batch_size %d." %
(embedding_type, gpu_count, batch_size), error.message)
print(
"[INFO]: For %s and gpu_count: %d, batch_size: %d, iteration: %d results is the same."
% (embedding_type, gpu_count, batch_size, iter_i))
hugectr_tf_ops.reset()
def _fprop_v4_VS_tf():
print("[INFO]: Testing fprop_v4 vs tf...")
if vocabulary_size < slot_num:
raise RuntimeError("vocabulary_size must > slot.")
with tf.GradientTape(persistent=True) as tape:
# initial embedding table
init_value = np.float32(
np.random.normal(loc=0,
scale=1,
size=(vocabulary_size, embedding_vec_size)))
# input keys
# TODO: Keys in different slots should be unique.
input_keys = np.ones(shape=(batch_size, slot_num, max_nnz),
dtype=np.int64) * -1
each_slot = vocabulary_size // slot_num
nnz_0_num = 0
for batch_id in range(batch_size):
for slot_id in range(slot_num):
nnz = np.random.randint(
low=nnz_0_num, high=max_nnz + 1,
size=1)[0] # how many keys in this slot
if nnz == 0:
nnz_0_num = 1
if (embedding_type == 'distributed'):
keys = np.random.randint(low=slot_id * each_slot,
high=(slot_id + 1) *
each_slot,
size=nnz)
elif (embedding_type == "localized"):
# TODO: key should belong to that slot.
keys = []
while len(keys) < nnz:
key = np.random.randint(low=slot_id * each_slot,
high=(slot_id + 1) *
each_slot,
size=1)
if key % slot_num == slot_id:
keys.append(key)
input_keys[batch_id, slot_id, 0:nnz] = keys
# hugectr ops
hugectr_tf_ops.init(visiable_gpus=gpus,
key_type='int64',
value_type='float',
batch_size=batch_size,
batch_size_eval=len(gpus))
embedding_name = hugectr_tf_ops.create_embedding(
init_value=init_value,
opt_hparams=[0.1, 0.9, 0.99, 1e-5],
name_='hugectr_embedding',
max_vocabulary_size_per_gpu=(vocabulary_size // len(gpus)) * 2
+ 1,
slot_num=slot_num,
embedding_vec_size=embedding_vec_size,
max_feature_num=slot_num * max_nnz,
embedding_type=embedding_type,
max_nnz=max_nnz,
update_type='Global')
reshape_input_keys = np.reshape(input_keys, [-1, max_nnz])
indices = tf.where(reshape_input_keys != -1)
values = tf.gather_nd(reshape_input_keys, indices)
row_indices = tf.transpose(indices, perm=[1, 0])[0]
bp_trigger = tf.Variable(initial_value=1.0,
trainable=True,
dtype=tf.float32)
hugectr_forward = hugectr_tf_ops.fprop_v4(
embedding_name=embedding_name,
row_indices=row_indices,
values=values,
bp_trigger=bp_trigger,
is_training=True,
output_shape=[batch_size, slot_num, max_nnz])
# print("hugectr_results=\n", hugectr_forward)
# tf ops
reshape_input_keys = np.reshape(input_keys, [-1, max_nnz])
tf_indices = tf.where(reshape_input_keys != -1)
tf_values = tf.gather_nd(reshape_input_keys, tf_indices)
sparse_tensor = tf.sparse.SparseTensor(tf_indices, tf_values,
reshape_input_keys.shape)
# FIXME: if there are too more nnz=0 slots, tf.nn.embedding_lookup_sparse may get wrong results?
tf_embedding_layer = OriginalEmbedding(
vocabulary_size=vocabulary_size,
embedding_vec_size=embedding_vec_size,
initializer=init_value,
combiner='sum',
gpus=gpus)
tf_forward = tf_embedding_layer(
sparse_tensor,
output_shape=[batch_size, slot_num, embedding_vec_size])
# print("tf_results=\n", tf_forward)
# compare first forward result
try:
tf.debugging.assert_near(hugectr_forward, tf_forward)
except tf.errors.InvalidArgumentError as error:
raise error
print(
"[INFO]: The results from HugeCTR and tf in the first forward propagation are the same."
)
# backward
hugectr_grads = tape.gradient(hugectr_forward, bp_trigger)
tf_opt = tf.keras.optimizers.Adam(learning_rate=0.1,
beta_1=0.9,
beta_2=0.99,
epsilon=1e-5)
tf_grads = tape.gradient(tf_forward,
tf_embedding_layer.trainable_weights)
tf_opt.apply_gradients(
zip(tf_grads, tf_embedding_layer.trainable_weights))
# compare second forward result
hugectr_forward_2 = hugectr_tf_ops.fprop_v4(
embedding_name=embedding_name,
row_indices=row_indices,
values=values,
bp_trigger=bp_trigger,
is_training=True,
output_shape=[batch_size, slot_num, max_nnz])
tf_forward_2 = tf_embedding_layer(
sparse_tensor,
output_shape=[batch_size, slot_num, embedding_vec_size])
# print("hugectr 2:\n", hugectr_forward_2)
# print("tf 2:\n", tf_forward_2)
try:
tf.debugging.assert_near(hugectr_forward_2,
tf_forward_2,
rtol=1e-4,
atol=1e-5)
except tf.errors.InvalidArgumentError as error:
raise error
print(
"[INFO]: The results from HugeCTR and tf in the second forward propagation are the same."
)
hugectr_tf_ops.reset()