def get_profile_device_list():
"""
Get profile device list.
Returns:
list, the available device list.
Raises:
ParamValueError: If the search condition contains some errors.
Examples:
>>> POST http://xxxx/v1/mindinsight/profile/devices
"""
profiler_dir = get_profiler_dir(request)
train_id = get_train_id(request)
if not profiler_dir or not train_id:
raise ParamValueError("No profiler_dir or train_id.")
profiler_dir_abs = os.path.join(settings.SUMMARY_BASE_DIR, train_id,
profiler_dir)
try:
profiler_dir_abs = validate_and_normalize_path(profiler_dir_abs,
"profiler")
except ValidationError:
raise ParamValueError("Invalid profiler dir")
check_train_job_and_profiler_dir(profiler_dir_abs)
device_list, _ = analyse_device_list_from_profiler_dir(profiler_dir_abs)
return jsonify(device_list)
def get_cluster_flops():
"""
Get cluster FLOPs.
Returns:
str, the cluster FLOPs.
Raises:
ParamValueError: If the cluster profiler dir is invalid.
Examples:
>>>GET http://xxx/v1/mindinsight/profile/cluster-flops
"""
train_id = get_train_id(request)
if not train_id:
raise ParamValueError('No train id.')
cluster_profiler_dir = os.path.join(settings.SUMMARY_BASE_DIR, train_id)
cluster_profiler_dir = validate_and_normalize_path(cluster_profiler_dir,
'cluster_profiler')
check_train_job_and_profiler_dir(cluster_profiler_dir)
analyser = AnalyserFactory.instance().get_analyser('cluster_flops',
cluster_profiler_dir)
flops = analyser.get_flops()
return jsonify(flops)
def query_image():
"""
Query image.
Returns:
bytes, image binary content for UI to demonstrate.
"""
train_id = get_train_id(request)
if train_id is None:
raise ParamMissError("train_id")
image_path = request.args.get("path")
if image_path is None:
raise ParamMissError("path")
image_type = request.args.get("type")
if image_type is None:
raise ParamMissError("type")
if image_type not in ("original", "overlay", "outcome"):
raise ParamValueError(
f"type:{image_type}, valid options: 'original' 'overlay' 'outcome'"
)
encapsulator = DatafileEncap(EXPLAIN_MANAGER)
image = encapsulator.query_image_binary(train_id, image_path, image_type)
return image
def get_cluster_link_info():
"""
Get cluster link info.
Returns:
Response, the cluster link info.
Raises:
ParamValueError: If the search condition contains some errors.
Examples:
>>>POST http://xxx/v1/mindinsight/profile/search-cluster-link
"""
train_id = get_train_id(request)
cluster_profiler_dir = os.path.join(settings.SUMMARY_BASE_DIR, train_id)
try:
cluster_profiler_dir = validate_and_normalize_path(
cluster_profiler_dir, 'cluster_profiler')
except ValidationError:
raise ParamValueError('Invalid cluster_profiler dir')
condition = request.stream.read()
try:
condition = json.loads(condition) if condition else {}
except (json.JSONDecodeError, ValueError):
raise ParamValueError("Json data parse failed.")
device_id = condition.get("device_id", "0")
to_int(device_id, 'device_id')
analyser = AnalyserFactory.instance().get_analyser('cluster_hccl',
cluster_profiler_dir,
device_id)
link_info = analyser.get_cluster_link_info(condition)
return jsonify(link_info)
def query_single_train_task():
"""Query single train task"""
plugin_name = request.args.get('plugin_name')
train_id = get_train_id(request)
processor = TrainTaskManager(DATA_MANAGER)
tasks = processor.get_single_train_task(train_id=train_id, plugin_name=plugin_name)
return jsonify(tasks)
def query_explain_job():
"""Query explain job meta-data."""
train_id = get_train_id(request)
if train_id is None:
raise ParamMissError("train_id")
encapsulator = ExplainJobEncap(EXPLAIN_MANAGER)
metadata = encapsulator.query_meta(train_id)
return jsonify(metadata)
def query_plugins():
"""Query plugins."""
train_id = get_train_id(request)
manual_update = request.args.get('manual_update', default='false')
manual_update = str_to_bool(manual_update, "manual_update")
processor = TrainTaskManager(DATA_MANAGER)
plugins = processor.get_plugins(train_id, manual_update)
return jsonify(plugins)
def query_evaluation():
"""Query saliency explainer evaluation scores."""
train_id = get_train_id(request)
if train_id is None:
raise ParamMissError("train_id")
encapsulator = EvaluationEncap(EXPLAIN_MANAGER)
scores = encapsulator.query_explainer_scores(train_id)
return jsonify({
"explainer_scores": scores,
})
def histogram():
"""
Interface to obtain histogram data.
Returns:
Response, which contains a JSON object.
"""
tag = request.args.get("tag", default=None)
train_id = get_train_id(request)
processor = HistogramProcessor(DATA_MANAGER)
response = processor.get_histograms(train_id, tag)
return jsonify(response)
def single_image():
"""
Interface to fetch raw image data for a particular image.
Returns:
Response, which contains a byte string of image.
"""
tag = request.args.get("tag")
step = request.args.get("step")
train_id = get_train_id(request)
processor = ImageProcessor(DATA_MANAGER)
img_data = processor.get_single_image(train_id, tag, step)
return img_data
def graph_search_single_node():
"""
Interface to search single node.
Returns:
Response, which contains a JSON object.
"""
name = request.args.get("name")
tag = request.args.get("tag", default=None)
train_id = get_train_id(request)
graph_process = GraphProcessor(train_id, DATA_MANAGER, tag)
resp = graph_process.search_single_node(name)
return jsonify(resp)
def graph_nodes():
"""
Interface to get graph nodes.
Returns:
Response, which contains a JSON object.
"""
name = request.args.get('name', default=None)
tag = request.args.get("tag", default=None)
train_id = get_train_id(request)
graph_process = GraphProcessor(train_id, DATA_MANAGER, tag)
response = graph_process.list_nodes(scope=name)
return jsonify(response)
def image_metadata():
"""
Interface to fetch metadata about the images for the particular run,tag, and zero-indexed sample.
Returns:
Response, which contains a list in JSON containing image events, each
one of which is an object containing items wall_time, step, width,
height, and query.
"""
tag = request.args.get("tag")
train_id = get_train_id(request)
processor = ImageProcessor(DATA_MANAGER)
response = processor.get_metadata_list(train_id, tag)
return jsonify(response)
def get_profile_summary_proposal():
"""
Get summary profiling proposal.
Returns:
str, the summary profiling proposal.
Raises:
ParamValueError: If the parameters contain some errors.
Examples:
>>> GET http://xxxx/v1/mindinsight/profile/summary/propose
"""
profiler_dir = get_profiler_dir(request)
train_id = get_train_id(request)
device_id = get_device_id(request)
if not profiler_dir or not train_id:
raise ParamValueError("No profiler_dir or train_id.")
to_int(device_id, 'device_id')
profiler_dir_abs = os.path.join(settings.SUMMARY_BASE_DIR, train_id,
profiler_dir)
try:
profiler_dir_abs = validate_and_normalize_path(profiler_dir_abs,
"profiler")
except ValidationError:
raise ParamValueError("Invalid profiler dir")
check_train_job_and_profiler_dir(profiler_dir_abs)
step_trace_condition = {
"filter_condition": {
"mode": "proc",
"proc_name": "iteration_interval",
"step_id": 0
}
}
options = {'step_trace': {"iter_interval": step_trace_condition}}
proposal_type_list = [
'step_trace', 'minddata', 'minddata_pipeline', 'common'
]
proposal_obj = ComposeProposal(profiler_dir_abs, device_id,
proposal_type_list)
proposal_info = proposal_obj.get_proposal(options)
# Use json.dumps for orderly return
return CustomResponse(json.dumps(proposal_info),
mimetype='application/json')
def graph_node_names():
"""
Interface to query node names.
Returns:
Response, which contains a JSON object.
"""
search_content = request.args.get("search")
offset = request.args.get("offset", default=0)
limit = request.args.get("limit", default=100)
tag = request.args.get("tag", default=None)
train_id = get_train_id(request)
graph_process = GraphProcessor(train_id, DATA_MANAGER, tag)
resp = graph_process.search_node_names(search_content, offset, limit)
return jsonify(resp)
def query_explain_job():
"""
Query explain job meta-data.
Returns:
Response, contains dict that stores metadata of the requested job.
Raises:
ParamMissError: If train_id info is not in the request.
"""
train_id = get_train_id(request)
if train_id is None:
raise ParamMissError("train_id")
encapsulator = ExplainJobEncap(EXPLAIN_MANAGER)
metadata = encapsulator.query_meta(train_id)
return jsonify(metadata)
def get_profile_op_info():
"""
Get operation profiling info.
Returns:
str, the operation profiling information.
Raises:
ParamValueError: If the search condition contains some errors.
Examples:
>>> POST http://xxxx/v1/mindinsight/profile/ops/search
"""
profiler_dir = get_profiler_dir(request)
train_id = get_train_id(request)
if not profiler_dir or not train_id:
raise ParamValueError("No profiler_dir or train_id.")
search_condition = request.stream.read()
try:
search_condition = json.loads(
search_condition if search_condition else "{}")
except (json.JSONDecodeError, ValueError):
raise ParamValueError("Json data parse failed.")
validate_condition(search_condition)
device_id = search_condition.get("device_id", "0")
to_int(device_id, 'device_id')
profiler_dir_abs = os.path.join(settings.SUMMARY_BASE_DIR, train_id,
profiler_dir)
try:
profiler_dir_abs = validate_and_normalize_path(profiler_dir_abs,
"profiler")
except ValidationError:
raise ParamValueError("Invalid profiler dir")
check_train_job_and_profiler_dir(profiler_dir_abs)
op_type = search_condition.get("op_type")
analyser = AnalyserFactory.instance().get_analyser(op_type,
profiler_dir_abs,
device_id)
op_info = analyser.query(search_condition)
return jsonify(op_info)
def query_evaluation():
"""
Query saliency explainer evaluation scores.
Returns:
Response, contains dict that stores evaluation scores.
Raises:
ParamMissError: If train_id info is not in the request.
"""
train_id = get_train_id(request)
if train_id is None:
raise ParamMissError("train_id")
encapsulator = EvaluationEncap(EXPLAIN_MANAGER)
scores = encapsulator.query_explainer_scores(train_id)
return jsonify({
"explainer_scores": scores,
})
def get_dataset_graph():
"""
Get dataset graph.
Returns:
str, the dataset graph information.
Raises:
MindInsightException: If method fails to be called.
ParamValueError: If summary_dir is invalid.
Examples:
>>> GET http://xxxx/v1/mindinsight/datasets/dataset_graph?train_id=xxx
"""
summary_base_dir = str(settings.SUMMARY_BASE_DIR)
summary_dir = get_train_id(request)
if summary_dir.startswith('/'):
validate_path(summary_dir)
elif summary_dir.startswith('./'):
summary_dir = os.path.join(summary_base_dir, summary_dir[2:])
summary_dir = validate_path(summary_dir)
else:
raise ParamValueError("Summary dir should be absolute path or "
"relative path that relate to summary base dir.")
try:
dataset_graph = get_summary_lineage(summary_dir=summary_dir,
keys=['dataset_graph'])
except MindInsightException as exception:
raise MindInsightException(exception.error,
exception.message,
http_code=400)
if dataset_graph:
summary_dir_result = dataset_graph.get('summary_dir')
base_dir_len = len(summary_base_dir)
if summary_base_dir == summary_dir_result:
relative_dir = './'
else:
relative_dir = os.path.join(os.curdir,
summary_dir[base_dir_len + 1:])
dataset_graph['summary_dir'] = relative_dir
return jsonify(dataset_graph)
def scalar_metadata():
"""
Interface to fetch metadata about the scalars for the particular run and tag.
Returns:
Response, which contains a list in JSON containing scalar events, each
one of which is an object containing items' wall_time, step and value.
"""
tag = request.args.get("tag")
train_id = get_train_id(request)
processor = ScalarsProcessor(DATA_MANAGER)
response = processor.get_metadata_list(train_id, tag)
metadatas = response['metadatas']
for metadata in metadatas:
value = metadata.get("value")
metadata["value"] = if_nan_inf_to_none('scalar_value', value)
return jsonify(response)
def get_minddata_pipeline_op_queue_info():
"""
Get minddata pipeline operator info and queue info.
Returns:
str, the operation information and queue information.
Raises:
ParamValueError: If the search condition contains some errors.
Examples:
>>> POST http://xxxx/v1/mindinsight/profile/minddata-pipeline/op-queue
"""
profiler_dir = get_profiler_dir(request)
train_id = get_train_id(request)
if not profiler_dir or not train_id:
raise ParamValueError("No profiler_dir or train_id.")
profiler_dir_abs = os.path.join(settings.SUMMARY_BASE_DIR, train_id,
profiler_dir)
try:
profiler_dir_abs = validate_and_normalize_path(profiler_dir_abs,
"profiler")
except ValidationError:
raise ParamValueError("Invalid profiler dir.")
check_train_job_and_profiler_dir(profiler_dir_abs)
condition = request.stream.read()
try:
condition = json.loads(condition) if condition else {}
except Exception:
raise ParamValueError("Json data parse failed.")
validate_minddata_pipeline_condition(condition)
device_id = condition.get("device_id", "0")
to_int(device_id, 'device_id')
analyser = AnalyserFactory.instance().get_analyser('minddata_pipeline',
profiler_dir_abs,
device_id)
op_info = analyser.query(condition)
return jsonify(op_info)
def update_lineage():
"""
Get lineage.
Returns:
str, update the lineage information about cache and tag.
Raises:
MindInsightException: If method fails to be called.
Examples:
>>> PUT http://xxxx/v1/mindinsight/lineagemgr/lineages?train_id=./run1
"""
train_id = get_train_id(request)
added_info = request.json
if not isinstance(added_info, dict):
raise ParamValueError("The request body should be a dict.")
update_lineage_object(DATA_MANAGER, train_id, added_info)
return jsonify({"status": "success"})
def query_train_jobs():
"""Query train jobs."""
offset = request.args.get("offset", default=0)
limit = request.args.get("limit", default=10)
train_id = get_train_id(request)
offset = Validation.check_offset(offset=offset)
limit = Validation.check_limit(
limit, min_value=1, max_value=SummaryWatcher.MAX_SUMMARY_DIR_COUNT)
processor = TrainTaskManager(DATA_MANAGER)
total, train_jobs = processor.query_train_jobs(offset, limit, train_id)
return jsonify({
'name':
os.path.basename(os.path.realpath(settings.SUMMARY_BASE_DIR)),
'total':
total,
'train_jobs':
train_jobs,
})
def get_minddata_cpu_utilization_info():
"""
Get minddata cpu utilization info.
Returns:
str, the minddata cpu utilization info.
Raises:
ParamValueError: If the search condition contains some errors.
Examples:
>>>POST http://xxx/v1/mindinsight/profile/minddata-cpu-utilization-summary
"""
profiler_dir = get_profiler_dir(request)
train_id = get_train_id(request)
if not profiler_dir or not train_id:
raise ParamValueError("No profiler_dir or train_id.")
profiler_dir_abs = os.path.join(settings.SUMMARY_BASE_DIR, train_id,
profiler_dir)
try:
profiler_dir_abs = validate_and_normalize_path(profiler_dir_abs,
"profiler")
except ValidationError:
raise ParamValueError("Invalid profiler dir.")
check_train_job_and_profiler_dir(profiler_dir_abs)
condition = request.stream.read()
try:
condition = json.loads(condition) if condition else {}
except (json.JSONDecodeError, ValueError):
raise ParamValueError("Json data parse failed.")
device_id = condition.get("device_id", "0")
to_int(device_id, 'device_id')
analyser = AnalyserFactory.instance().get_analyser(
'minddata_cpu_utilization', profiler_dir_abs, device_id)
cpu_utilization = analyser.query(condition)
return jsonify(cpu_utilization)
def get_minddata_pipeline_queue_info():
"""
Get the special minddata pipeline queue info.
Returns:
str, the queue information.
Raises:
ParamValueError: If the search condition contains some errors.
Examples:
>>> GET http://xxxx/v1/mindinsight/profile/minddata-pipeline/queue
"""
profiler_dir = get_profiler_dir(request)
train_id = get_train_id(request)
if not profiler_dir or not train_id:
raise ParamValueError("No profiler_dir or train_id.")
profiler_dir_abs = os.path.join(settings.SUMMARY_BASE_DIR, train_id,
profiler_dir)
try:
profiler_dir_abs = validate_and_normalize_path(profiler_dir_abs,
"profiler")
except ValidationError:
raise ParamValueError("Invalid profiler dir.")
check_train_job_and_profiler_dir(profiler_dir_abs)
device_id = request.args.get('device_id', default='0')
to_int(device_id, 'device_id')
op_id = request.args.get('op_id', type=int)
if op_id is None:
raise ParamValueError(
"Invalid operator id or operator id does not exist.")
analyser = AnalyserFactory.instance().get_analyser('minddata_pipeline',
profiler_dir_abs,
device_id)
op_queue_info = analyser.get_op_and_parent_op_info(op_id)
return jsonify(op_queue_info)
def get_dataset_graph():
"""
Get dataset graph.
Returns:
str, the dataset graph information.
Raises:
MindInsightException: If method fails to be called.
ParamValueError: If summary_dir is invalid.
Examples:
>>> GET http://xxxx/v1/mindinsight/datasets/dataset_graph?train_id=xxx
"""
train_id = get_train_id(request)
validate_train_id(train_id)
search_condition = {
'summary_dir': {
'in': [train_id]
}
}
result = {}
try:
objects = filter_summary_lineage(data_manager=DATA_MANAGER, search_condition=search_condition).get('object')
except MindInsightException as exception:
raise MindInsightException(exception.error, exception.message, http_code=400)
if objects:
lineage_obj = objects[0]
dataset_graph = lineage_obj.get('dataset_graph')
if dataset_graph:
result.update({'dataset_graph': dataset_graph})
result.update({'summary_dir': lineage_obj.get('summary_dir')})
return jsonify(result)
def get_dataset_graph():
"""
Get dataset graph.
Returns:
str, the dataset graph information.
Raises:
MindInsightException: If method fails to be called.
ParamValueError: If summary_dir is invalid.
Examples:
>>> GET http://xxxx/v1/mindinsight/datasets/dataset_graph?train_id=xxx
"""
summary_base_dir = str(settings.SUMMARY_BASE_DIR)
summary_dir = get_train_id(request)
try:
dataset_graph = general_get_summary_lineage(DATA_MANAGER,
summary_dir=summary_dir,
keys=['dataset_graph'])
except MindInsightException as exception:
raise MindInsightException(exception.error,
exception.message,
http_code=400)
if dataset_graph:
summary_dir_result = dataset_graph.get('summary_dir')
base_dir_len = len(summary_base_dir)
if summary_base_dir == summary_dir_result:
relative_dir = './'
else:
relative_dir = os.path.join(os.curdir,
summary_dir[base_dir_len + 1:])
dataset_graph['summary_dir'] = relative_dir
return jsonify(dataset_graph)
def get_profiler_abs_dir(requests):
"""
Get interval process summary.
Args:
requests (LocalProxy): The requests.
Returns:
str, the profiler abs dir.
"""
profiler_dir = get_profiler_dir(requests)
train_id = get_train_id(requests)
if not profiler_dir or not train_id:
raise ParamValueError("No profiler_dir or train_id.")
profiler_dir_abs = os.path.join(settings.SUMMARY_BASE_DIR, train_id,
profiler_dir)
try:
profiler_dir_abs = validate_and_normalize_path(profiler_dir_abs,
"profiler")
except ValidationError:
raise ParamValueError("Invalid profiler dir")
return profiler_dir_abs
