Update README.md

README.md

@@ -35,7 +35,8 @@ The model was also trained with a multi-user, multi-character paradigm, where us
 	- There is no real "system" role or label. System information or instructions can be added by using `::::user:` without any attached character name.
 - OOC messages have been trained without character names, but in practice this doesn't seem to be a huge issue. The format is `(OOC: {{message}})`
 - It is **necessary** to add the BOS token at the start of the prompt (`<|begin_of_text|>` in the case of Llama-3.1), otherwise performance will be significantly reduced.
-- The model wasn't trained with an EOS token after each model/assistant turn. `::::` or `::::user` or `::::assistant` can be used as stopping strings just as effectively.
+- The model wasn't trained with an EOS token after each model/assistant turn. `::::` or `::::user` or `::::assistant` can be used as stopping strings just as effectively, at least in theory.
+	- Upon testing it appears that vLLM/Aphrodite cannot truly stop text generation without an EOS token, unlike text-generation-webui and llama.cpp. Your mileage may vary.
 
 ### Schematical example
 The BOS token was omitted here. Messages and descriptions are short for the sake of brevity.
@@ -72,6 +73,17 @@ After meeting together, Anon and Nanashi have a little chat.
 
 Note that in reality OOC messages might not necessarily always have the intended effect. This is something I'm trying to look into.
 
+## Sampling settings
+There are the settings I use for testing:
+
+- Neutralize samplers
+- Temperature: 1.0
+- Min-p: 0.03
+- DRY Repetition penalty
+	- Multiplier: 0.8
+	- Base: 1.75
+	- Allowed Length: 2
+
 ## Training details
 [Unsloth](https://github.com/unslothai/unsloth) was used on a single RTX3090 24GB GPU with QLoRA finetuning.
 
@@ -102,6 +114,9 @@ lr_scheduler_kwargs = {
 }
 ```
 
+### Eval / Train loss graph
+![Train/loss graph](https://files.catbox.moe/y7rp37.png)
+
 ## Questions and answers
 **Q. What's up with the name?**