Buckets:

HuggingFaceDocBuilder's picture
download
raw
34 kB
import"../chunks/DsnmJJEf.js";import{i as A,h as H,C as L,H as r,b as $,D as l,E as D,s as O,a as m}from"../chunks/D6eQvJIN.js";import{p as K,o as ee,s as e,f as p,a,b as oe,c as i,d as c,r as d,n as I}from"../chunks/D9TCMYxZ.js";import{T as F}from"../chunks/eoUhp3Ag.js";import{E as z}from"../chunks/D8azpo8f.js";import{H as B}from"../chunks/DIDglxF-.js";const te='{"title":"Convolutional Vision Transformer (CvT)","local":"convolutional-vision-transformer-cvt","sections":[{"title":"Resources","local":"resources","sections":[],"depth":2},{"title":"CvtConfig","local":"transformers.CvtConfig","sections":[],"depth":2},{"title":"CvtModel","local":"transformers.CvtModel","sections":[],"depth":2},{"title":"CvtForImageClassification","local":"transformers.CvtForImageClassification","sections":[],"depth":2}],"depth":1}';var se=c('<meta name="hf:doc:metadata"/>'),ae=c("<!> <!>",1),R=c("<p>Example:</p> <!>",1),G=c(`<p>Although the recipe for forward pass needs to be defined within this function, one should call the <code>Module</code> instance afterwards instead of this since the former takes care of running the pre and post processing steps while
the latter silently ignores them.</p>`),ne=c(`<p></p> <p><em>This model was published in HF papers on 2021-03-29 and contributed to Hugging Face Transformers on 2022-05-18.</em></p> <!> <!> <p><a href="https://huggingface.co/papers/2103.15808" rel="nofollow">Convolutional Vision Transformer (CvT)</a> is a model that combines the strengths of convolutional neural networks (CNNs) and Vision transformers for the computer vision tasks. It introduces convolutional layers into the vision transformer architecture, allowing it to capture local patterns in images while maintaining the global context provided by self-attention mechanisms.</p> <p>You can find all the CvT checkpoints under the <a href="https://huggingface.co/microsoft?search_models=cvt" rel="nofollow">Microsoft</a> organization.</p> <blockquote class="tip"><p>This model was contributed by <a href="https://huggingface.co/anugunj" rel="nofollow">anujunj</a>.</p> <p>Click on the CvT models in the right sidebar for more examples of how to apply CvT to different computer vision tasks.</p></blockquote> <p>The example below demonstrates how to classify an image with <a href="/docs/transformers/pr_49542/en/main_classes/pipelines#transformers.Pipeline">Pipeline</a> or the <a href="/docs/transformers/pr_49542/en/model_doc/auto#transformers.AutoModel">AutoModel</a> class.</p> <!> <!> <p>Refer to this set of ViT <a href="https://github.com/NielsRogge/Transformers-Tutorials/tree/master/VisionTransformer" rel="nofollow">notebooks</a> for examples of inference and fine-tuning on custom datasets. Replace <code>ViTFeatureExtractor</code> and <a href="/docs/transformers/pr_49542/en/model_doc/vit#transformers.ViTForImageClassification">ViTForImageClassification</a> in these notebooks with <a href="/docs/transformers/pr_49542/en/model_doc/auto#transformers.AutoImageProcessor">AutoImageProcessor</a> and <a href="/docs/transformers/pr_49542/en/model_doc/cvt#transformers.CvtForImageClassification">CvtForImageClassification</a>.</p> <!> <div class="docstring border-l-2 border-t-2 pl-4 pt-3.5 border-gray-100 rounded-tl-xl mb-6 mt-8"><!> <p>This is the configuration class to store the configuration of a CvtModel. It is used to instantiate a Cvt
model according to the specified arguments, defining the model architecture. Instantiating a configuration with the
defaults will yield a similar configuration to that of the <a href="https://huggingface.co/microsoft/cvt-13" rel="nofollow">microsoft/cvt-13</a></p> <p>Configuration objects inherit from <a href="/docs/transformers/pr_49542/en/main_classes/configuration#transformers.PreTrainedConfig">PreTrainedConfig</a> and can be used to control the model outputs. Read the
documentation from <a href="/docs/transformers/pr_49542/en/main_classes/configuration#transformers.PreTrainedConfig">PreTrainedConfig</a> for more information.</p> <!></div> <!> <div class="docstring border-l-2 border-t-2 pl-4 pt-3.5 border-gray-100 rounded-tl-xl mb-6 mt-8"><!> <p>The bare Cvt Model outputting raw hidden-states without any specific head on top.</p> <p>This model inherits from <a href="/docs/transformers/pr_49542/en/main_classes/model#transformers.PreTrainedModel">PreTrainedModel</a>. Check the superclass documentation for the generic methods the
library implements for all its model (such as downloading or saving, resizing the input embeddings, pruning heads
etc.)</p> <p>This model is also a PyTorch <a href="https://pytorch.org/docs/stable/nn.html#torch.nn.Module" rel="nofollow">torch.nn.Module</a> subclass.
Use it as a regular PyTorch Module and refer to the PyTorch documentation for all matter related to general usage
and behavior.</p> <div class="docstring border-l-2 border-t-2 pl-4 pt-3.5 border-gray-100 rounded-tl-xl mb-6 mt-8"><!> <p>The <a href="/docs/transformers/pr_49542/en/model_doc/cvt#transformers.CvtModel">CvtModel</a> forward method, overrides the <code>__call__</code> special method.</p> <!> <ul><li><p><strong>last_hidden_state</strong> (<code>torch.FloatTensor</code> of shape <code>(batch_size, sequence_length, hidden_size)</code>, <em>optional</em>) — Sequence of hidden-states at the output of the last layer of the model.</p></li> <li><p><strong>cls_token_value</strong> (<code>torch.FloatTensor</code> of shape <code>(batch_size, 1, hidden_size)</code>) — Classification token at the output of the last layer of the model.</p></li> <li><p><strong>hidden_states</strong> (<code>tuple[torch.FloatTensor, ...]</code>, <em>optional</em>, returned when <code>output_hidden_states=True</code> is passed or when <code>config.output_hidden_states=True</code>) — Tuple of <code>torch.FloatTensor</code> (one for the output of the embeddings, if the model has an embedding layer, +
one for the output of each layer) of shape <code>(batch_size, sequence_length, hidden_size)</code>.</p> <p>Hidden-states of the model at the output of each layer plus the optional initial embedding outputs.</p></li></ul></div></div> <!> <div class="docstring border-l-2 border-t-2 pl-4 pt-3.5 border-gray-100 rounded-tl-xl mb-6 mt-8"><!> <p>Cvt Model transformer with an image classification head on top (a linear layer on top of the final hidden state of
the [CLS] token) e.g. for ImageNet.</p> <p>This model inherits from <a href="/docs/transformers/pr_49542/en/main_classes/model#transformers.PreTrainedModel">PreTrainedModel</a>. Check the superclass documentation for the generic methods the
library implements for all its model (such as downloading or saving, resizing the input embeddings, pruning heads
etc.)</p> <p>This model is also a PyTorch <a href="https://pytorch.org/docs/stable/nn.html#torch.nn.Module" rel="nofollow">torch.nn.Module</a> subclass.
Use it as a regular PyTorch Module and refer to the PyTorch documentation for all matter related to general usage
and behavior.</p> <div class="docstring border-l-2 border-t-2 pl-4 pt-3.5 border-gray-100 rounded-tl-xl mb-6 mt-8"><!> <p>The <a href="/docs/transformers/pr_49542/en/model_doc/cvt#transformers.CvtForImageClassification">CvtForImageClassification</a> forward method, overrides the <code>__call__</code> special method.</p> <!> <ul><li><strong>loss</strong> (<code>torch.FloatTensor</code> of shape <code>(1,)</code>, <em>optional</em>, returned when <code>labels</code> is provided) — Classification (or regression if config.num_labels==1) loss.</li> <li><strong>logits</strong> (<code>torch.FloatTensor</code> of shape <code>(batch_size, config.num_labels)</code>) — Classification (or regression if config.num_labels==1) scores (before SoftMax).</li> <li><strong>hidden_states</strong> (<code>tuple(torch.FloatTensor)</code>, <em>optional</em>, returned when <code>output_hidden_states=True</code> is passed or when <code>config.output_hidden_states=True</code>) — Tuple of <code>torch.FloatTensor</code> (one for the output of the embeddings, if the model has an embedding layer, +
one for the output of each stage) of shape <code>(batch_size, num_channels, height, width)</code>. Hidden-states (also
called feature maps) of the model at the output of each stage.</li></ul> <!></div></div> <!> <p></p>`,1);function pe(Y,E){K(E,!1),ee(()=>{new URLSearchParams(window.location.search).get("fw")}),A();var _=ne();H("13t0mzz",o=>{var s=se();O(s,"content",te),a(o,s)});var v=e(p(_),4);L(v,{containerStyle:"float: right; margin-left: 10px; display: inline-flex; position: relative; z-index: 10;"});var b=e(v,2);r(b,{title:"Convolutional Vision Transformer (CvT)",local:"convolutional-vision-transformer-cvt",headingTag:"h1"});var M=e(b,10);$(M,{id:"usage",options:["Pipeline","AutoModel"],children:(o,s)=>{var t=ae(),n=p(t);B(n,{id:"usage",option:"Pipeline",children:(u,S)=>{m(u,{code:"ZnJvbSUyMHRyYW5zZm9ybWVycyUyMGltcG9ydCUyMHBpcGVsaW5lJTBBJTBBJTBBcGlwZWxpbmUlMjAlM0QlMjBwaXBlbGluZSglMEElMjAlMjAlMjAlMjB0YXNrJTNEJTIyaW1hZ2UtY2xhc3NpZmljYXRpb24lMjIlMkMlMEElMjAlMjAlMjAlMjBtb2RlbCUzRCUyMm1pY3Jvc29mdCUyRmN2dC0xMyUyMiUyQyUwQSUyMCUyMCUyMCUyMGRldmljZSUzRDAlMEEpJTBBcGlwZWxpbmUoJTIyaHR0cHMlM0ElMkYlMkZodWdnaW5nZmFjZS5jbyUyRmRhdGFzZXRzJTJGaHVnZ2luZ2ZhY2UlMkZkb2N1bWVudGF0aW9uLWltYWdlcyUyRnJlc29sdmUlMkZtYWluJTJGcGlwZWxpbmUtY2F0LWNob25rLmpwZWclMjIp",highlighted:`<span class="hljs-keyword">from</span> transformers <span class="hljs-keyword">import</span> pipeline
pipeline = pipeline(
task=<span class="hljs-string">&quot;image-classification&quot;</span>,
model=<span class="hljs-string">&quot;microsoft/cvt-13&quot;</span>,
device=<span class="hljs-number">0</span>
)
pipeline(<span class="hljs-string">&quot;https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/pipeline-cat-chonk.jpeg&quot;</span>)`,lang:"python",wrap:!1})},$$slots:{default:!0}});var P=e(n,2);B(P,{id:"usage",option:"AutoModel",children:(u,S)=>{m(u,{code:"aW1wb3J0JTIwcmVxdWVzdHMlMEFpbXBvcnQlMjB0b3JjaCUwQWZyb20lMjBQSUwlMjBpbXBvcnQlMjBJbWFnZSUwQSUwQWZyb20lMjB0cmFuc2Zvcm1lcnMlMjBpbXBvcnQlMjBBdXRvSW1hZ2VQcm9jZXNzb3IlMkMlMjBBdXRvTW9kZWxGb3JJbWFnZUNsYXNzaWZpY2F0aW9uJTBBJTBBJTBBaW1hZ2VfcHJvY2Vzc29yJTIwJTNEJTIwQXV0b0ltYWdlUHJvY2Vzc29yLmZyb21fcHJldHJhaW5lZCglMjJtaWNyb3NvZnQlMkZjdnQtMTMlMjIpJTBBbW9kZWwlMjAlM0QlMjBBdXRvTW9kZWxGb3JJbWFnZUNsYXNzaWZpY2F0aW9uLmZyb21fcHJldHJhaW5lZCglMEElMjAlMjAlMjAlMjAlMjJtaWNyb3NvZnQlMkZjdnQtMTMlMjIlMkMlMEElMjAlMjAlMjAlMjBkZXZpY2VfbWFwJTNEJTIyYXV0byUyMiUwQSklMEElMEF1cmwlMjAlM0QlMjAlMjJodHRwcyUzQSUyRiUyRmh1Z2dpbmdmYWNlLmNvJTJGZGF0YXNldHMlMkZodWdnaW5nZmFjZSUyRmRvY3VtZW50YXRpb24taW1hZ2VzJTJGcmVzb2x2ZSUyRm1haW4lMkZwaXBlbGluZS1jYXQtY2hvbmsuanBlZyUyMiUwQWltYWdlJTIwJTNEJTIwSW1hZ2Uub3BlbihyZXF1ZXN0cy5nZXQodXJsJTJDJTIwc3RyZWFtJTNEVHJ1ZSkucmF3KSUwQWlucHV0cyUyMCUzRCUyMGltYWdlX3Byb2Nlc3NvcihpbWFnZSUyQyUyMHJldHVybl90ZW5zb3JzJTNEJTIycHQlMjIpLnRvKG1vZGVsLmRldmljZSklMEElMEF3aXRoJTIwdG9yY2gubm9fZ3JhZCgpJTNBJTBBJTIwJTIwbG9naXRzJTIwJTNEJTIwbW9kZWwoKippbnB1dHMpLmxvZ2l0cyUwQXByZWRpY3RlZF9jbGFzc19pZCUyMCUzRCUyMGxvZ2l0cy5hcmdtYXgoZGltJTNELTEpLml0ZW0oKSUwQSUwQWNsYXNzX2xhYmVscyUyMCUzRCUyMG1vZGVsLmNvbmZpZy5pZDJsYWJlbCUwQXByZWRpY3RlZF9jbGFzc19sYWJlbCUyMCUzRCUyMGNsYXNzX2xhYmVscyU1QnByZWRpY3RlZF9jbGFzc19pZCU1RCUwQXByaW50KGYlMjJUaGUlMjBwcmVkaWN0ZWQlMjBjbGFzcyUyMGxhYmVsJTIwaXMlM0ElMjAlN0JwcmVkaWN0ZWRfY2xhc3NfbGFiZWwlN0QlMjIp",highlighted:`<span class="hljs-keyword">import</span> requests
<span class="hljs-keyword">import</span> torch
<span class="hljs-keyword">from</span> PIL <span class="hljs-keyword">import</span> Image
<span class="hljs-keyword">from</span> transformers <span class="hljs-keyword">import</span> AutoImageProcessor, AutoModelForImageClassification
image_processor = AutoImageProcessor.from_pretrained(<span class="hljs-string">&quot;microsoft/cvt-13&quot;</span>)
model = AutoModelForImageClassification.from_pretrained(
<span class="hljs-string">&quot;microsoft/cvt-13&quot;</span>,
device_map=<span class="hljs-string">&quot;auto&quot;</span>
)
url = <span class="hljs-string">&quot;https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/pipeline-cat-chonk.jpeg&quot;</span>
image = Image.<span class="hljs-built_in">open</span>(requests.get(url, stream=<span class="hljs-literal">True</span>).raw)
inputs = image_processor(image, return_tensors=<span class="hljs-string">&quot;pt&quot;</span>).to(model.device)
<span class="hljs-keyword">with</span> torch.no_grad():
logits = model(**inputs).logits
predicted_class_id = logits.argmax(dim=-<span class="hljs-number">1</span>).item()
class_labels = model.config.id2label
predicted_class_label = class_labels[predicted_class_id]
<span class="hljs-built_in">print</span>(<span class="hljs-string">f&quot;The predicted class label is: <span class="hljs-subst">{predicted_class_label}</span>&quot;</span>)`,lang:"python",wrap:!1})},$$slots:{default:!0}}),a(o,t)},$$slots:{default:!0}});var y=e(M,2);r(y,{title:"Resources",local:"resources",headingTag:"h2"});var C=e(y,4);r(C,{title:"CvtConfig",local:"transformers.CvtConfig",headingTag:"h2"});var h=e(C,2),T=i(h);l(T,{name:"class transformers.CvtConfig",anchor:"transformers.CvtConfig",source:"https://github.com/huggingface/transformers/blob/vr_49542/src/transformers/models/cvt/configuration_cvt.py#L24",parameters:[{name:"transformers_version",val:": str | None = None"},{name:"architectures",val:": list[str] | None = None"},{name:"output_hidden_states",val:": bool | None = False"},{name:"return_dict",val:": bool | None = True"},{name:"dtype",val:": str | torch.dtype | None = None"},{name:"chunk_size_feed_forward",val:": int = 0"},{name:"is_encoder_decoder",val:": bool = False"},{name:"id2label",val:": dict[int, str] | dict[str, str] | None = None"},{name:"label2id",val:": dict[str, int] | dict[str, str] | None = None"},{name:"problem_type",val:": Literal['regression', 'single_label_classification', 'multi_label_classification'] | None = None"},{name:"num_channels",val:": int = 3"},{name:"patch_sizes",val:": list[int] | tuple[int, ...] = (7, 3, 3)"},{name:"patch_stride",val:": list[int] | tuple[int, ...] = (4, 2, 2)"},{name:"patch_padding",val:": list[int] | tuple[int, ...] = (2, 1, 1)"},{name:"embed_dim",val:": list[int] | tuple[int, ...] = (64, 192, 384)"},{name:"num_heads",val:": list[int] | tuple[int, ...] = (1, 3, 6)"},{name:"depth",val:": list[int] | tuple[int, ...] = (1, 2, 10)"},{name:"mlp_ratio",val:": list[float] | tuple[float, ...] = (4.0, 4.0, 4.0)"},{name:"attention_drop_rate",val:": list[float] | tuple[float, ...] = (0.0, 0.0, 0.0)"},{name:"drop_rate",val:": list[float] | tuple[float, ...] = (0.0, 0.0, 0.0)"},{name:"drop_path_rate",val:": list[float] | tuple[float, ...] = (0.0, 0.0, 0.1)"},{name:"qkv_bias",val:": list[bool] | tuple[bool, ...] = (True, True, True)"},{name:"cls_token",val:": list[bool] | tuple[bool, ...] = (False, False, True)"},{name:"qkv_projection_method",val:": list[str] | tuple[str, ...] = ('dw_bn', 'dw_bn', 'dw_bn')"},{name:"kernel_qkv",val:": list[int] | tuple[int, ...] = (3, 3, 3)"},{name:"padding_kv",val:": list[int] | tuple[int, ...] = (1, 1, 1)"},{name:"stride_kv",val:": list[int] | tuple[int, ...] = (2, 2, 2)"},{name:"padding_q",val:": list[int] | tuple[int, ...] = (1, 1, 1)"},{name:"stride_q",val:": list[int] | tuple[int, ...] = (1, 1, 1)"},{name:"initializer_range",val:": float = 0.02"},{name:"layer_norm_eps",val:": float = 1e-12"}],parametersDescription:[{anchor:"transformers.CvtConfig.num_channels",description:`<strong>num_channels</strong> (<code>int</code>, <em>optional</em>, defaults to <code>3</code>) &#x2014;
The number of input channels.`,name:"num_channels"},{anchor:"transformers.CvtConfig.patch_sizes",description:`<strong>patch_sizes</strong> (<code>Union[list[int], tuple[int, ...]]</code>, <em>optional</em>, defaults to <code>(7, 3, 3)</code>) &#x2014;
Patch size at each stage of the model.`,name:"patch_sizes"},{anchor:"transformers.CvtConfig.patch_stride",description:`<strong>patch_stride</strong> (<code>list[int]</code>, <em>optional</em>, defaults to <code>[4, 2, 2]</code>) &#x2014;
The stride size of each encoder&#x2019;s patch embedding.`,name:"patch_stride"},{anchor:"transformers.CvtConfig.patch_padding",description:`<strong>patch_padding</strong> (<code>list[int]</code>, <em>optional</em>, defaults to <code>[2, 1, 1]</code>) &#x2014;
The padding size of each encoder&#x2019;s patch embedding.`,name:"patch_padding"},{anchor:"transformers.CvtConfig.embed_dim",description:`<strong>embed_dim</strong> (<code>Union[list[int], tuple[int, ...]]</code>, <em>optional</em>, defaults to <code>(64, 192, 384)</code>) &#x2014;
Dimensionality of the embeddings and hidden states.`,name:"embed_dim"},{anchor:"transformers.CvtConfig.num_heads",description:`<strong>num_heads</strong> (<code>Union[list[int], tuple[int, ...]]</code>, <em>optional</em>, defaults to <code>(1, 3, 6)</code>) &#x2014;
Number of attention heads for each attention layer in the Transformer decoder.`,name:"num_heads"},{anchor:"transformers.CvtConfig.depth",description:`<strong>depth</strong> (<code>list[int]</code>, <em>optional</em>, defaults to <code>[1, 2, 10]</code>) &#x2014;
The number of layers in each encoder block.`,name:"depth"},{anchor:"transformers.CvtConfig.mlp_ratio",description:`<strong>mlp_ratio</strong> (<code>Union[list[float], tuple[float, ...]]</code>, <em>optional</em>, defaults to <code>(4.0, 4.0, 4.0)</code>) &#x2014;
Ratio of the MLP hidden dim to the embedding dim.`,name:"mlp_ratio"},{anchor:"transformers.CvtConfig.attention_drop_rate",description:`<strong>attention_drop_rate</strong> (<code>list[float]</code>, <em>optional</em>, defaults to <code>[0.0, 0.0, 0.0]</code>) &#x2014;
The dropout ratio for the attention probabilities.`,name:"attention_drop_rate"},{anchor:"transformers.CvtConfig.drop_rate",description:`<strong>drop_rate</strong> (<code>list[float]</code>, <em>optional</em>, defaults to <code>[0.0, 0.0, 0.0]</code>) &#x2014;
The dropout ratio for the patch embeddings probabilities.`,name:"drop_rate"},{anchor:"transformers.CvtConfig.drop_path_rate",description:`<strong>drop_path_rate</strong> (<code>Union[list[float], tuple[float, ...]]</code>, <em>optional</em>, defaults to <code>(0.0, 0.0, 0.1)</code>) &#x2014;
Drop path rate for the patch fusion.`,name:"drop_path_rate"},{anchor:"transformers.CvtConfig.qkv_bias",description:`<strong>qkv_bias</strong> (<code>Union[list[bool], tuple[bool, ...]]</code>, <em>optional</em>, defaults to <code>(True, True, True)</code>) &#x2014;
Whether to add a bias to the queries, keys and values.`,name:"qkv_bias"},{anchor:"transformers.CvtConfig.cls_token",description:`<strong>cls_token</strong> (<code>list[bool]</code>, <em>optional</em>, defaults to <code>[False, False, True]</code>) &#x2014;
Whether or not to add a classification token to the output of each of the last 3 stages.`,name:"cls_token"},{anchor:"transformers.CvtConfig.qkv_projection_method",description:`<strong>qkv_projection_method</strong> (<code>list[string]</code>, <em>optional</em>, defaults to [&#x201C;dw_bn&#x201D;, &#x201C;dw_bn&#x201D;, &#x201C;dw_bn&#x201D;]\`) &#x2014;
The projection method for query, key and value Default is depth-wise convolutions with batch norm. For
Linear projection use &#x201C;avg&#x201D;.`,name:"qkv_projection_method"},{anchor:"transformers.CvtConfig.kernel_qkv",description:`<strong>kernel_qkv</strong> (<code>list[int]</code>, <em>optional</em>, defaults to <code>[3, 3, 3]</code>) &#x2014;
The kernel size for query, key and value in attention layer`,name:"kernel_qkv"},{anchor:"transformers.CvtConfig.padding_kv",description:`<strong>padding_kv</strong> (<code>list[int]</code>, <em>optional</em>, defaults to <code>[1, 1, 1]</code>) &#x2014;
The padding size for key and value in attention layer`,name:"padding_kv"},{anchor:"transformers.CvtConfig.stride_kv",description:`<strong>stride_kv</strong> (<code>list[int]</code>, <em>optional</em>, defaults to <code>[2, 2, 2]</code>) &#x2014;
The stride size for key and value in attention layer`,name:"stride_kv"},{anchor:"transformers.CvtConfig.padding_q",description:`<strong>padding_q</strong> (<code>list[int]</code>, <em>optional</em>, defaults to <code>[1, 1, 1]</code>) &#x2014;
The padding size for query in attention layer`,name:"padding_q"},{anchor:"transformers.CvtConfig.stride_q",description:`<strong>stride_q</strong> (<code>list[int]</code>, <em>optional</em>, defaults to <code>[1, 1, 1]</code>) &#x2014;
The stride size for query in attention layer`,name:"stride_q"},{anchor:"transformers.CvtConfig.initializer_range",description:`<strong>initializer_range</strong> (<code>float</code>, <em>optional</em>, defaults to <code>0.02</code>) &#x2014;
The standard deviation of the truncated_normal_initializer for initializing all weight matrices.`,name:"initializer_range"},{anchor:"transformers.CvtConfig.layer_norm_eps",description:`<strong>layer_norm_eps</strong> (<code>float</code>, <em>optional</em>, defaults to <code>1e-12</code>) &#x2014;
The epsilon used by the layer normalization layers.`,name:"layer_norm_eps"}]});var Q=e(T,6);z(Q,{anchor:"transformers.CvtConfig.example",children:(o,s)=>{var t=R(),n=e(p(t),2);m(n,{code:"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",highlighted:`<span class="hljs-meta">&gt;&gt;&gt; </span><span class="hljs-keyword">from</span> transformers <span class="hljs-keyword">import</span> CvtConfig, CvtModel
<span class="hljs-meta">&gt;&gt;&gt; </span><span class="hljs-comment"># Initializing a Cvt msft/cvt style configuration</span>
<span class="hljs-meta">&gt;&gt;&gt; </span>configuration = CvtConfig()
<span class="hljs-meta">&gt;&gt;&gt; </span><span class="hljs-comment"># Initializing a model (with random weights) from the msft/cvt style configuration</span>
<span class="hljs-meta">&gt;&gt;&gt; </span>model = CvtModel(configuration)
<span class="hljs-meta">&gt;&gt;&gt; </span><span class="hljs-comment"># Accessing the model configuration</span>
<span class="hljs-meta">&gt;&gt;&gt; </span>configuration = model.config`,lang:"python",wrap:!1}),a(o,t)},$$slots:{default:!0}}),d(h);var j=e(h,2);r(j,{title:"CvtModel",local:"transformers.CvtModel",headingTag:"h2"});var f=e(j,2),w=i(f);l(w,{name:"class transformers.CvtModel",anchor:"transformers.CvtModel",source:"https://github.com/huggingface/transformers/blob/vr_49542/src/transformers/models/cvt/modeling_cvt.py#L499",parameters:[{name:"config",val:""},{name:"add_pooling_layer",val:" = True"}],parametersDescription:[{anchor:"transformers.CvtModel.config",description:`<strong>config</strong> (<a href="/docs/transformers/pr_49542/en/model_doc/cvt#transformers.CvtModel">CvtModel</a>) &#x2014;
Model configuration class with all the parameters of the model. Initializing with a config file does not
load the weights associated with the model, only the configuration. Check out the
<a href="/docs/transformers/pr_49542/en/main_classes/model#transformers.PreTrainedModel.from_pretrained">from_pretrained()</a> method to load the model weights.`,name:"config"},{anchor:"transformers.CvtModel.add_pooling_layer",description:`<strong>add_pooling_layer</strong> (<code>bool</code>, <em>optional</em>, defaults to <code>True</code>) &#x2014;
Whether to add a pooling layer`,name:"add_pooling_layer"}]});var Z=e(w,8),U=i(Z);l(U,{name:"forward",anchor:"transformers.CvtModel.forward",source:"https://github.com/huggingface/transformers/blob/vr_49542/src/transformers/models/cvt/modeling_cvt.py#L510",parameters:[{name:"pixel_values",val:": typing.Optional[torch.Tensor] = None"},{name:"output_hidden_states",val:": bool | None = None"},{name:"return_dict",val:": bool | None = None"},{name:"**kwargs",val:""}],parametersDescription:[{anchor:"transformers.CvtModel.forward.pixel_values",description:`<strong>pixel_values</strong> (<code>torch.Tensor</code> of shape <code>(batch_size, num_channels, image_size, image_size)</code>, <em>optional</em>) &#x2014;
The tensors corresponding to the input images. Pixel values can be obtained using
<a href="/docs/transformers/pr_49542/en/model_doc/convnext#transformers.ConvNextImageProcessor">ConvNextImageProcessor</a>. See <a href="/docs/transformers/pr_49542/en/model_doc/molmo2#transformers.Molmo2ImageProcessor.__call__">ConvNextImageProcessor.<strong>call</strong>()</a> for details (<code>processor_class</code> uses
<a href="/docs/transformers/pr_49542/en/model_doc/convnext#transformers.ConvNextImageProcessor">ConvNextImageProcessor</a> for processing images).`,name:"pixel_values"},{anchor:"transformers.CvtModel.forward.output_hidden_states",description:`<strong>output_hidden_states</strong> (<code>bool</code>, <em>optional</em>) &#x2014;
Whether or not to return the hidden states of all layers. See <code>hidden_states</code> under returned tensors for
more detail.`,name:"output_hidden_states"},{anchor:"transformers.CvtModel.forward.return_dict",description:`<strong>return_dict</strong> (<code>bool</code>, <em>optional</em>) &#x2014;
Whether or not to return a <a href="/docs/transformers/pr_49542/en/main_classes/output#transformers.utils.ModelOutput">ModelOutput</a> instead of a plain tuple.`,name:"return_dict"}],returnDescription:`<script context="module">export const metadata = 'undefined';<\/script>
<p>A <code>BaseModelOutputWithCLSToken</code> or a tuple of
<code>torch.FloatTensor</code> (if <code>return_dict=False</code> is passed or when <code>config.return_dict=False</code>) comprising various
elements depending on the configuration (<a
href="/docs/transformers/pr_49542/en/model_doc/cvt#transformers.CvtConfig"
>CvtConfig</a>) and inputs.</p>
`,returnType:`<script context="module">export const metadata = 'undefined';<\/script>
<p><code>BaseModelOutputWithCLSToken</code> or <code>tuple(torch.FloatTensor)</code></p>
`});var q=e(U,4);F(q,{children:(o,s)=>{var t=G();a(o,t)},$$slots:{default:!0}}),I(2),d(Z),d(f);var W=e(f,2);r(W,{title:"CvtForImageClassification",local:"transformers.CvtForImageClassification",headingTag:"h2"});var g=e(W,2),k=i(g);l(k,{name:"class transformers.CvtForImageClassification",anchor:"transformers.CvtForImageClassification",source:"https://github.com/huggingface/transformers/blob/vr_49542/src/transformers/models/cvt/modeling_cvt.py#L549",parameters:[{name:"config",val:""}],parametersDescription:[{anchor:"transformers.CvtForImageClassification.config",description:`<strong>config</strong> (<a href="/docs/transformers/pr_49542/en/model_doc/cvt#transformers.CvtForImageClassification">CvtForImageClassification</a>) &#x2014;
Model configuration class with all the parameters of the model. Initializing with a config file does not
load the weights associated with the model, only the configuration. Check out the
<a href="/docs/transformers/pr_49542/en/main_classes/model#transformers.PreTrainedModel.from_pretrained">from_pretrained()</a> method to load the model weights.`,name:"config"}]});var J=e(k,8),x=i(J);l(x,{name:"forward",anchor:"transformers.CvtForImageClassification.forward",source:"https://github.com/huggingface/transformers/blob/vr_49542/src/transformers/models/cvt/modeling_cvt.py#L564",parameters:[{name:"pixel_values",val:": typing.Optional[torch.Tensor] = None"},{name:"labels",val:": typing.Optional[torch.Tensor] = None"},{name:"output_hidden_states",val:": bool | None = None"},{name:"return_dict",val:": bool | None = None"},{name:"**kwargs",val:""}],parametersDescription:[{anchor:"transformers.CvtForImageClassification.forward.pixel_values",description:`<strong>pixel_values</strong> (<code>torch.Tensor</code> of shape <code>(batch_size, num_channels, image_size, image_size)</code>, <em>optional</em>) &#x2014;
The tensors corresponding to the input images. Pixel values can be obtained using
<a href="/docs/transformers/pr_49542/en/model_doc/convnext#transformers.ConvNextImageProcessor">ConvNextImageProcessor</a>. See <a href="/docs/transformers/pr_49542/en/model_doc/molmo2#transformers.Molmo2ImageProcessor.__call__">ConvNextImageProcessor.<strong>call</strong>()</a> for details (<code>processor_class</code> uses
<a href="/docs/transformers/pr_49542/en/model_doc/convnext#transformers.ConvNextImageProcessor">ConvNextImageProcessor</a> for processing images).`,name:"pixel_values"},{anchor:"transformers.CvtForImageClassification.forward.labels",description:`<strong>labels</strong> (<code>torch.LongTensor</code> of shape <code>(batch_size,)</code>, <em>optional</em>) &#x2014;
Labels for computing the image classification/regression loss. Indices should be in <code>[0, ..., config.num_labels - 1]</code>. If <code>config.num_labels == 1</code> a regression loss is computed (Mean-Square loss), If
<code>config.num_labels &gt; 1</code> a classification loss is computed (Cross-Entropy).`,name:"labels"},{anchor:"transformers.CvtForImageClassification.forward.output_hidden_states",description:`<strong>output_hidden_states</strong> (<code>bool</code>, <em>optional</em>) &#x2014;
Whether or not to return the hidden states of all layers. See <code>hidden_states</code> under returned tensors for
more detail.`,name:"output_hidden_states"},{anchor:"transformers.CvtForImageClassification.forward.return_dict",description:`<strong>return_dict</strong> (<code>bool</code>, <em>optional</em>) &#x2014;
Whether or not to return a <a href="/docs/transformers/pr_49542/en/main_classes/output#transformers.utils.ModelOutput">ModelOutput</a> instead of a plain tuple.`,name:"return_dict"}],returnDescription:`<script context="module">export const metadata = 'undefined';<\/script>
<p>A <a
href="/docs/transformers/pr_49542/en/main_classes/output#transformers.modeling_outputs.ImageClassifierOutputWithNoAttention"
>ImageClassifierOutputWithNoAttention</a> or a tuple of
<code>torch.FloatTensor</code> (if <code>return_dict=False</code> is passed or when <code>config.return_dict=False</code>) comprising various
elements depending on the configuration (<a
href="/docs/transformers/pr_49542/en/model_doc/cvt#transformers.CvtConfig"
>CvtConfig</a>) and inputs.</p>
`,returnType:`<script context="module">export const metadata = 'undefined';<\/script>
<p><a
href="/docs/transformers/pr_49542/en/main_classes/output#transformers.modeling_outputs.ImageClassifierOutputWithNoAttention"
>ImageClassifierOutputWithNoAttention</a> or <code>tuple(torch.FloatTensor)</code></p>
`});var N=e(x,4);F(N,{children:(o,s)=>{var t=G();a(o,t)},$$slots:{default:!0}});var X=e(N,4);z(X,{anchor:"transformers.CvtForImageClassification.forward.example",children:(o,s)=>{var t=R(),n=e(p(t),2);m(n,{code:"ZnJvbSUyMHRyYW5zZm9ybWVycyUyMGltcG9ydCUyMEF1dG9JbWFnZVByb2Nlc3NvciUyQyUyMEN2dEZvckltYWdlQ2xhc3NpZmljYXRpb24lMEFpbXBvcnQlMjB0b3JjaCUwQWZyb20lMjBkYXRhc2V0cyUyMGltcG9ydCUyMGxvYWRfZGF0YXNldCUwQSUwQWRhdGFzZXQlMjAlM0QlMjBsb2FkX2RhdGFzZXQoJTIyaHVnZ2luZ2ZhY2UlMkZjYXRzLWltYWdlJTIyKSUwQWltYWdlJTIwJTNEJTIwZGF0YXNldCU1QiUyMnRlc3QlMjIlNUQlNUIlMjJpbWFnZSUyMiU1RCU1QjAlNUQlMEElMEFpbWFnZV9wcm9jZXNzb3IlMjAlM0QlMjBBdXRvSW1hZ2VQcm9jZXNzb3IuZnJvbV9wcmV0cmFpbmVkKCUyMm1pY3Jvc29mdCUyRmN2dC0xMyUyMiklMEFtb2RlbCUyMCUzRCUyMEN2dEZvckltYWdlQ2xhc3NpZmljYXRpb24uZnJvbV9wcmV0cmFpbmVkKCUyMm1pY3Jvc29mdCUyRmN2dC0xMyUyMiklMEElMEFpbnB1dHMlMjAlM0QlMjBpbWFnZV9wcm9jZXNzb3IoaW1hZ2UlMkMlMjByZXR1cm5fdGVuc29ycyUzRCUyMnB0JTIyKSUwQSUwQXdpdGglMjB0b3JjaC5ub19ncmFkKCklM0ElMEElMjAlMjAlMjAlMjBsb2dpdHMlMjAlM0QlMjBtb2RlbCgqKmlucHV0cykubG9naXRzJTBBJTBBJTIzJTIwbW9kZWwlMjBwcmVkaWN0cyUyMG9uZSUyMG9mJTIwdGhlJTIwMTAwMCUyMEltYWdlTmV0JTIwY2xhc3NlcyUwQXByZWRpY3RlZF9sYWJlbCUyMCUzRCUyMGxvZ2l0cy5hcmdtYXgoLTEpLml0ZW0oKSUwQXByaW50KG1vZGVsLmNvbmZpZy5pZDJsYWJlbCU1QnByZWRpY3RlZF9sYWJlbCU1RCk=",highlighted:`<span class="hljs-meta">&gt;&gt;&gt; </span><span class="hljs-keyword">from</span> transformers <span class="hljs-keyword">import</span> AutoImageProcessor, CvtForImageClassification
<span class="hljs-meta">&gt;&gt;&gt; </span><span class="hljs-keyword">import</span> torch
<span class="hljs-meta">&gt;&gt;&gt; </span><span class="hljs-keyword">from</span> datasets <span class="hljs-keyword">import</span> load_dataset
<span class="hljs-meta">&gt;&gt;&gt; </span>dataset = load_dataset(<span class="hljs-string">&quot;huggingface/cats-image&quot;</span>)
<span class="hljs-meta">&gt;&gt;&gt; </span>image = dataset[<span class="hljs-string">&quot;test&quot;</span>][<span class="hljs-string">&quot;image&quot;</span>][<span class="hljs-number">0</span>]
<span class="hljs-meta">&gt;&gt;&gt; </span>image_processor = AutoImageProcessor.from_pretrained(<span class="hljs-string">&quot;microsoft/cvt-13&quot;</span>)
<span class="hljs-meta">&gt;&gt;&gt; </span>model = CvtForImageClassification.from_pretrained(<span class="hljs-string">&quot;microsoft/cvt-13&quot;</span>)
<span class="hljs-meta">&gt;&gt;&gt; </span>inputs = image_processor(image, return_tensors=<span class="hljs-string">&quot;pt&quot;</span>)
<span class="hljs-meta">&gt;&gt;&gt; </span><span class="hljs-keyword">with</span> torch.no_grad():
<span class="hljs-meta">... </span> logits = model(**inputs).logits
<span class="hljs-meta">&gt;&gt;&gt; </span><span class="hljs-comment"># model predicts one of the 1000 ImageNet classes</span>
<span class="hljs-meta">&gt;&gt;&gt; </span>predicted_label = logits.argmax(-<span class="hljs-number">1</span>).item()
<span class="hljs-meta">&gt;&gt;&gt; </span><span class="hljs-built_in">print</span>(model.config.id2label[predicted_label])
...`,lang:"python",wrap:!1}),a(o,t)},$$slots:{default:!0}}),d(J),d(g);var V=e(g,2);D(V,{source:"https://github.com/huggingface/transformers/blob/main/docs/source/en/model_doc/cvt.md"}),I(2),a(Y,_),oe()}export{pe as component};

Xet Storage Details

Size:
34 kB
·
Xet hash:
01b215a8274c66efe72492f70be4f96d281e594434f185079578f6c1407093aa

Xet efficiently stores files, intelligently splitting them into unique chunks and accelerating uploads and downloads. More info.