Skip to main content

​
convert_from_keras_model

Convert a Keras model to an hls4ml ModelGraph. 
hls4ml.converters.convert_from_keras_model(
    model,
    output_dir='my-hls-test',
    project_name='myproject',
    input_data_tb=None,
    output_data_tb=None,
    backend='Vivado',
    hls_config=None,
    bit_exact=None,
    allow_da_fallback=True,
    allow_v2_fallback=True,
    **kwargs
)

​
Parameters

​
model
keras.Model
required
Keras model to convert. Must be a compiled Keras model.
​
output_dir
str
default:"my-hls-test"
Output directory for the generated HLS project.
​
project_name
str
default:"myproject"
Name of the HLS project. Used as the top-level function name.
​
input_data_tb
str
default:"None"
Path to input test data in .npy or .dat format for C simulation and co-simulation.
​
output_data_tb
str
default:"None"
Path to expected output data in .npy or .dat format for verification.
​
backend
str
default:"Vivado"
Backend to use. Options: 'Vivado', 'Vitis', 'Quartus', 'Catapult'.
​
board
str
default:"None"
Target board from supported_board.json. Overrides part parameter.
​
part
str
default:"None"
FPGA part number (e.g., 'xcvu13p-flga2577-2-e'). Backend-specific defaults used if not provided.
​
clock_period
int
default:"5"
Clock period in nanoseconds.
​
clock_uncertainty
str
default:"12.5%"
Clock uncertainty percentage. Defaults: 12.5% (Vivado HLS), 27% (Vitis HLS).
​
io_type
str
default:"io_parallel"
Interface type: 'io_parallel' or 'io_stream'.
​
hls_config
dict
default:"None"
HLS configuration dictionary. See Configuration Guide for details.
​
bit_exact
bool
default:"None"
Enable model-wise fixed-point precision propagation. If None, automatically enabled for HGQ models.
​
allow_da_fallback
bool
default:"True"
(Keras v3 only) Allow fallback to direct acyclic graph combinational logic for unsupported layers.
​
allow_v2_fallback
bool
default:"True"
(Keras v3 only) Allow fallback to Keras v2 layer handlers for unsupported layers.

​
Returns

​
hls_model
ModelGraph
The converted hls4ml model ready for compilation and synthesis.

​
Example

import tensorflow as tf
import hls4ml

# Create a simple Keras model
model = tf.keras.Sequential([
    tf.keras.layers.Dense(64, activation='relu', input_shape=(10,)),
    tf.keras.layers.Dense(32, activation='relu'),
    tf.keras.layers.Dense(1, activation='sigmoid')
])

# Configure HLS settings
hls_config = {
    'Model': {
        'Precision': 'ap_fixed<16,6>',
        'ReuseFactor': 4
    }
}

# Convert to HLS
hls_model = hls4ml.converters.convert_from_keras_model(
    model,
    output_dir='my_prj',
    project_name='my_keras_model',
    backend='Vivado',
    board='pynq-z2',
    io_type='io_parallel',
    hls_config=hls_config
)

# Compile the model
hls_model.compile()

# Test predictions
import numpy as np
X_test = np.random.rand(100, 10)
predictions = hls_model.predict(X_test)

​
Advanced Configuration

# Layer-specific configuration
hls_config = {
    'Model': {
        'Precision': 'ap_fixed<16,6>',
        'ReuseFactor': 1,
        'Strategy': 'Latency'
    },
    'LayerName': {
        'dense_1': {
            'Precision': 'ap_fixed<18,8>',
            'ReuseFactor': 2
        },
        'output': {
            'Precision': 'ap_fixed<12,4>'
        }
    }
}

hls_model = hls4ml.converters.convert_from_keras_model(
    model,
    hls_config=hls_config,
    output_dir='custom_config_prj'
)

​
parse_yaml_config

Parse conversion configuration from a YAML file. 
hls4ml.converters.parse_yaml_config(config_file)

​
Parameters

​
config_file
str
required
Path to the YAML configuration file.

​
Returns

​
config
dict
Parsed configuration dictionary.

​
Example YAML

KerasModel: my_model.h5
OutputDir: my-hls-test
ProjectName: myproject
Part: xcvu13p-flga2577-2-e
ClockPeriod: 5
IOType: io_stream
HLSConfig:
  Model:
    Precision: ap_fixed<16,6>
    ReuseFactor: 10

​
Usage

import hls4ml

# Parse YAML config
config = hls4ml.converters.parse_yaml_config('config.yml')

# Convert model using config
hls_model = hls4ml.converters.convert_from_config(config)

​
convert_from_config

Convert a model using a configuration dictionary or YAML file path. 
hls4ml.converters.convert_from_config(config)

​
Parameters

​
config
str | dict
required
Either a path to a YAML configuration file or a parsed configuration dictionary.

​
Returns

​
hls_model
ModelGraph
The converted hls4ml model.

​
Example

import hls4ml

# From YAML file
hls_model = hls4ml.converters.convert_from_config('config.yml')

# From dictionary
config = {
    'KerasModel': model,
    'OutputDir': 'my-hls-test',
    'ProjectName': 'myproject',
    'Backend': 'Vivado',
    'HLSConfig': {
        'Model': {
            'Precision': 'ap_fixed<16,6>',
            'ReuseFactor': 4
        }
    }
}
hls_model = hls4ml.converters.convert_from_config(config)

​
Supported Layers

Keras converter supports the following layer types:

​
Core Layers

  • Dense - Fully connected layer
  • Activation - Activation functions (ReLU, tanh, sigmoid, etc.)
  • Dropout - Dropout (removed during conversion)
  • Flatten - Flatten layer
  • Reshape - Reshape layer

​
Convolutional Layers

  • Conv1D - 1D convolution
  • Conv2D - 2D convolution
  • SeparableConv1D - Depthwise separable convolution
  • SeparableConv2D - 2D depthwise separable convolution
  • DepthwiseConv2D - Depthwise convolution

​
Pooling Layers

  • MaxPooling1D / MaxPooling2D
  • AveragePooling1D / AveragePooling2D
  • GlobalMaxPooling1D / GlobalMaxPooling2D
  • GlobalAveragePooling1D / GlobalAveragePooling2D

​
Normalization Layers

  • BatchNormalization
  • LayerNormalization

​
Recurrent Layers

  • SimpleRNN
  • LSTM
  • GRU
  • Bidirectional

​
Merge Layers

  • Add, Subtract, Multiply
  • Average, Maximum, Minimum
  • Concatenate
  • Dot

​
Other Layers

  • Embedding
  • Softmax
  • ZeroPadding1D / ZeroPadding2D
  • UpSampling1D / UpSampling2D
  • Resize

​
See Also

Build docs developers (and LLMs) love

Get started for freeTalk to us