updated description of input features (#356)

* updated description of input features

* updates based on comments from Alex

* A few more updates based on comments from Alex

* Another check for more suggested changes

* Changes after Arnav's comments

* More changes after Arnav's comments

* Minor change suggested by Alex

docs/configuration/large_language_model.md

@@ -88,20 +88,84 @@ ludwig train ...
 
 ## Input Features
 
-Currently, the LLM model type only supports a single input feature of type `text`.
+Because of the way LLMs work, they accept exactly one input, which is of type `text`. 
+This input can be built up from one or more columns of the input dataset and optional 
+interspersed static text. The following three examples illustrate the range of possibilities.
 
-If no `prompt` template is provided, this feature must correspond to a column
-in the input dataset. If a prompt template is provided, the rendered prompt
-will be used as the input feature value during training and inference.
+### Single Dataset Column Only
+
+If it is intended for the input to the LLM to be just the content of a single dataset column (without 
+any other prefixed or suffixed text), no `prompt` template should be provided and the `name` of the 
+feature must correspond to that specific dataset column. See the following example.
 
 ```yaml
 input_features:
   - name: input
     type: text
 ```
 
-See [Text Features](./features/text_features.md) for
-configuration options.
+The value of the `name` attribute (`input` in the example) is the name of a 
+dataset column. 
+
+See [Text Features](./features/text_features.md) for configuration options.
+
+### Single Dataset Column with Additional Text
+
+If the input to the LLM must be created by prefixing and/or suffixing some static text 
+to the content of one dataset column, then a `prompt` `template` must be provided to specify how 
+the content of the chosen column should be formatted for the LLM. See the following example.
+
+```yaml
+prompt:
+  template: |
+    Translate into French 
+    Input: {english_input}
+    Translation:
+
+input_features:
+  - name: prompt
+    type: text
+```
+
+In the example above `english_input` is the name of a column in the input dataset. 
+In this case the `name` of the `input_feature` (`prompt`) is not the name of a 
+dataset column (as in the previous example). It is just a placeholder that is replaced 
+by the formatted text obtained by applying the template to the selected dataset column. 
+The actual name used has no significance, so you can choose any name that is intuitive 
+in the context of your application.
+
+### Multiple Dataset Columns with Interspersed Static Text
+
+This case is a generalization of the last example to situations that have to process 
+two or more dataset columns. See the example below.
+
+```yaml
+prompt:
+  template: |
+    [INST] <<SYS>>
+    You are a helpful, detailed, and polite AI assistant. 
+    Answer the question using only the provided context.
+    <</SYS>>
+    
+    ### Context:
+    {context}
+    
+    ### Question:
+    {question}
+
+    ### Response:
+    [/INST]
+
+input_features:
+  - name: prompt
+    type: text
+```
+
+As in the previous example `context` and `question` are names of columns in the input dataset. 
+The `name` of the `input_feature` (`prompt` here) is again just a placeholder that will be 
+replaced by formatted text obtained by applying the template to the selected dataset columns. 
+The name used (`prompt`) is not significant, so any intuitive name could have been used without 
+changing the results obtained.
 
 ## Output Features