Fix images urls in docs

docs/life_cycle.md

@@ -13,7 +13,7 @@ This component contains all the code for working with your model, including all
 When you create a `mlup.UP` object and pass a model `mlup.UP(ml_model=your_model)` into it, mlup does nothing.
 At this point, you have simply created a class. This is done in order to avoid starting the processes of loading and analyzing the model, which can be lengthy, without explicit instructions.
 
-![mlup load process](https://github.com/nxexox/pymlup/tree/main/docs/assets/img/mlup-load-scheme.png)
+![mlup load process](https://github.com/nxexox/pymlup/blob/main/docs/assets/img/mlup-load-scheme.png?raw=true)
 
 #### up.ml.load()
 
@@ -155,14 +155,14 @@ Behind it lies a process of several stages:
 * If the prediction caused an error, a [PredictError](https://github.com/nxexox/pymlup/blob/main/mlup/errors.py) exception will be thrown.
 * And if the prediction is successful, the prediction results will be sent to the data transformer from the `data_transformer_for_predicted` configuration parameter for conversion to a valid JSON format and returned from the method. Called `up.ml._transform_predicted_data`.
 
-![mlup predict process](https://github.com/nxexox/pymlup/tree/main/docs/assets/img/mlup-predict-scheme.png)
+![mlup predict process](https://github.com/nxexox/pymlup/blob/main/docs/assets/img/mlup-predict-scheme.png?raw=true)
 
 ## Web application work
 
 After launching the web application, it is ready to accept requests from users. When a developer independently writes a web application with a model, he has full control and knowledge of how the request processing process occurs.
 mlup takes care of this, so the entire request processing process is described here.
 
-![mlup web application work process](https://github.com/nxexox/pymlup/tree/main/docs/assets/img/mlup-request-processed.png)
+![mlup web application work process](https://github.com/nxexox/pymlup/blob/main/docs/assets/img/mlup-request-processed.png?raw=true)
 
 The process looks like this:
 * A request has been received from a user.

docs/web_app_architectures.md

@@ -78,7 +78,7 @@ This can lead to various problems:
 
 This architecture is well suited for fast models that can process multiple requests per second.
 
-![mlup architecture directly to predict](https://github.com/nxexox/pymlup/tree/main/docs/assets/img/archi_directly_to_predict.png)
+![mlup architecture directly to predict](https://github.com/nxexox/pymlup/blob/main/docs/assets/img/archi_directly_to_predict.png?raw=true)
 
 ### Initialization of the architecture
 
@@ -118,7 +118,7 @@ To configure queue sizes and response prediction lifetimes, there are configurat
 * `ttl_client_wait: float` - the maximum time to wait for client results in the `GET: /get-predict/{predict_id}` method in seconds.
   See "[Web app API](https://github.com/nxexox/pymlup/tree/main/docs/web_app_api.md)".
 
-![mlup architecture worker and queue](https://github.com/nxexox/pymlup/tree/main/docs/assets/img/archi_worker_and_queue.png)
+![mlup architecture worker and queue](https://github.com/nxexox/pymlup/blob/main/docs/assets/img/archi_worker_and_queue.png?raw=true)
 
 ### Initialization of the architecture
 
@@ -208,7 +208,7 @@ In addition to the `batch_worker_timeout` and `min_batch_len` batching parameter
 * `ttl_client_wait: float` - the maximum time to wait for client results in the `GET: /get-predict/{predict_id}` method in seconds.
   See "[Web app API](https://github.com/nxexox/pymlup/tree/main/docs/web_app_api.md)".
 
-![mlup architecture batching](https://github.com/nxexox/pymlup/tree/main/docs/assets/img/archi_batching.png)
+![mlup architecture batching](https://github.com/nxexox/pymlup/blob/main/docs/assets/img/archi_batching.png?raw=true)
 
 ### Initialization of the architecture