Fix documentation for generator, vector DB and reranker modules (#95)

* include init.py file in root directory for correct package installation

* fix platform arg in OpenAI usage

* Update README.md

* update doc for GPU usage

* update doc for GPU usage

* Update README.md

* Update README.md

* formatting

* Update README.md

---------

Co-authored-by: Shuai Zhang <cheungdaven@gmail.com>

src/agrag/modules/generator/README.md

@@ -26,13 +26,13 @@ generator:
 
   generator_model_platform_args: Additional platform-specific parameters to use when initializing the model, generating text, etc.
   
-  num_gpus: Number of GPUs to use for generating responses (default is 0).
+  num_gpus: Number of GPUs to use for generating responses. If no value is provided, the maximum available GPUs will be used. Otherwise, the minimum of the provided value and maximum available GPUs will be used.
 
-  generator_query_prefix: Prefix to add to each query that will be passed into the generator.
+  generator_query_prefix: Prefix to be added to each query that will be passed into the generator.
 ```
 
 #### `generator_model_platform_args` structure
-If you are using `huggingface` platform, the arguments must be structured as:
+If you are using `openai` platform, the arguments must be structured as:
   ```python
   generator_model_platform_args = {
       "gpt_generate_params": {}, #Additional parameters to pass to the OpenAI GPT model's `create` method.

src/agrag/modules/retriever/rerankers/reranker.py

@@ -25,7 +25,8 @@ class Reranker:
     batch_size : int
         The size of the batch. If you have limited CUDA memory, decrease the size of the batch (default is 64).
     num_gpus: int
-        Number of GPUs to use for reranking.
+        Number of GPUs to use for generating responses. If no value is provided, the maximum available GPUs will be used.
+        Otherwise, the minimum of the provided value and maximum available GPUs will be used.
     top_k: int,
         The top-k documents to use as context for generation (default is 10).
     **kwargs : dict

src/agrag/modules/vector_db/README.md

@@ -35,7 +35,7 @@ vector_db:
   metadata_index_load_path:  The path to an existing, stored Metadata JSON file. This can be a local path or an S3 path.
   Note that we do not support using an existing Milvus database
 
-  num_gpus: Number of GPUs to use when building the index
+  num_gpus: Number of GPUs to use for generating responses. If no value is provided, the maximum available GPUs will be used. Otherwise, the minimum of the provided value and maximum available GPUs will be used.
   
   metadata: Metadata for each embedding stored in the database