...
Code Block | ||
---|---|---|
| ||
def convert_model(sym, arg_params, aux_params, target_dtype="float16", target_dtype_ops=None, fp32_ops=None, widest_dtype_ops=None, conditional_fp32_ops=None, excluded_sym_names=None): """API for converting a model from FP32 model to a mixed precision model. MXNet tries to convert the FP32 model to mixed precision model by adding cast layers using amp_cast and amp_multicast operators. The decision on which cast layer to add is based on hardcoded lists for Automatic Mixed Precision in MXNet. These lists can be overridden by the user by providing their own lists using : targe_precision_ops, fp32_ops, widest_precision_ops, conditional_fp32_ops Parameters ---------- sym : str or Symbol Defines the structure of a neural network for FP32 types. arg_params : dict Dictionary of name to `NDArray`. aux_params : dict Dictionary of name to `NDArray`. target_dtype : str Currently only supports float16. The target dtype indicates to add cast layers when possible so that lower precision computation can be leveraged. target_dtype_ops : list of strs Override the list of operator names casted to target_dtype. If None, uses the framework's default list to be casted to target dtype. fp32_ops : list of strs Override the lists of operator names casted to FP32. If None, uses the framework's default list to be casted to FP32. widestconditional_dtypefp32_ops : list of strs A(string, string, list of op names provided by user which should run in widest precision among its inputs. If None, uses the framework's default list of widest_precision_ops. conditional_fp32_ops : list of (string, string, list of string) Override the list of operatorsstring) Override the list of operators to be casted to FP32. The format of the list is (name of the function, name of the parameter, list of values of the parameter that make the operator to be casted to FP32. The format of the list is (name of the function, name of the parameter, list of values of the parameter that make the operator to be casted to fp32) excluded_ fp32) excluded_sym_names : list of strs A list of strings that represent the names of symbols that users want to exclude from being quantized. """ |
...
Code Block | ||
---|---|---|
| ||
def convert_hybrid_block(block, target_dtype="float16", target_dtype_ops=None, fp32_ops=None, widest_dtype_ops=None, conditional_fp32_ops=None, excluded_sym_names=None, input_names=['data']): """Given a hybrid block/symbol block representing a neural network of data type FP32 and target_dtype, return a block with mixed precision support Parameters ---------- block : HybridBlock or SymbolBlock object FP32 HybridBlock or SymbolBlock object target_dtype : str or numpy currently only supports float16. The target dtype indicates to add cast layers when possible so that lower precision computation can be leveraged. target_precision_ops : list of strs Override the list of operator names casted to target_dtype. If None, uses the framework's default list to be casted to target dtype. fp32_ops : list of strs Override the lists of operator names casted to FP32. If None, uses the framework's default list to be casted to FP32. widestconditional_precisionfp32_ops : list of (string, string, list of strsstring) Override the list of operatorfunctions namescasted whichto shouldFP32. run in widest precision among its The format of the list input arguments.is If(name None,of uses the framework's default list of widest_precision_ops.function, name of the parameter, conditional_fp32_ops : list of (string, string, list of string) values of the parameter that make the operator Override the list of functionsto be casted to FP32. fp32) The format of theexcluded_sym_names : list isof strs A (namelist of thestrings function, name ofthat represent the parameter, names of symbols that users want to exclude list of values of the parameter thatfrom make the operator to be casted tobeing quantized. fp32) excluded_syminput_names : list of strs A list of strings that representrepresenting the names of symbols that users want to exclude from being quantized. input_names : list of strs A list of strings representing the names of input variables """input variables """ |
User experience will User experience will be similar to the export API experience today. Users will have to call hybridize followed by one forward pass before calling convert_model.
...
After mixed precision pass is done and amp_cast and amp_multicast layers are added, the symbolic representation needs to be modified to store the right dtype attrs for some of its inputs. This will require running InferType pass after the NNVM ReducePrecision pass and then using the obtained information to set the data types of inputsweights and auxiliary states.
This will ensure that the dtype corresponding to each param or aux input will have the right dtypeis correct, by casting the arg_params and aux_params accordingly inside .
Thus the symbol returned by convert_model .
Gluon Changes
For Gluon code, we need to add an internal API to retrieve sym, arg_params and aux_params from a hybrid_block. Following this, convert_model can be used to convert a symbol json, model params and auxiliary params. After conversion, the symbolic model (json, arg_params, aux_params) can be imported back into gluon with SymbolBlock.imports. The returned symbolblock is ready to use for inference.
Frontend Bindings
Need to add amp convert_model API support for different bindings like C++, Scala etc.
FAQ
...
API will have amp_cast and amp_multicast symbols and the "__dtype__" attribute of weight and aux symbols will be updated. Also the returned arg_params and aux_params ndarrays will have the same dtype as the "__dtype__" attribute in the returned symbol.
Gluon Changes
For Gluon code, we need to add an internal API to retrieve sym, arg_params and aux_params
...
from a hybrid_block. Following this, convert_model can be used to convert a symbol json, model params and auxiliary params. After conversion, the symbolic model (json, arg_params, aux_params) can be imported back into gluon with SymbolBlock.imports. The returned symbolblock is ready to use for inference.
Frontend Bindings
Need to add amp convert_model API support for different bindings like C++, Scala etc.
Performance
Setup
EC2 Instance: p3.8xlarge
CUDNN: 7.4.2
CUDA: 10.0
Commit Hash: b3b952f9d5490ee2707209ab866e6c3f094e2046 (PoC changes made on top of this built from source)
Mixed Precision Models:
Resnet50_v1: JSON File, Params File
imagenet1k-resnet-152: JSON File, Params File
Results
Model (Samples/sec) | Batch Size | Original Model (Samples/sec) | Mixed Precision Model (Samples/sec) | Original Model with Implicit Type Conversion (MXNET_CUDA_TENSOR_OP_MATH_ALLOW_CONVERSION=1) (Samples/sec) |
---|---|---|---|---|
imagenet1k-resnet-152 | 1 | 85 | 72 | 72 |
2 | 140 | 140 | 142 | |
4 | 240 | 270 | 228 | |
8 | 320 | 470 | 261 | |
16 | 405 | 680 | 315 | |
resnet50_v1 | 1 | 215 | 165 | 205 |
2 | 370 | 330 | 365 | |
4 | 560 | 600 | 545 | |
8 | 760 | 980 | 635 | |
16 | 935 | 1400 | 790 |
FAQ
Will the arg_params and aux_params be casted to fp16 ?
Inputs of ops in FP16 will be casted. Other params may or may not be casted based on the type inference logic.
How is this different from casting inputs to FP16 and casting params to FP16 in Gluon ?
Casting inputs to FP16 and params to FP16 for gluon ensures that you are able to execute the model in FP16 precision. Generally, there may be some ops which may need to run in FP16 while other in FP32 for accuracy and performance considerations. This is where the AMP APIs will be useful.
Will the dtype attribute in the serialized model change after convert_model is called ?
Yes dtype attribute in the serialized model can change after convert_model is called. This depends on how the whitelist affects the model in question and if the type inference decides that certain params needs to be in float16.
Is there a need for hybridizing and running a forward pass for the AMP converted gluon model ?
No there is no need to hybridize since it will return SymbolBlocks which are already hybridized.
What changes need to be made to existing script to convert and run inference mixed precision model ?
Adding the line, amp.convert_model or amp.convert_block should be sufficient to convert and run inference on a mixed precision model. Below are two user experience examples to convert a model to mixed precision model and run inference:
Module API
Code Block |
---|
sym, arg_params, aux_params = mx.model.load_checkpoint("resnet18", 0)
# Additional line below to convert to a mixed precision model. Everything else remains the same
result_sym, arg_params, aux_params = mx.contrib.amp.convert_model(sym, arg_params, aux_params, target_dtype="float16")
mod = mx.mod.Module(result_sym, data_names=['data'], label_names=None, context=mx.cpu())
mod.bind(data_shapes=[['data', (1, 3, 224, 224)]])
mod.set_params(arg_params, aux_params)
mod.forward(mx.io.DataBatch(data=[mx.nd.ones((1, 3, 224, 224))], label=None))
result = mod.get_outputs()[0].asnumpy() |
Gluon API
Code Block |
---|
net = get_model(name="resnet50_v1", classes=1000, pretrained=True)
net.hybridize()
x = mx.nd.random.uniform(0, 1, shape=(1, 3, 224, 224))
out = net(x)
# Additional line below to convert to a mixed precision model. Everything else remains the same
net = mx.contrib.amp.convert_block(net, target_dtype="float16")
out = net(x) |
Depends on the whitelists provided. The default whitelists have been selected in a way to avoid casting of the params, for commonly used convnet networks. If the whitelist is such that the type inference decides that certain param needs to be float16 then it will be casted.
How is this different from casting inputs to FP16 and casting params to FP16 in Gluon ?
Casting inputs to FP16 and params to FP16 for gluon ensures that you are able to execute the model in FP16 precision. Generally, there may be some ops which may need to run in FP16 while other in FP32 for accuracy and performance considerations. This is where the AMP APIs will be useful.
Will the dtype attribute in the serialized model change after convert_model is called ?
Yes dtype attribute in the serialized model can change after convert_model is called. This depends on how the whitelist affects the model in question and if the type inference decides that certain params needs to be in float16.
Is there a need for hybridizing and running a forward pass for the AMP converted gluon model ?
No there is no need to hybridize since it will return SymbolBlocks which are already hybridized.
What changes need to be made to existing script to convert and run inference mixed precision model ?
...