clarifai 11.1.5rc6__py3-none-any.whl → 11.1.5rc8__py3-none-any.whl
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- clarifai/__init__.py +1 -1
- clarifai/cli/__pycache__/model.cpython-310.pyc +0 -0
- clarifai/client/#model_client.py# +430 -0
- clarifai/client/model.py +95 -61
- clarifai/client/model_client.py +64 -49
- clarifai/runners/__pycache__/__init__.cpython-310.pyc +0 -0
- clarifai/runners/models/__pycache__/base_typed_model.cpython-310.pyc +0 -0
- clarifai/runners/models/__pycache__/model_builder.cpython-310.pyc +0 -0
- clarifai/runners/models/__pycache__/model_class.cpython-310.pyc +0 -0
- clarifai/runners/models/__pycache__/model_runner.cpython-310.pyc +0 -0
- clarifai/runners/models/model_class.py +31 -48
- clarifai/runners/utils/__pycache__/data_handler.cpython-310.pyc +0 -0
- clarifai/runners/utils/__pycache__/data_types.cpython-310.pyc +0 -0
- clarifai/runners/utils/__pycache__/method_signatures.cpython-310.pyc +0 -0
- clarifai/runners/utils/__pycache__/serializers.cpython-310.pyc +0 -0
- clarifai/runners/utils/data_types.py +62 -10
- clarifai/runners/utils/method_signatures.py +278 -295
- clarifai/runners/utils/serializers.py +143 -67
- {clarifai-11.1.5rc6.dist-info → clarifai-11.1.5rc8.dist-info}/METADATA +1 -1
- {clarifai-11.1.5rc6.dist-info → clarifai-11.1.5rc8.dist-info}/RECORD +24 -23
- {clarifai-11.1.5rc6.dist-info → clarifai-11.1.5rc8.dist-info}/LICENSE +0 -0
- {clarifai-11.1.5rc6.dist-info → clarifai-11.1.5rc8.dist-info}/WHEEL +0 -0
- {clarifai-11.1.5rc6.dist-info → clarifai-11.1.5rc8.dist-info}/entry_points.txt +0 -0
- {clarifai-11.1.5rc6.dist-info → clarifai-11.1.5rc8.dist-info}/top_level.txt +0 -0
clarifai/__init__.py
CHANGED
@@ -1 +1 @@
|
|
1
|
-
__version__ = "11.1.
|
1
|
+
__version__ = "11.1.5rc8"
|
Binary file
|
@@ -0,0 +1,430 @@
|
|
1
|
+
import inspect
|
2
|
+
import time
|
3
|
+
from typing import Any, Dict, Iterator, List
|
4
|
+
|
5
|
+
from clarifai_grpc.grpc.api import resources_pb2, service_pb2
|
6
|
+
from clarifai_grpc.grpc.api.status import status_code_pb2
|
7
|
+
|
8
|
+
from clarifai.constants.model import MAX_MODEL_PREDICT_INPUTS
|
9
|
+
from clarifai.errors import UserError
|
10
|
+
from clarifai.runners.utils.method_signatures import (deserialize, get_stream_from_signature,
|
11
|
+
serialize, signatures_from_json,
|
12
|
+
unflatten_nested_keys)
|
13
|
+
from clarifai.utils.misc import BackoffIterator, status_is_retryable
|
14
|
+
|
15
|
+
from clarifai.utils.logging import logger
|
16
|
+
|
17
|
+
|
18
|
+
class ModelClient:
|
19
|
+
'''
|
20
|
+
Client for calling model predict, generate, and stream methods.
|
21
|
+
'''
|
22
|
+
|
23
|
+
def __init__(self, stub, request_template: service_pb2.PostModelOutputsRequest = None):
|
24
|
+
'''
|
25
|
+
Initialize the model client.
|
26
|
+
|
27
|
+
Args:
|
28
|
+
stub: The gRPC stub for the model.
|
29
|
+
request_template: The template for the request to send to the model, including
|
30
|
+
common fields like model_id, model_version, cluster, etc.
|
31
|
+
'''
|
32
|
+
self.STUB = stub
|
33
|
+
self.request_template = request_template or service_pb2.PostModelOutputsRequest()
|
34
|
+
self._fetch_signatures()
|
35
|
+
self._define_functions()
|
36
|
+
|
37
|
+
def _fetch_signatures(self):
|
38
|
+
'''
|
39
|
+
Fetch the method signatures from the model.
|
40
|
+
|
41
|
+
Returns:
|
42
|
+
Dict: The method signatures.
|
43
|
+
'''
|
44
|
+
#request = resources_pb2.GetModelSignaturesRequest()
|
45
|
+
#response = self.stub.GetModelSignatures(request)
|
46
|
+
#self._method_signatures = json.loads(response.signatures) # or define protos
|
47
|
+
# TODO this could use a new endpoint to get the signatures
|
48
|
+
# for local grpc models, we'll also have to add the endpoint to the model servicer
|
49
|
+
# for now we'll just use the predict endpoint with a special method name
|
50
|
+
|
51
|
+
request = service_pb2.PostModelOutputsRequest()
|
52
|
+
request.CopyFrom(self.request_template)
|
53
|
+
# request.model.model_version.output_info.params['_method_name'] = '_GET_SIGNATURES'
|
54
|
+
inp = request.inputs.add() # empty input for this method
|
55
|
+
inp.data.parts.add() # empty part for this input
|
56
|
+
inp.data.metadata['_method_name'] = '_GET_SIGNATURES'
|
57
|
+
start_time = time.time()
|
58
|
+
backoff_iterator = BackoffIterator(10)
|
59
|
+
while True:
|
60
|
+
response = self.STUB.PostModelOutputs(request)
|
61
|
+
if status_is_retryable(
|
62
|
+
response.status.code) and time.time() - start_time < 60 * 10: # 10 minutes
|
63
|
+
self.logger.info(f"Retrying model info fetch with response {response.status!r}")
|
64
|
+
time.sleep(next(backoff_iterator))
|
65
|
+
continue
|
66
|
+
break
|
67
|
+
if response.status.code == status_code_pb2.INPUT_UNSUPPORTED_FORMAT:
|
68
|
+
# return code from older models that don't support _GET_SIGNATURES
|
69
|
+
self._method_signatures = {}
|
70
|
+
return
|
71
|
+
if response.status.code != status_code_pb2.SUCCESS:
|
72
|
+
raise Exception(f"Model failed with response {response!r}")
|
73
|
+
self._method_signatures = signatures_from_json(response.outputs[0].data.text.raw)
|
74
|
+
|
75
|
+
def _define_functions(self):
|
76
|
+
'''
|
77
|
+
Define the functions based on the method signatures.
|
78
|
+
'''
|
79
|
+
for method_name, method_signature in self._method_signatures.items():
|
80
|
+
# define the function in this client instance
|
81
|
+
if method_signature.method_type == 'predict':
|
82
|
+
call_func = self._predict
|
83
|
+
elif method_signature.method_type == 'generate':
|
84
|
+
call_func = self._generate
|
85
|
+
elif method_signature.method_type == 'stream':
|
86
|
+
call_func = self._stream
|
87
|
+
else:
|
88
|
+
raise ValueError(f"Unknown method type {method_signature.method_type}")
|
89
|
+
|
90
|
+
# method argnames, in order, collapsing nested keys to corresponding user function args
|
91
|
+
method_argnames = []
|
92
|
+
for var in method_signature.inputs:
|
93
|
+
outer = var.name.split('.', 1)[0]
|
94
|
+
if outer in method_argnames:
|
95
|
+
continue
|
96
|
+
method_argnames.append(outer)
|
97
|
+
|
98
|
+
def bind_f(method_name, method_argnames, call_func):
|
99
|
+
|
100
|
+
def f(*args, **kwargs):
|
101
|
+
if len(args) > len(method_argnames):
|
102
|
+
raise TypeError(
|
103
|
+
f"{method_name}() takes {len(method_argnames)} positional arguments but {len(args)} were given"
|
104
|
+
)
|
105
|
+
for name, arg in zip(method_argnames, args): # handle positional with zip shortest
|
106
|
+
if name in kwargs:
|
107
|
+
raise TypeError(f"Multiple values for argument {name}")
|
108
|
+
kwargs[name] = arg
|
109
|
+
return call_func(kwargs, method_name)
|
110
|
+
|
111
|
+
return f
|
112
|
+
|
113
|
+
# need to bind method_name to the value, not the mutating loop variable
|
114
|
+
f = bind_f(method_name, method_argnames, call_func)
|
115
|
+
|
116
|
+
# set names, annotations and docstrings
|
117
|
+
f.__name__ = method_name
|
118
|
+
f.__qualname__ = f'{self.__class__.__name__}.{method_name}'
|
119
|
+
input_annos = {var.name: var.data_type for var in method_signature.inputs}
|
120
|
+
output_annos = {var.name: var.data_type for var in method_signature.outputs}
|
121
|
+
# unflatten nested keys to match the user function args for docs
|
122
|
+
input_annos = unflatten_nested_keys(input_annos, method_signature.inputs, is_output=False)
|
123
|
+
output_annos = unflatten_nested_keys(output_annos, method_signature.outputs, is_output=True)
|
124
|
+
|
125
|
+
# add Stream[] to the stream input annotations for docs
|
126
|
+
input_stream_argname, _ = get_stream_from_signature(method_signature.inputs)
|
127
|
+
if input_stream_argname:
|
128
|
+
input_annos[input_stream_argname] = 'Stream[' + str(
|
129
|
+
input_annos[input_stream_argname]) + ']'
|
130
|
+
|
131
|
+
# handle multiple outputs in the return annotation
|
132
|
+
return_annotation = output_annos
|
133
|
+
name = next(iter(output_annos.keys()))
|
134
|
+
if len(output_annos) == 1 and name == 'return':
|
135
|
+
# single output
|
136
|
+
return_annotation = output_annos[name]
|
137
|
+
elif name.startswith('return.') and name.split('.', 1)[1].isnumeric():
|
138
|
+
# tuple output
|
139
|
+
return_annotation = '(' + ", ".join(output_annos[f'return.{i}']
|
140
|
+
for i in range(len(output_annos))) + ')'
|
141
|
+
else:
|
142
|
+
# named output
|
143
|
+
return_annotation = f'Output({", ".join(f"{k}={t}" for k, t in output_annos.items())})'
|
144
|
+
if method_signature.method_type in ['generate', 'stream']:
|
145
|
+
return_annotation = f'Stream[{return_annotation}]'
|
146
|
+
|
147
|
+
# set annotations and docstrings
|
148
|
+
sig = inspect.signature(f).replace(
|
149
|
+
parameters=[
|
150
|
+
inspect.Parameter(k, inspect.Parameter.POSITIONAL_OR_KEYWORD, annotation=v)
|
151
|
+
for k, v in input_annos.items()
|
152
|
+
],
|
153
|
+
return_annotation=return_annotation,
|
154
|
+
)
|
155
|
+
f.__signature__ = sig
|
156
|
+
f.__doc__ = method_signature.docstring
|
157
|
+
setattr(self, method_name, f)
|
158
|
+
|
159
|
+
def _predict(
|
160
|
+
self,
|
161
|
+
inputs, # TODO set up functions according to fetched signatures?
|
162
|
+
method_name: str = 'predict',
|
163
|
+
) -> Any:
|
164
|
+
input_signature = self._method_signatures[method_name].inputs
|
165
|
+
output_signature = self._method_signatures[method_name].outputs
|
166
|
+
|
167
|
+
batch_input = True
|
168
|
+
if isinstance(inputs, dict):
|
169
|
+
inputs = [inputs]
|
170
|
+
batch_input = False
|
171
|
+
|
172
|
+
proto_inputs = []
|
173
|
+
for input in inputs:
|
174
|
+
proto = resources_pb2.Input()
|
175
|
+
serialize(input, input_signature, proto.data)
|
176
|
+
proto_inputs.append(proto)
|
177
|
+
|
178
|
+
response = self._predict_by_proto(proto_inputs, method_name)
|
179
|
+
#print(response)
|
180
|
+
|
181
|
+
outputs = []
|
182
|
+
for output in response.outputs:
|
183
|
+
outputs.append(deserialize(output.data, output_signature, is_output=True))
|
184
|
+
if batch_input:
|
185
|
+
return outputs
|
186
|
+
return outputs[0]
|
187
|
+
|
188
|
+
def _predict_by_proto(
|
189
|
+
self,
|
190
|
+
inputs: List[resources_pb2.Input],
|
191
|
+
method_name: str = None,
|
192
|
+
inference_params: Dict = None,
|
193
|
+
output_config: Dict = None,
|
194
|
+
) -> service_pb2.MultiOutputResponse:
|
195
|
+
"""Predicts the model based on the given inputs.
|
196
|
+
|
197
|
+
Args:
|
198
|
+
inputs (List[resources_pb2.Input]): The inputs to predict.
|
199
|
+
method_name (str): The remote method name to call.
|
200
|
+
inference_params (Dict): Inference parameters to override.
|
201
|
+
output_config (Dict): Output configuration to override.
|
202
|
+
|
203
|
+
Returns:
|
204
|
+
service_pb2.MultiOutputResponse: The prediction response(s).
|
205
|
+
"""
|
206
|
+
if not isinstance(inputs, list):
|
207
|
+
raise UserError('Invalid inputs, inputs must be a list of Input objects.')
|
208
|
+
if len(inputs) > MAX_MODEL_PREDICT_INPUTS:
|
209
|
+
raise UserError(f"Too many inputs. Max is {MAX_MODEL_PREDICT_INPUTS}.")
|
210
|
+
|
211
|
+
request = service_pb2.PostModelOutputsRequest()
|
212
|
+
request.CopyFrom(self.request_template)
|
213
|
+
|
214
|
+
request.inputs.extend(inputs)
|
215
|
+
|
216
|
+
if method_name:
|
217
|
+
# TODO put in new proto field?
|
218
|
+
for inp in request.inputs:
|
219
|
+
inp.data.metadata['_method_name'] = method_name
|
220
|
+
if inference_params:
|
221
|
+
request.model.model_version.output_info.params.update(inference_params)
|
222
|
+
if output_config:
|
223
|
+
request.model.model_version.output_info.output_config.MergeFrom(
|
224
|
+
resources_pb2.OutputConfig(**output_config))
|
225
|
+
|
226
|
+
start_time = time.time()
|
227
|
+
backoff_iterator = BackoffIterator(10)
|
228
|
+
while True:
|
229
|
+
response = self.STUB.PostModelOutputs(request)
|
230
|
+
if status_is_retryable(
|
231
|
+
response.status.code) and time.time() - start_time < 60 * 10: # 10 minutes
|
232
|
+
self.logger.info(f"Model predict failed with response {response!r}")
|
233
|
+
time.sleep(next(backoff_iterator))
|
234
|
+
continue
|
235
|
+
|
236
|
+
if response.status.code != status_code_pb2.SUCCESS:
|
237
|
+
raise Exception(f"Model predict failed with response {response!r}")
|
238
|
+
break
|
239
|
+
|
240
|
+
return response
|
241
|
+
|
242
|
+
def _generate(
|
243
|
+
self,
|
244
|
+
inputs, # TODO set up functions according to fetched signatures?
|
245
|
+
method_name: str = 'generate',
|
246
|
+
) -> Any:
|
247
|
+
input_signature = self._method_signatures[method_name].inputs
|
248
|
+
output_signature = self._method_signatures[method_name].outputs
|
249
|
+
|
250
|
+
batch_input = True
|
251
|
+
if isinstance(inputs, dict):
|
252
|
+
inputs = [inputs]
|
253
|
+
batch_input = False
|
254
|
+
|
255
|
+
proto_inputs = []
|
256
|
+
for input in inputs:
|
257
|
+
proto = resources_pb2.Input()
|
258
|
+
serialize(input, input_signature, proto.data)
|
259
|
+
proto_inputs.append(proto)
|
260
|
+
|
261
|
+
response_stream = self._generate_by_proto(proto_inputs, method_name)
|
262
|
+
#print(response)
|
263
|
+
|
264
|
+
for response in response_stream:
|
265
|
+
outputs = []
|
266
|
+
for output in response.outputs:
|
267
|
+
outputs.append(deserialize(output.data, output_signature, is_output=True))
|
268
|
+
if batch_input:
|
269
|
+
yield outputs
|
270
|
+
yield outputs[0]
|
271
|
+
|
272
|
+
def _generate_by_proto(
|
273
|
+
self,
|
274
|
+
inputs: List[resources_pb2.Input],
|
275
|
+
method_name: str = None,
|
276
|
+
inference_params: Dict = {},
|
277
|
+
output_config: Dict = {},
|
278
|
+
):
|
279
|
+
"""Generate the stream output on model based on the given inputs.
|
280
|
+
|
281
|
+
Args:
|
282
|
+
inputs (list[Input]): The inputs to generate, must be less than 128.
|
283
|
+
method_name (str): The remote method name to call.
|
284
|
+
inference_params (dict): The inference params to override.
|
285
|
+
output_config (dict): The output config to override.
|
286
|
+
"""
|
287
|
+
if not isinstance(inputs, list):
|
288
|
+
raise UserError('Invalid inputs, inputs must be a list of Input objects.')
|
289
|
+
if len(inputs) > MAX_MODEL_PREDICT_INPUTS:
|
290
|
+
raise UserError(f"Too many inputs. Max is {MAX_MODEL_PREDICT_INPUTS}."
|
291
|
+
) # TODO Use Chunker for inputs len > 128
|
292
|
+
|
293
|
+
request = service_pb2.PostModelOutputsRequest()
|
294
|
+
request.CopyFrom(self.request_template)
|
295
|
+
|
296
|
+
request.inputs.extend(inputs)
|
297
|
+
|
298
|
+
if method_name:
|
299
|
+
# TODO put in new proto field?
|
300
|
+
for inp in request.inputs:
|
301
|
+
inp.data.metadata['_method_name'] = method_name
|
302
|
+
if inference_params:
|
303
|
+
request.model.model_version.output_info.params.update(inference_params)
|
304
|
+
if output_config:
|
305
|
+
request.model.model_version.output_info.output_config.MergeFromDict(output_config)
|
306
|
+
|
307
|
+
start_time = time.time()
|
308
|
+
backoff_iterator = BackoffIterator(10)
|
309
|
+
started = False
|
310
|
+
while not started:
|
311
|
+
stream_response = self.STUB.GenerateModelOutputs(request)
|
312
|
+
try:
|
313
|
+
response = next(stream_response) # get the first response
|
314
|
+
except StopIteration:
|
315
|
+
raise Exception("Model Generate failed with no response")
|
316
|
+
if status_is_retryable(response.status.code) and \
|
317
|
+
time.time() - start_time < 60 * 10:
|
318
|
+
self.logger.info("Model is still deploying, please wait...")
|
319
|
+
time.sleep(next(backoff_iterator))
|
320
|
+
continue
|
321
|
+
if response.status.code != status_code_pb2.SUCCESS:
|
322
|
+
raise Exception(f"Model Generate failed with response {response.status!r}")
|
323
|
+
started = True
|
324
|
+
|
325
|
+
yield response # yield the first response
|
326
|
+
|
327
|
+
for response in stream_response:
|
328
|
+
if response.status.code != status_code_pb2.SUCCESS:
|
329
|
+
raise Exception(f"Model Generate failed with response {response.status!r}")
|
330
|
+
yield response
|
331
|
+
|
332
|
+
def _stream(
|
333
|
+
self,
|
334
|
+
inputs,
|
335
|
+
method_name: str = 'stream',
|
336
|
+
) -> Any:
|
337
|
+
input_signature = self._method_signatures[method_name].inputs
|
338
|
+
output_signature = self._method_signatures[method_name].outputs
|
339
|
+
|
340
|
+
if isinstance(inputs, list):
|
341
|
+
assert len(inputs) == 1, 'streaming methods do not support batched calls'
|
342
|
+
inputs = inputs[0]
|
343
|
+
assert isinstance(inputs, dict)
|
344
|
+
kwargs = inputs
|
345
|
+
|
346
|
+
# find the streaming vars in the input signature, and the streaming input python param
|
347
|
+
stream_argname, streaming_var_signatures = get_stream_from_signature(input_signature)
|
348
|
+
|
349
|
+
# get the streaming input generator from the user-provided function arg values
|
350
|
+
user_inputs_generator = kwargs.pop(stream_argname)
|
351
|
+
|
352
|
+
def _input_proto_stream():
|
353
|
+
# first item contains all the inputs and the first stream item
|
354
|
+
proto = resources_pb2.Input()
|
355
|
+
try:
|
356
|
+
item = next(user_inputs_generator)
|
357
|
+
except StopIteration:
|
358
|
+
return # no items to stream
|
359
|
+
kwargs[stream_argname] = item
|
360
|
+
serialize(kwargs, input_signature, proto.data)
|
361
|
+
|
362
|
+
yield proto
|
363
|
+
|
364
|
+
# subsequent items are just the stream items
|
365
|
+
for item in user_inputs_generator:
|
366
|
+
proto = resources_pb2.Input()
|
367
|
+
serialize({stream_argname: item}, streaming_var_signatures, proto.data)
|
368
|
+
yield proto
|
369
|
+
|
370
|
+
response_stream = self._stream_by_proto(_input_proto_stream(), method_name)
|
371
|
+
#print(response)
|
372
|
+
|
373
|
+
for response in response_stream:
|
374
|
+
assert len(response.outputs) == 1, 'streaming methods must have exactly one output'
|
375
|
+
yield deserialize(response.outputs[0].data, output_signature, is_output=True)
|
376
|
+
|
377
|
+
def _req_iterator(self,
|
378
|
+
input_iterator: Iterator[List[resources_pb2.Input]],
|
379
|
+
method_name: str = None,
|
380
|
+
inference_params: Dict = {},
|
381
|
+
output_config: Dict = {}):
|
382
|
+
request = service_pb2.PostModelOutputsRequest()
|
383
|
+
request.CopyFrom(self.request_template)
|
384
|
+
if inference_params:
|
385
|
+
request.model.model_version.output_info.params.update(inference_params)
|
386
|
+
if output_config:
|
387
|
+
request.model.model_version.output_info.output_config.MergeFromDict(output_config)
|
388
|
+
for inputs in input_iterator:
|
389
|
+
req = service_pb2.PostModelOutputsRequest()
|
390
|
+
req.CopyFrom(request)
|
391
|
+
if isinstance(inputs, list):
|
392
|
+
req.inputs.extend(inputs)
|
393
|
+
else:
|
394
|
+
req.inputs.append(inputs)
|
395
|
+
# TODO: put into new proto field?
|
396
|
+
for inp in req.inputs:
|
397
|
+
inp.data.metadata['_method_name'] = method_name
|
398
|
+
yield req
|
399
|
+
|
400
|
+
def _stream_by_proto(self,
|
401
|
+
inputs: Iterator[List[resources_pb2.Input]],
|
402
|
+
method_name: str = None,
|
403
|
+
inference_params: Dict = {},
|
404
|
+
output_config: Dict = {}):
|
405
|
+
"""Generate the stream output on model based on the given stream of inputs.
|
406
|
+
"""
|
407
|
+
# if not isinstance(inputs, Iterator[List[Input]]):
|
408
|
+
# raise UserError('Invalid inputs, inputs must be a iterator of list of Input objects.')
|
409
|
+
|
410
|
+
request = self._req_iterator(inputs, method_name, inference_params, output_config)
|
411
|
+
|
412
|
+
start_time = time.time()
|
413
|
+
backoff_iterator = BackoffIterator(10)
|
414
|
+
generation_started = False
|
415
|
+
while True:
|
416
|
+
if generation_started:
|
417
|
+
break
|
418
|
+
stream_response = self.STUB.StreamModelOutputs(request)
|
419
|
+
for response in stream_response:
|
420
|
+
if status_is_retryable(response.status.code) and \
|
421
|
+
time.time() - start_time < 60 * 10:
|
422
|
+
self.logger.info("Model is still deploying, please wait...")
|
423
|
+
time.sleep(next(backoff_iterator))
|
424
|
+
break
|
425
|
+
if response.status.code != status_code_pb2.SUCCESS:
|
426
|
+
raise Exception(f"Model Predict failed with response {response.status!r}")
|
427
|
+
else:
|
428
|
+
if not generation_started:
|
429
|
+
generation_started = True
|
430
|
+
yield response
|
clarifai/client/model.py
CHANGED
@@ -1,7 +1,8 @@
|
|
1
|
+
import itertools
|
1
2
|
import json
|
2
3
|
import os
|
3
4
|
import time
|
4
|
-
from typing import Any, Dict, Generator, Iterator, List, Tuple, Union
|
5
|
+
from typing import Any, Dict, Generator, Iterable, Iterator, List, Tuple, Union
|
5
6
|
|
6
7
|
import numpy as np
|
7
8
|
import requests
|
@@ -77,7 +78,8 @@ class Model(Lister, BaseClient):
|
|
77
78
|
self.logger = logger
|
78
79
|
self.training_params = {}
|
79
80
|
self.input_types = None
|
80
|
-
self.
|
81
|
+
self._client = None
|
82
|
+
self._added_methods = False
|
81
83
|
self._set_runner_selector(
|
82
84
|
compute_cluster_id=compute_cluster_id,
|
83
85
|
nodepool_id=nodepool_id,
|
@@ -418,8 +420,8 @@ class Model(Lister, BaseClient):
|
|
418
420
|
**dict(self.kwargs, model_version=model_version_info))
|
419
421
|
|
420
422
|
@property
|
421
|
-
def
|
422
|
-
if self.
|
423
|
+
def client(self):
|
424
|
+
if self._client is None:
|
423
425
|
request_template = service_pb2.PostModelOutputsRequest(
|
424
426
|
user_app_id=self.user_app_id,
|
425
427
|
model_id=self.id,
|
@@ -427,30 +429,46 @@ class Model(Lister, BaseClient):
|
|
427
429
|
model=self.model_info,
|
428
430
|
runner_selector=self._runner_selector,
|
429
431
|
)
|
430
|
-
self.
|
431
|
-
return self.
|
432
|
+
self._client = ModelClient(self.STUB, request_template=request_template)
|
433
|
+
return self._client
|
432
434
|
|
433
|
-
def predict(self,
|
434
|
-
"""Predicts the model based on the given inputs.
|
435
|
-
|
436
|
-
Args:
|
437
|
-
inputs (list[Input]): The inputs to predict, must be less than 128.
|
435
|
+
def predict(self, *args, **kwargs):
|
438
436
|
"""
|
437
|
+
Calls the model's predict() method with the given arguments.
|
439
438
|
|
440
|
-
|
441
|
-
|
442
|
-
|
443
|
-
output_config=output_config,
|
444
|
-
)
|
439
|
+
If passed in request_pb2.PostModelOutputsRequest values, will send the model the raw
|
440
|
+
protos directly for compatibility with previous versions of the SDK.
|
441
|
+
"""
|
445
442
|
|
446
|
-
|
447
|
-
|
443
|
+
inputs = None
|
444
|
+
if 'inputs' in kwargs:
|
445
|
+
inputs = kwargs['inputs']
|
446
|
+
elif args:
|
447
|
+
inputs = args[0]
|
448
|
+
if inputs and isinstance(inputs, list) and isinstance(inputs[0], resources_pb2.Input):
|
449
|
+
assert not args, "Cannot pass in raw protos and additional arguments at the same time."
|
450
|
+
inference_params = kwargs.get('inference_params', {})
|
451
|
+
output_config = kwargs.get('output_config', {})
|
452
|
+
return self.client._predict_by_proto(
|
453
|
+
inputs=inputs, inference_params=inference_params, output_config=output_config)
|
448
454
|
|
449
|
-
|
450
|
-
inputs (list[Input]): The inputs to predict, must be less than 128.
|
451
|
-
"""
|
455
|
+
return self.client.predict(*args, **kwargs)
|
452
456
|
|
453
|
-
|
457
|
+
def __getattr__(self, name):
|
458
|
+
try:
|
459
|
+
return getattr(self.model_info, name)
|
460
|
+
except AttributeError:
|
461
|
+
pass
|
462
|
+
if not self._added_methods:
|
463
|
+
# fetch and set all the model methods
|
464
|
+
self._added_methods = True
|
465
|
+
self.client.fetch()
|
466
|
+
for method_name in self.client._method_signatures.keys():
|
467
|
+
if not hasattr(self, method_name):
|
468
|
+
setattr(self, method_name, getattr(self.client, method_name))
|
469
|
+
if hasattr(self.client, name):
|
470
|
+
return getattr(self.client, name)
|
471
|
+
raise AttributeError(f"'{self.__class__.__name__}' object has no attribute '{name}'")
|
454
472
|
|
455
473
|
def _check_predict_input_type(self, input_type: str) -> None:
|
456
474
|
"""Checks if the input type is valid for the model.
|
@@ -635,24 +653,27 @@ class Model(Lister, BaseClient):
|
|
635
653
|
return self.predict(
|
636
654
|
inputs=[input_proto], inference_params=inference_params, output_config=output_config)
|
637
655
|
|
638
|
-
def generate(
|
639
|
-
|
640
|
-
|
641
|
-
inference_params: Dict = {},
|
642
|
-
output_config: Dict = {},
|
643
|
-
):
|
644
|
-
"""Generate the stream output on model based on the given inputs.
|
656
|
+
def generate(self, *args, **kwargs):
|
657
|
+
"""
|
658
|
+
Calls the model's generate() method with the given arguments.
|
645
659
|
|
646
|
-
|
647
|
-
|
648
|
-
inference_params (dict): The inference params to override.
|
649
|
-
output_config (dict): The output config to override.
|
660
|
+
If passed in request_pb2.PostModelOutputsRequest values, will send the model the raw
|
661
|
+
protos directly for compatibility with previous versions of the SDK.
|
650
662
|
"""
|
651
|
-
|
652
|
-
|
653
|
-
|
654
|
-
|
655
|
-
|
663
|
+
|
664
|
+
inputs = None
|
665
|
+
if 'inputs' in kwargs:
|
666
|
+
inputs = kwargs['inputs']
|
667
|
+
elif args:
|
668
|
+
inputs = args[0]
|
669
|
+
if inputs and isinstance(inputs, list) and isinstance(inputs[0], resources_pb2.Input):
|
670
|
+
assert not args, "Cannot pass in raw protos and additional arguments at the same time."
|
671
|
+
inference_params = kwargs.get('inference_params', {})
|
672
|
+
output_config = kwargs.get('output_config', {})
|
673
|
+
return self.client._generate_by_proto(
|
674
|
+
inputs=inputs, inference_params=inference_params, output_config=output_config)
|
675
|
+
|
676
|
+
return self.client.generate(*args, **kwargs)
|
656
677
|
|
657
678
|
def generate_by_filepath(self,
|
658
679
|
filepath: str,
|
@@ -766,28 +787,44 @@ class Model(Lister, BaseClient):
|
|
766
787
|
return self.generate(
|
767
788
|
inputs=[input_proto], inference_params=inference_params, output_config=output_config)
|
768
789
|
|
769
|
-
def stream(self,
|
770
|
-
|
771
|
-
|
772
|
-
output_config: Dict = {}):
|
773
|
-
"""Generate the stream output on model based on the given stream of inputs.
|
774
|
-
|
775
|
-
Args:
|
776
|
-
inputs (Iterator[list[Input]]): stream of inputs to predict, must be less than 128.
|
790
|
+
def stream(self, *args, **kwargs):
|
791
|
+
"""
|
792
|
+
Calls the model's stream() method with the given arguments.
|
777
793
|
|
778
|
-
|
779
|
-
|
780
|
-
>>> model = Model("url") # Example URL: https://clarifai.com/clarifai/main/models/general-image-recognition
|
781
|
-
or
|
782
|
-
>>> model = Model(model_id='model_id', user_id='user_id', app_id='app_id')
|
783
|
-
>>> stream_response = model.stream(inputs=inputs, runner_selector=runner_selector)
|
784
|
-
>>> list_stream_response = [response for response in stream_response]
|
794
|
+
If passed in request_pb2.PostModelOutputsRequest values, will send the model the raw
|
795
|
+
protos directly for compatibility with previous versions of the SDK.
|
785
796
|
"""
|
786
|
-
|
787
|
-
|
788
|
-
|
789
|
-
|
790
|
-
|
797
|
+
|
798
|
+
use_proto_call = False
|
799
|
+
inputs = None
|
800
|
+
if 'inputs' in kwargs:
|
801
|
+
inputs = kwargs['inputs']
|
802
|
+
elif args:
|
803
|
+
inputs = args[0]
|
804
|
+
if inputs and isinstance(inputs, Iterable):
|
805
|
+
inputs_iter = iter(inputs)
|
806
|
+
try:
|
807
|
+
peek = next(inputs_iter)
|
808
|
+
except StopIteration:
|
809
|
+
pass
|
810
|
+
else:
|
811
|
+
use_proto_call = isinstance(peek, resources_pb2.Input)
|
812
|
+
# put back the peeked value
|
813
|
+
if inputs_iter is inputs:
|
814
|
+
inputs = itertools.chain([peek], inputs_iter)
|
815
|
+
if 'inputs' in kwargs:
|
816
|
+
kwargs['inputs'] = inputs
|
817
|
+
else:
|
818
|
+
args = (inputs,) + args[1:]
|
819
|
+
|
820
|
+
if use_proto_call:
|
821
|
+
assert not args, "Cannot pass in raw protos and additional arguments at the same time."
|
822
|
+
inference_params = kwargs.get('inference_params', {})
|
823
|
+
output_config = kwargs.get('output_config', {})
|
824
|
+
return self.client._stream_by_proto(
|
825
|
+
inputs=inputs, inference_params=inference_params, output_config=output_config)
|
826
|
+
|
827
|
+
return self.client.stream(*args, **kwargs)
|
791
828
|
|
792
829
|
def stream_by_filepath(self,
|
793
830
|
filepath: str,
|
@@ -946,9 +983,6 @@ class Model(Lister, BaseClient):
|
|
946
983
|
self.kwargs = self.process_response_keys(dict_response['model'])
|
947
984
|
self.model_info = resources_pb2.Model(**self.kwargs)
|
948
985
|
|
949
|
-
def __getattr__(self, name):
|
950
|
-
return getattr(self.model_info, name)
|
951
|
-
|
952
986
|
def __str__(self):
|
953
987
|
if len(self.kwargs) < 10:
|
954
988
|
self.load_info()
|