Clarify sharding restrictions and data-parallel support

README.md

@@ -78,8 +78,8 @@ tokenizer = AutoTokenizer.from_pretrained(
 model = AutoModelForCausalLM.from_pretrained(
     model_id,
     trust_remote_code=True,
-    torch_dtype="auto",
-    device_map="auto",
+    dtype="auto",
+    device_map="cuda",
     attn_implementation="sdpa",
 )
 
@@ -101,6 +101,10 @@ answer = tokenizer.decode(
 print(answer)
 ```
 
+The Transformers implementation supports one complete model replica per process. Automatic or explicit model sharding, tensor parallelism, and pipeline parallelism are not currently supported.
+
+For offline multi-GPU inference, independent replicas can be launched with one process per GPU using `device_map={"": local_rank}`. For production serving, continuous batching, and multi-GPU request scheduling, use the [Limite vLLM integration](https://github.com/paradigma-inc/limite-violetto).
+
 ### Supported Transformers inference paths
 
 | Attention backend | Cache | Status |