Merge branch 'main' of https://github.com/EmileAydar/transformers

Author: EmileAydar

docs/source/en/model_doc/altclip.md

@@ -17,8 +17,6 @@ rendered properly in your Markdown viewer.
 <div style="float: right;">
   <div class="flex flex-wrap space-x-1">
     <img alt="PyTorch" src="https://img.shields.io/badge/PyTorch-DE3412?style=flat&logo=pytorch&logoColor=white">
-    <img alt="Transformers" src="https://img.shields.io/badge/Transformers-6B5B95?style=flat&logo=transformers&logoColor=white">
-  </div>
 </div>
 
 # AltCLIP
@@ -34,11 +32,6 @@ The examples below demonstrates how to calculate similarity scores between an im
 
 <hfoptions id="usage">
 
-<hfoption id="Pipeline">
-
-`pipeline()` isn’t available because altCLIP currently exposes only its contrastive head (image × text similarity) and no dedicated *visual-question-answering* or *captioning* head.  
-Use the AutoModel path shown below instead.
-
 </hfoption>
 
 <hfoption id="AutoModel">
@@ -76,9 +69,7 @@ AltCLIP does **not** require `transformers-cli` at inference time, but the tool
 
 </hfoptions>
 
----
 
-## Quantization
 
 Quantization reduces the memory burden of large models by representing the weights in a lower precision. Refer to the [Quantization](https://huggingface.co/docs/transformers/main/en/quantization/overview) overview for more available quantization backends.
 
@@ -113,12 +104,9 @@ for label, prob in zip(labels, probs[0]):
     print(f"{label}: {prob.item():.4f}")
 ```
 
-On a typical machine the INT-8 checkpoint occupies **≈ ½ the RAM** of the full-precision model with negligible accuracy drop.
 
-> ℹ️ Embedding layers can be quantized with *float-qparams weight-only* configs once PyTorch exposes them via the public API.
 
 
----
 
 ## Attention visualisation