Update README.md

README.md

@@ -1,34 +1,27 @@
 ---
-library_name: peft
+license: cc-by-nc-4.0
+datasets: Norquinal/OpenCAI
+language: en
+tags:
+- art
+- not-for-all-audiences
 ---
-## Training procedure
+# OpenCAI
 
+OpenCAI is a model fine-tuned from [Llama-2-13B](https://huggingface.co/NousResearch/Llama-2-13b-hf) and is an attempted open-source recreation of the style of roleplay found at [C.AI](https://beta.character.ai/). It was trained on [4800 samples](https://huggingface.co/datasets/Norquinal/OpenCAI/tree/main) of Discord roleplay interactions rather than C.AI outputs, as Discord is the likely origin of a majority of the material used to train C.AI's original model.
 
-The following `bitsandbytes` quantization config was used during training:
-- quant_method: bitsandbytes
-- load_in_8bit: False
-- load_in_4bit: True
-- llm_int8_threshold: 6.0
-- llm_int8_skip_modules: None
-- llm_int8_enable_fp32_cpu_offload: False
-- llm_int8_has_fp16_weight: False
-- bnb_4bit_quant_type: nf4
-- bnb_4bit_use_double_quant: False
-- bnb_4bit_compute_dtype: float16
+This model is primarily focused on chat and roleplay without any alignment. As such, it may output content that can be considered "unsafe" or "harmful." Please use the responsibly and to your best judgement.
 
-The following `bitsandbytes` quantization config was used during training:
-- quant_method: bitsandbytes
-- load_in_8bit: False
-- load_in_4bit: True
-- llm_int8_threshold: 6.0
-- llm_int8_skip_modules: None
-- llm_int8_enable_fp32_cpu_offload: False
-- llm_int8_has_fp16_weight: False
-- bnb_4bit_quant_type: nf4
-- bnb_4bit_use_double_quant: False
-- bnb_4bit_compute_dtype: float16
-### Framework versions
+## Prompt Format
 
-- PEFT 0.4.0
+This model uses the Pygmalion-2/Metharme prompt format. The model has been trained on prompts using three different roles, which are denoted by the following tokens: `<|system|>`, `<|user|>` and `<|model|>`.
 
-- PEFT 0.4.0
+The `<|system|>` prompt can be used to inject out-of-channel information behind the scenes, while the `<|user|>` prompt should be used to indicate user input. The `<|model|>` token should then be used to indicate that the model should generate a response. These tokens can happen multiple times and be chained up to form a conversation history.
+
+### Example Prompt
+```
+<|system|>{system_prompt}
+Characters:
+   [char]: [description]
+Summary: [summary of events]
+```