Add clarifications and disclaimer

README.md

@@ -19,6 +19,14 @@ inference: false
 
 # MPT-30B-Instruct
 
+This is the MPT-30B-Instruct but with added support to finetune using peft (tested with qlora). It is not finetuned further, the weights are the same as the original MPT-30B-Instruct.
+
+I have not traced through the whole huggingface stack to see if this is working correctly but it does finetune with qlora and outputs are reasonable.
+Inspired by implementations here https://huggingface.co/cekal/mpt-7b-peft-compatible/commits/main
+https://huggingface.co/mosaicml/mpt-7b/discussions/42.
+
+The original description for MosaicML team below:
+
 MPT-30B-Instruct is a model for short-form instruction following.
 It is built by finetuning [MPT-30B](https://huggingface.co/mosaicml/mpt-30b) on [Dolly HHRLHF](https://huggingface.co/datasets/mosaicml/dolly_hhrlhf) derived from the [Databricks Dolly-15k](https://huggingface.co/datasets/databricks/databricks-dolly-15k) and the [Anthropic Helpful and Harmless (HH-RLHF)](https://huggingface.co/datasets/Anthropic/hh-rlhf) datasets. It is also trained on [Competition Math](https://huggingface.co/datasets/competition_math), [Duorc](https://huggingface.co/datasets/duorc), [CoT GSM8k](https://huggingface.co/datasets/conceptofmind/cot_submix_original), [Qasper](https://huggingface.co/datasets/allenai/qasper), [Quality](https://huggingface.co/datasets/emozilla/quality), [Summ Screen FD](https://huggingface.co/datasets/tau/scrolls) and [Spider](https://huggingface.co/datasets/spider).
   * License: _CC-By-SA-3.0_
@@ -232,4 +240,4 @@ for open-source foundation models},
     note      = {Accessed: 2023-06-22},
     urldate   = {2023-06-22}
 }
-```
+```