Update lora docs

TUTORIAL.md

@@ -357,19 +357,25 @@ Currently we support [Learned Positional Embeddings](https://arxiv.org/pdf/1706.
 | RoPE (Hugging<code>&nbsp;</code>Face Implementation)  | <pre>model:<br>     attn_config:<br>         rope:&nbsp;True<br>         rope_impl:&nbsp;hf</pre>| 62.3                                                |                                                                                                                                                                             |
 
 ### Can I finetune using PEFT / LoRA?
-- The LLM Foundry codebase does not directly have examples of PEFT or LORA workflows. However, our MPT model is a subclass of HuggingFace `PretrainedModel`, and https://github.com/mosaicml/llm-foundry/pull/346 added required features to enable HuggingFace’s [PEFT](https://huggingface.co/docs/peft/index) / [LORA](https://huggingface.co/docs/peft/conceptual_guides/lora) workflows for MPT. MPT models with LoRA modules can be trained either using LLM Foundry or Hugging Face's [accelerate](https://huggingface.co/docs/accelerate/index). Within LLM Foundry, run (`scripts/train/train.py`), adding `lora` arguments to the config `.yaml`, like so:
+- LLM Foundry does support LoRA via an integration with the [PEFT](https://github.com/huggingface/peft) library. Within LLM Foundry, run (`scripts/train/train.py`), adding `peft_config` arguments to the `model` section of the config `.yaml`, like so:
 <!--pytest.mark.skip-->
 ```yaml
-lora:
-  args:
-    r: 16
-    lora_alpha: 32
-    lora_dropout: 0.05
-    target_modules: ['Wqkv']
+model:
+  ...
+  peft_config:
+      r: 16
+      peft_type: LORA
+      task_type: CAUSAL_LM
+      lora_alpha: 32
+      lora_dropout: 0.05
+      target_modules:
+      - q_proj
+      - k_proj
+    target_modules:
+    - 'Wqkv'
 ```
-- In the current release, these features have Beta support.
 - For efficiency, The MPT model concatenates the `Q`, `K`, and `V` matrices in each attention block into a single `Wqkv` matrix that is three times wider. Currently, LoRA supports a low-rank approximation to this `Wqkv` matrix.
-- When evaluating with PEFT / LoRA seperated weight, just set `pretrained_lora_id_or_path` in `model`(Find an example [here](scripts/eval/yamls/hf_lora_eval.yml#L19)).
+- When evaluating with PEFT / LoRA separated weight, just set `pretrained_lora_id_or_path` in `model`(Find an example [here](scripts/eval/yamls/hf_lora_eval.yml#L19)).
 
 ### Can I quantize these models and/or run on CPU?
 - The LLM Foundry codebase does not directly have examples of quantization or limited-resource inference. But you can check out [GGML](https://github.com/ggerganov/ggml) (same library that powers llama.cpp) which has built support for efficiently running MPT models on CPU! You _can_ load your model in 8-bit precision for inference using the [bitsandbytes library](https://github.com/TimDettmers/bitsandbytes) and Hugging Face's [accelerate](https://huggingface.co/docs/accelerate/index) via `load model = AutoModelForCausalLM.from_pretrained(model_name, load_in_8bit=True, device_map="auto", trust_remote_code=True)`, although we have not extensively benchmarked the performance (see the Hugging Face [quantization documentation](https://huggingface.co/docs/transformers/main/main_classes/quantization) for more detail).