54 lines
1.8 KiB
Markdown
54 lines
1.8 KiB
Markdown
---
|
|
license: other
|
|
license_name: lfm1.0
|
|
license_link: LICENSE
|
|
language:
|
|
- en
|
|
base_model:
|
|
- LiquidAI/LFM2.5-1.2B-Instruct
|
|
pipeline_tag: text-generation
|
|
tags:
|
|
- roleplay
|
|
- rp
|
|
- writer
|
|
- Super
|
|
- creative
|
|
- lfm2.5
|
|
- experimental
|
|
- finetune
|
|
- lora
|
|
---
|
|
|
|
Super Slop Machina: Complex Roleplay Finetune of LiquidAI's LFM 2.5 1.2b instruct.
|
|
-
|
|
|
|
|
|
"Its back, stronger than ever before. Its all yours, on your phone, on your toaster, its always here."
|
|
---
|
|
|
|
# Overview:
|
|
|
|
This is, at the moment, my most recent finetune on LFM 1.2b, trained on 35M~ tokens, it overall improves on LFM 2.5 1.2b instruct roleplay tuned v2.
|
|
|
|
# In comparison to previous version:
|
|
|
|
* Writes better
|
|
* Coherence was improved
|
|
* Maintains creativity
|
|
|
|
# Quants(Speaking from personal experience with this specific model):
|
|
* BF16- Recommended, highest quality, least logical mistakes.
|
|
* Q8_0- Recommended, high quality, makes slightly more mistakes but nonetheless near lossless.
|
|
* Q6_K- Recommended if Q8_0 is too much, degradation begins, not exactly notable here, but you will notice minor detail loss.(When Mradermacher quantizes this model, I recommend getting his i1 Q6_K quant instead of the one I got in my repo, but in any case I still recommend Q8_0 or BF16)
|
|
* Q5_K_M- Recommended if hardware is really, REALLY bad, degradation becomes noticeable.
|
|
* Q4_K_M- Not recommended for most use cases, degradation is clearly noticeable.
|
|
|
|
Quants can be found in the repository, along with safetensors.
|
|
|
|
# Note:
|
|
|
|
I may improve it further later, I will still have to say that its fundamentally the same 1.2b base model, even with how much I refined its style.
|
|
|
|
# V2:
|
|
|
|
v2 is now out with some improvements: https://huggingface.co/Indexnusrefather/Super-Slop-Machina-Roleplay-1.2b-ERP-Tolerant-V2 |