You can deploy this model and start earning money today!

RichardErkhov/
Evillain_-_StarDust_20B_v0.2-gguf

Name: Evillain_-_StarDust_20B_v0.2-gguf
Availability: InStock

endpoints_compatible

gguf

template

Views

No views yet

Quantization made by Richard Erkhov.

Github

Discord

Request more models

StarDust_20B_v0.2 - GGUF

Model creator: https://huggingface.co/Evillain/
Original model: https://huggingface.co/Evillain/StarDust_20B_v0.2/

Name	Quant method	Size
StarDust_20B_v0.2.Q2_K.gguf	Q2_K	6.91GB
StarDust_20B_v0.2.IQ3_XS.gguf	IQ3_XS	7.63GB
StarDust_20B_v0.2.IQ3_S.gguf	IQ3_S	8.06GB
StarDust_20B_v0.2.Q3_K_S.gguf	Q3_K_S	8.06GB
StarDust_20B_v0.2.IQ3_M.gguf	IQ3_M	8.53GB
StarDust_20B_v0.2.Q3_K.gguf	Q3_K	9.04GB
StarDust_20B_v0.2.Q3_K_M.gguf	Q3_K_M	9.04GB
StarDust_20B_v0.2.Q3_K_L.gguf	Q3_K_L	9.9GB
StarDust_20B_v0.2.IQ4_XS.gguf	IQ4_XS	10.01GB
StarDust_20B_v0.2.Q4_0.gguf	Q4_0	10.52GB
StarDust_20B_v0.2.IQ4_NL.gguf	IQ4_NL	10.57GB
StarDust_20B_v0.2.Q4_K_S.gguf	Q4_K_S	7.56GB
StarDust_20B_v0.2.Q4_K.gguf	Q4_K	5.57GB
StarDust_20B_v0.2.Q4_K_M.gguf	Q4_K_M	4.12GB
StarDust_20B_v0.2.Q4_1.gguf	Q4_1	3.66GB
StarDust_20B_v0.2.Q5_0.gguf	Q5_0	2.97GB
StarDust_20B_v0.2.Q5_K_S.gguf	Q5_K_S	2.04GB
StarDust_20B_v0.2.Q5_K.gguf	Q5_K	1.65GB
StarDust_20B_v0.2.Q5_K_M.gguf	Q5_K_M	1.39GB
StarDust_20B_v0.2.Q5_1.gguf	Q5_1	1.31GB
StarDust_20B_v0.2.Q6_K.gguf	Q6_K	1.05GB
StarDust_20B_v0.2.Q8_0.gguf	Q8_0	0.99GB

Original model description:

license: other library_name: transformers tags:

mergekit
merge
not-for-all-audiences base_model:
Kooten/DaringMaid-20B
TeeZee/DarkForest-20B-v2.0
athirdpath/Iambe-RP-v3-20b license_name: microsoft-research-license model-index:
name: StarDust_20B_v0.2 results:
- task: type: text-generation name: Text Generation dataset: name: AI2 Reasoning Challenge (25-Shot) type: ai2_arc config: ARC-Challenge split: test args: num_few_shot: 25 metrics:
  - type: acc_norm value: 61.01 name: normalized accuracy source: url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Evillain/StarDust_20B_v0.2 name: Open LLM Leaderboard
- task: type: text-generation name: Text Generation dataset: name: HellaSwag (10-Shot) type: hellaswag split: validation args: num_few_shot: 10 metrics:
  - type: acc_norm value: 83.76 name: normalized accuracy source: url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Evillain/StarDust_20B_v0.2 name: Open LLM Leaderboard
- task: type: text-generation name: Text Generation dataset: name: MMLU (5-Shot) type: cais/mmlu config: all split: test args: num_few_shot: 5 metrics:
  - type: acc value: 59.29 name: accuracy source: url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Evillain/StarDust_20B_v0.2 name: Open LLM Leaderboard
- task: type: text-generation name: Text Generation dataset: name: TruthfulQA (0-shot) type: truthful_qa config: multiple_choice split: validation args: num_few_shot: 0 metrics:
  - type: mc2 value: 51.43 source: url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Evillain/StarDust_20B_v0.2 name: Open LLM Leaderboard
- task: type: text-generation name: Text Generation dataset: name: Winogrande (5-shot) type: winogrande config: winogrande_xl split: validation args: num_few_shot: 5 metrics:
  - type: acc value: 77.27 name: accuracy source: url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Evillain/StarDust_20B_v0.2 name: Open LLM Leaderboard
- task: type: text-generation name: Text Generation dataset: name: GSM8k (5-shot) type: gsm8k config: main split: test args: num_few_shot: 5 metrics:
  - type: acc value: 24.03 name: accuracy source: url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Evillain/StarDust_20B_v0.2 name: Open LLM Leaderboard

Model Info

GGUF Version: StarDust_20B_v0.2-GGUF

Exllamav2 4.55bpw: Evillain/StarDust_20B_v0.2_exl2_4.55bpw

Experimental merge of 3 capable models. Main purpose was to keep the DarkForest adventuring quality while make it more consistent with prompt and context following, and also increase RP and dialogues variety.

This model is suitable for adventuring, should be fine also for storytelling, RP and ERP, but I don't have much of a time and imagination for testing.

I haven't tested this model much, actually, so it's all about an exploration of it's capabilities.

Prompt

Should work well with Alpaca format(but I'm not sure, since I don't really know what I'm doing :D)

Below is an instruction that describes a task. Write a response that appropriately completes the request.
### Instruction:
{prompt}
### Input:
{input}
### Response:

DaringMaid was tested with Undi/Ikaris SillyTavern presets for Noromaid: Context template, Instruct template, so maybe it have a sense to try my model with those templates too.

Merged

This is a merge of pre-trained language models created using mergekit.

Merge Details

My second ever merge, created in order to understand how everything works.

Firstly I merged athirdpath/Iambe-RP-v3-20b and Kooten/DaringMaid-20B, named it Dust and repeated the operation with TeeZee/DarkForest-20B-v2.0 as a base and Dust as second model.

Merge Method

This model was merged using the SLERP merge method.

Models Merged

The following models were included in the merge:

Configuration

The following YAML configuration was used to produce this model:

yaml

slices:
  - sources:
      - model: TeeZee/DarkForest-20B-v2.0
        layer_range: [0, 62]
      - model: Dust
        layer_range: [0, 62]
merge_method: slerp
base_model: TeeZee/DarkForest-20B-v2.0
parameters:
  t:
    - filter: self_attn
      value: [0, 0.5, 0.3, 0.7, 1]
    - filter: mlp
      value: [1, 0.5, 0.7, 0.3, 0]
    - value: 0.5
dtype: bfloat16
name: StarDust_20B_v0.2

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

Metric	Value
Avg.	59.46
AI2 Reasoning Challenge (25-Shot)	61.01
HellaSwag (10-Shot)	83.76
MMLU (5-Shot)	59.29
TruthfulQA (0-shot)	51.43
Winogrande (5-shot)	77.27
GSM8k (5-shot)	24.03

RichardErkhov/Evillain_-_StarDust_20B_v0.2-gguf