Upload model

README.md

@@ -0,0 +1,203 @@
+---
+library_name: transformers
+tags:
+- feature-extraction
+- audio
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+This is the model card of a 🤗 transformers model that has been pushed on the Hub. This model card has been automatically generated.
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure 
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+
+

config.json

@@ -0,0 +1,27 @@
+{
+  "architectures": [
+    "SincNetModel"
+  ],
+  "auto_map": {
+    "AutoConfig": "config.SincNetConfig",
+    "AutoModel": "model_hf.SincNetModel"
+  },
+  "conv_filter_length": 5,
+  "min_band_hz": 50,
+  "min_low_hz": 50,
+  "model_type": "sincnet",
+  "num_conv_filters": 60,
+  "num_sinc_filters": 80,
+  "num_wavform_channels": 1,
+  "pool_kernel_size": 3,
+  "pool_stride": 3,
+  "sample_rate": 16000,
+  "sinc_filter_dilation": 1,
+  "sinc_filter_in_channels": 1,
+  "sinc_filter_length": 251,
+  "sinc_filter_padding": 0,
+  "sinc_filter_stride": 10,
+  "stride": 10,
+  "torch_dtype": "float32",
+  "transformers_version": "4.37.2"
+}

config.py

@@ -0,0 +1,40 @@
+from transformers import PretrainedConfig
+
+class SincNetConfig(PretrainedConfig):
+    model_type = "sincnet"
+    
+    def __init__(
+        self,
+        stride: int = 10,
+        num_sinc_filters: int = 80,
+        sinc_filter_length: int = 251,
+        num_conv_filters: int = 60,
+        conv_filter_length: int = 5,
+        pool_kernel_size: int = 3,
+        pool_stride: int = 3,
+        sample_rate: int = 16000,
+        sinc_filter_stride: int = 10,
+        sinc_filter_padding: int = 0,
+        sinc_filter_dilation: int = 1,
+        min_low_hz: int = 50,
+        min_band_hz: int = 50,
+        sinc_filter_in_channels: int = 1,
+        num_wavform_channels: int = 1,
+        **kwargs
+    ):
+        self.sample_rate = sample_rate
+        self.stride = stride
+        self.num_sinc_filters = num_sinc_filters
+        self.sinc_filter_length = sinc_filter_length
+        self.num_conv_filters = num_conv_filters
+        self.conv_filter_length = conv_filter_length
+        self.pool_kernel_size = pool_kernel_size
+        self.pool_stride = pool_stride
+        self.sinc_filter_stride = sinc_filter_stride
+        self.sinc_filter_padding = sinc_filter_padding
+        self.sinc_filter_dilation = sinc_filter_dilation
+        self.min_low_hz = min_low_hz
+        self.min_band_hz = min_band_hz
+        self.sinc_filter_in_channels = sinc_filter_in_channels
+        self.num_wavform_channels = num_wavform_channels
+        super().__init__(**kwargs)   

model.py

@@ -0,0 +1,166 @@
+
+""" SincNet model """
+from functools import lru_cache
+import numpy as np
+import logging
+
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+
+logger = logging.getLogger(__name__)
+
+class SincNetFilterConvLayer(nn.Module):
+    """SincNet fast convolution filter layer"""
+
+    def __init__(self, out_channels: int, kernel_size: int, sample_rate=16000, 
+                stride=1, padding=0, dilation=1, min_low_hz=50, min_band_hz=50, 
+                in_channels=1, requires_grad=False):
+        """
+        Args:
+            out_channels : `int` number of filters.
+            kernel_size : `int` filter length.
+            sample_rate : `int`, optional sample rate. Defaults to 16000.
+        """
+        super(SincNetFilterConvLayer, self).__init__()
+
+        if in_channels != 1:
+            raise ValueError(f"SincNetFilterConvLayer only support in_channels = 1, was in_channels = {in_channels}")
+
+        self._out_channels = out_channels
+        self._kernel_size = kernel_size
+
+        if kernel_size % 2 == 0:
+            self._kernel_size += 1 # Forcing the filters to be odd
+            
+        self._stride = stride
+        self._padding = padding
+        self._dilation = dilation
+        self._sample_rate = sample_rate
+        self._min_low_hz = min_low_hz
+        self._min_band_hz = min_band_hz
+
+        # initialize filterbanks such that they are equally spaced in Mel scale
+        low_hz = 30
+        high_hz = self._sample_rate / 2 - (self._min_low_hz + self._min_band_hz)
+        mel = np.linspace(
+            2595 * np.log10(1 + low_hz / 700), # Convert Hz to Mel
+            2595 * np.log10(1 + high_hz / 700), # Convert Hz to Mel
+            self._out_channels // 2 + 1
+        )
+        hz = 700 * (10 ** (mel / 2595) - 1) # Convert Mel to Hz
+        
+        self._low_hz = nn.Parameter(
+            torch.Tensor(hz[:-1]).view(-1, 1),
+            requires_grad=requires_grad
+        )
+        self._band_hz = nn.Parameter(
+            torch.Tensor(np.diff(hz)).view(-1, 1),
+            requires_grad=requires_grad
+        )
+        self.register_buffer(
+            "_window", 
+            torch.from_numpy(np.hamming(self._kernel_size)[: self._kernel_size // 2]).float()
+        )
+        self.register_buffer(
+            "_n",
+            (2* np.pi * torch.arange(-(self._kernel_size // 2), 0.0).view(1, -1) / self._sample_rate)
+        )
+
+    @property
+    @lru_cache(maxsize=1)
+    def filters(self) -> torch.Tensor:
+        low = self._min_low_hz + torch.abs(self._low_hz)
+        high = torch.clamp(low + self._min_band_hz + torch.abs(self._band_hz), self._min_low_hz, self._sample_rate/2)
+        band = (high-low)[:,0]
+
+        f_times_t_low = torch.matmul(low, self._n)
+        f_times_t_high = torch.matmul(high, self._n)
+
+        band_pass_left = ((torch.sin(f_times_t_high)-torch.sin(f_times_t_low))/(self._n/2))*self._window
+        band_pass_center = 2 * band.view(-1, 1)
+        band_pass_right = torch.flip(band_pass_left, dims=[1])
+
+        band_pass = torch.cat([band_pass_left,band_pass_center,band_pass_right],dim=1)
+        band_pass = band_pass / (2*band[:,None])
+        return band_pass.view(self._out_channels, 1, self._kernel_size)
+
+    def forward(self, waveforms: torch.Tensor) -> torch.Tensor:
+        """
+        Args:
+            waveforms : (batch_size, 1, n_samples) batch of waveforms.
+        
+        Returns:
+            features : (batch_size, out_channels, n_samples_out) batch of sinc filters activations.
+        """
+        return F.conv1d(waveforms, self.filters, stride=self._stride,
+                        padding=self._padding, dilation=self._dilation,
+        ).abs_() # https://github.com/mravanelli/SincNet/issues/4
+
+class SincNet(nn.Module):
+    """SincNet"""
+
+    def __init__(
+        self,  
+        num_sinc_filters: int = 80,
+        sinc_filter_length: int = 251,
+        num_conv_filters: int = 60,
+        conv_filter_length: int = 5,
+        pool_kernel_size: int = 3,
+        pool_stride: int = 3,
+        sample_rate: int = 16000,
+        sinc_filter_stride: int = 10,
+        sinc_filter_padding: int = 0,
+        sinc_filter_dilation: int = 1,
+        min_low_hz: int = 50,
+        min_band_hz: int = 50,
+        sinc_filter_in_channels: int = 1,
+        num_wavform_channels: int = 1,
+    ):
+        super().__init__()
+
+        if sample_rate != 16000:
+            raise NotImplementedError(f"SincNet only supports 16kHz audio (sample_rate = 16000), was sample_rate = {sample_rate}")
+
+        self.wav_norm1d = nn.InstanceNorm1d(num_wavform_channels, affine=True)
+
+        self.conv1d = nn.ModuleList([
+            SincNetFilterConvLayer(
+                num_sinc_filters, 
+                sinc_filter_length, 
+                sample_rate=sample_rate, 
+                stride=sinc_filter_stride,
+                padding=sinc_filter_padding,
+                dilation=sinc_filter_dilation,
+                min_low_hz=min_low_hz,
+                min_band_hz=min_band_hz,
+                in_channels=sinc_filter_in_channels,
+            ),
+            nn.Conv1d(num_sinc_filters, num_conv_filters, conv_filter_length),
+            nn.Conv1d(num_conv_filters, num_conv_filters, conv_filter_length),
+        ])
+        self.pool1d = nn.ModuleList([
+            nn.MaxPool1d(pool_kernel_size, stride=pool_stride),
+            nn.MaxPool1d(pool_kernel_size, stride=pool_stride),
+            nn.MaxPool1d(pool_kernel_size, stride=pool_stride),
+        ])
+        self.norm1d = nn.ModuleList([
+            nn.InstanceNorm1d(num_sinc_filters, affine=True),
+            nn.InstanceNorm1d(num_conv_filters, affine=True),
+            nn.InstanceNorm1d(num_conv_filters, affine=True),
+        ])
+
+    def forward(self, waveforms: torch.Tensor) -> torch.Tensor:
+        """
+        Args:
+            waveforms : (batch, channel, sample)
+        """
+        outputs = self.wav_norm1d(waveforms)
+
+        for _, (conv1d, pool1d, norm1d) in enumerate(
+            zip(self.conv1d, self.pool1d, self.norm1d)
+        ):
+            outputs = conv1d(outputs)
+            outputs = F.leaky_relu(norm1d(pool1d(outputs)))
+
+        return outputs

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d1a33ae4bb439f0732b21ba7a5239c20e918b04573eec7233d07da310011dd17
+size 172768

model_hf.py

@@ -0,0 +1,27 @@
+from transformers import PreTrainedModel, AutoConfig, AutoModel
+from .model import SincNet
+from .config import SincNetConfig
+
+class SincNetModel(PreTrainedModel):
+    config_class = SincNetConfig
+    base_model_prefix = "sincnet"
+
+    def __init__(self, config: SincNetConfig):
+        super().__init__(config)
+
+        self.model = SincNet(
+            sinc_filter_stride=config.stride,
+            num_sinc_filters=config.num_sinc_filters,
+            sinc_filter_length=config.sinc_filter_length,
+            num_conv_filters=config.num_conv_filters,
+            conv_filter_length=config.conv_filter_length,
+            pool_kernel_size=config.pool_kernel_size,
+            pool_stride=config.pool_stride,
+            sample_rate=config.sample_rate,
+        )
+    
+    def forward(self, waveforms):
+        return self.model(waveforms)
+
+AutoConfig.register('sincnet', SincNetConfig)
+AutoModel.register(SincNetConfig, SincNetModel)