---
base_model: Sao10K/L3-8B-Stheno-v3.3-32K
quantized_by: Lewdiculous
library_name: transformers
license: cc-by-nc-4.0
inference: false
language:
- en
tags:
- roleplay
- llama3
- sillytavern
---
# #roleplay #sillytavern #llama3
My GGUF-IQ-Imatrix quants for [**Sao10K/L3-8B-Stheno-v3.3-32K**](https://huggingface.co./Sao10K/L3-8B-Stheno-v3.3-32K).
**Sao10K** with Stheno **yet** again, now bigger and better than ever!
I recommend checking his page for feedback and support.
> [!IMPORTANT]
> **Quantization process:**
> Imatrix data was generated from the FP16-GGUF and conversions directly from the BF16-GGUF.
> This is a bit more disk and compute intensive but hopefully avoids any losses during conversion.
> To run this model, please use the [**latest version of KoboldCpp**](https://github.com/LostRuins/koboldcpp/releases/latest).
> If you noticed any issues let me know in the discussions.
> [!NOTE]
> **General usage:**
> For **8GB VRAM** GPUs, I recommend the **Q4_K_M-imat** (4.89 BPW) quant for up to 12288 context sizes.
>
> **Presets:**
> Some compatible SillyTavern presets can be found [**here (Virt's Roleplay Presets)**](https://huggingface.co./Virt-io/SillyTavern-Presets).
> Check [**discussions such as this one**](https://huggingface.co./Virt-io/SillyTavern-Presets/discussions/5#664d6fb87c563d4d95151baa) for other recommendations and samplers.
⇲ Click here to expand/hide information – General chart with relative quant parformances.
> [!NOTE]
> **Recommended read:**
>
> [**"Which GGUF is right for me? (Opinionated)" by Artefact2**](https://gist.github.com/Artefact2/b5f810600771265fc1e39442288e8ec9)
>
> *Click the image to view full size.*
> 
> [!TIP]
> **Personal-support:**
> I apologize for disrupting your experience.
> Eventually I may be able to use a dedicated server for this, but for now hopefully these quants are helpful.
> If you **want** and you are **able to**...
> You can [**spare some change over here (Ko-fi)**](https://ko-fi.com/Lewdiculous).
>
> **Author-support:**
> You can support the author [**at their own page**](https://ko-fi.com/sao10k).
Original model card information.
## **Original card:**
Trained with compute from [Backyard.ai](https://backyard.ai/) | Thanks to them and @dynafire for helping me out.
---
Training Details:
Trained at 8K Context -> Expanded to 32K Context with PoSE training.
Dataset Modifications:
\- Further Cleaned up Roleplaying Samples -> Quality Check
\- Removed Low Quality Samples from Manual Check -> Increased Baseline Quality Floor
\- More Creative Writing Samples -> 2x Samples
\- Remade and Refined Detailed Instruct Data
Notes:
\- Training run is much less aggressive than previous Stheno versions.
\- This model works when tested in bf16 with the same configs as within the file.
\- I do not know the effects quantisation has on it.
\- Roleplays pretty well. Feels nice in my opinion.
\- It has some issues on long context understanding and reasoning. Much better vs rope scaling normally though, so that is a plus.
\- Reminder, this isn't a native 32K model. It has it's issues, but it's coherent and working well.
Sanity Check // Needle in a Haystack Results:
\- This is not as complex as RULER or NIAN, but it's a basic evaluator. Some improper train examples had Haystack scores ranging from Red to Orange for most of the extended contexts.
Wandb Run:
---
Relevant Axolotl Configurations:
-> Taken from [winglian/Llama-3-8b-64k-PoSE](https://huggingface.co./winglian/Llama-3-8b-64k-PoSE)
\- I tried to find my own configs, hours of tinkering but the one he used worked best, so I stuck to it.
\- 2M Rope Theta had the best loss results during training compared to other values.
\- Leaving it at 500K rope wasn't that much worse, but 4M and 8M Theta made the grad_norm values worsen even if loss drops fast.
\- Mixing in Pretraining Data was a PITA. Made it a lot worse with formatting.
\- Pretraining / Noise made it worse at Haystack too? It wasn't all Green, Mainly Oranges.
\- Improper / Bad Rope Theta shows in Grad_Norm exploding to thousands. It'll drop to low values alright, but it's a scary fast drop even with gradient clipping.
```
sequence_len: 8192
use_pose: true
pose_max_context_len: 32768
overrides_of_model_config:
rope_theta: 2000000.0
max_position_embeddings: 32768
# peft_use_dora: true
adapter: lora
peft_use_rslora: true
lora_model_dir:
lora_r: 256
lora_alpha: 256
lora_dropout: 0.1
lora_target_linear: true
lora_target_modules:
- gate_proj
- down_proj
- up_proj
- q_proj
- v_proj
- k_proj
- o_proj
warmup_steps: 80
gradient_accumulation_steps: 6
micro_batch_size: 1
num_epochs: 2
optimizer: adamw_bnb_8bit
lr_scheduler: cosine_with_min_lr
learning_rate: 0.00004
lr_scheduler_kwargs:
min_lr: 0.000004
```