Whisper

Whisper is a versatile speech recognition model trained on a large dataset of diverse audio. It can perform multilingual speech recognition, speech translation, and language identification. The model uses an encoder-decoder architecture and exhibits robust performance across different acoustic environments. RBLN NPUs can accelerate Whisper model inference using Optimum RBLN.

Key Classes

• RBLNWhisperForConditionalGeneration: Whisper model implementation for speech recognition and transcription on RBLN NPU
• RBLNWhisperForConditionalGenerationConfig: Configuration class for Whisper conditional generation model

API Reference

Classes

RBLNWhisperForConditionalGeneration

Bases: RBLNModel

Whisper model for speech recognition and transcription optimized for RBLN NPU.

This model inherits from [RBLNModel]. It implements the methods to convert and run pre-trained transformers based Whisper model on RBLN devices by: - transferring the checkpoint weights of the original into an optimized RBLN graph, - compiling the resulting graph using the RBLN compiler.

Example (Short form):

import torch
from transformers import AutoProcessor
from datasets import load_dataset
from optimum.rbln import RBLNWhisperForConditionalGeneration

# Load processor and dataset
model_id = "openai/whisper-tiny"
processor = AutoProcessor.from_pretrained(model_id)
ds = load_dataset("hf-internal-testing/librispeech_asr_dummy", "clean", split="validation")

# Prepare input features
input_features = processor(
    ds[0]["audio"]["array"],
    sampling_rate=ds[0]["audio"]["sampling_rate"],
    return_tensors="pt"
).input_features

# Load and compile model (or load pre-compiled model)
model = RBLNWhisperForConditionalGeneration.from_pretrained(
    model_id=model_id,
    export=True,
    rbln_batch_size=1
)

# Generate transcription
outputs = model.generate(input_features=input_features, return_timestamps=True)
transcription = processor.batch_decode(outputs, skip_special_tokens=True)[0]
print(f"Transcription: {transcription}")

Functions

from_pretrained(model_id, export=False, rbln_config=None, **kwargs) classmethod

The from_pretrained() function is utilized in its standard form as in the HuggingFace transformers library. User can use this function to load a pre-trained model from the HuggingFace library and convert it to a RBLN model to be run on RBLN NPUs.

Parameters:

Name	Type	Description	Default
model_id	Union[str, Path]	The model id of the pre-trained model to be loaded. It can be downloaded from the HuggingFace model hub or a local path, or a model id of a compiled model using the RBLN Compiler.	required
export	bool	A boolean flag to indicate whether the model should be compiled.	False
rbln_config	Optional[Union[Dict, RBLNModelConfig]]	Configuration for RBLN model compilation and runtime. This can be provided as a dictionary or an instance of the model's configuration class (e.g., RBLNLlamaForCausalLMConfig for Llama models). For detailed configuration options, see the specific model's configuration class documentation.	None
kwargs	Dict[str, Any]	Additional keyword arguments. Arguments with the prefix 'rbln_' are passed to rbln_config, while the remaining arguments are passed to the HuggingFace library.	{}

Returns:

Type	Description
Self	A RBLN model instance ready for inference on RBLN NPU devices.

from_model(model, *, rbln_config=None, **kwargs) classmethod

Converts and compiles a pre-trained HuggingFace library model into a RBLN model. This method performs the actual model conversion and compilation process.

Parameters:

Name	Type	Description	Default
model	PreTrainedModel	The PyTorch model to be compiled. The object must be an instance of the HuggingFace transformers PreTrainedModel class.	required
rbln_config	Optional[Union[Dict, RBLNModelConfig]]	Configuration for RBLN model compilation and runtime. This can be provided as a dictionary or an instance of the model's configuration class (e.g., RBLNLlamaForCausalLMConfig for Llama models). For detailed configuration options, see the specific model's configuration class documentation.	None
kwargs	Dict[str, Any]	Additional keyword arguments. Arguments with the prefix 'rbln_' are passed to rbln_config, while the remaining arguments are passed to the HuggingFace library.	{}

The method performs the following steps:

1. Compiles the PyTorch model into an optimized RBLN graph
2. Configures the model for the specified NPU device
3. Creates the necessary runtime objects if requested
4. Saves the compiled model and configurations

Returns:

Type	Description
Self	A RBLN model instance ready for inference on RBLN NPU devices.

save_pretrained(save_directory)

Saves a model and its configuration file to a directory, so that it can be re-loaded using the [from_pretrained] class method.

Parameters:

Name	Type	Description	Default
save_directory	Union[str, PathLike]	The directory to save the model and its configuration files. Will be created if it doesn't exist.	required

Classes

RBLNWhisperForConditionalGenerationConfig

Bases: RBLNModelConfig

Functions

__init__(batch_size=None, token_timestamps=None, use_attention_mask=None, enc_max_seq_len=None, dec_max_seq_len=None, **kwargs)

Parameters:

Name	Type	Description	Default
batch_size	int	The batch size for inference. Defaults to 1.	None
token_timestamps	bool	Whether to output token timestamps during generation. Defaults to False.	None
use_attention_mask	bool	Whether to use attention masks during inference. This is automatically set to True for RBLN-CA02 devices.	None
enc_max_seq_len	int	Maximum sequence length for the encoder.	None
dec_max_seq_len	int	Maximum sequence length for the decoder.	None
**kwargs	Dict[str, Any]	Additional arguments passed to the parent RBLNModelConfig.	{}

Raises:

Type	Description
ValueError	If batch_size is not a positive integer.