Astrorocketboros-3B-GGUF/README.md

base_model, inference, language, library_name, license, model_creator, model_name, pipeline_tag, quantized_by, tags

base_model	inference	language	library_name	license	model_creator	model_name	pipeline_tag	quantized_by	tags
Aryanne/Astrorocketboros-3B	false	en	transformers	cc-by-sa-4.0	Aryanne	Astrorocketboros-3B	text-generation	afrideva	gpt llm stablelm gguf ggml quantized q2_k q3_k_m q4_k_m q5_k_m q6_k q8_0

Aryanne/Astrorocketboros-3B-GGUF

Quantized GGUF model files for Astrorocketboros-3B from Aryanne

Name	Quant method	Size
astrorocketboros-3b.fp16.gguf	fp16	5.59 GB
astrorocketboros-3b.q2_k.gguf	q2_k	1.20 GB
astrorocketboros-3b.q3_k_m.gguf	q3_k_m	1.39 GB
astrorocketboros-3b.q4_k_m.gguf	q4_k_m	1.71 GB
astrorocketboros-3b.q5_k_m.gguf	q5_k_m	1.99 GB
astrorocketboros-3b.q6_k.gguf	q6_k	2.30 GB
astrorocketboros-3b.q8_0.gguf	q8_0	2.97 GB

Original Model Card:

This model is a task_arithmetic merge of pansophic/rocket-3B, jondurbin/airoboros-3b-3p0 and Aryanne/Astrohermes-3B, as shown in the yaml(see Astrorocketboros.yml or below).

I used this model as base ayoubkirouane/StableLM-3B

I'm not sure if all the .json files are all right, but it seems to work at the moment.

merge_method: task_arithmetic
base_model: ayoubkirouane/StableLM-3B
models:
  - model: ayoubkirouane/StableLM-3B
  - model: pansophic/rocket-3B
    parameters:
      weight: 1.0
  - model: Aryanne/Astrohermes-3B
    parameters:
      weight: 0.22
  - model: jondurbin/airoboros-3b-3p0
    parameters:
      weight: 0.1
dtype: float16

I recommend the use of chatml prompt format, but alpaca seems to work too but it's necessary to write something before the instruction:

You are an Assistant
### Instruction:
write a poem.
### Response:
Amidst the tranquil hues of twilight's hue,
As shadows stretch and dance with whispered dew,
The trees weave tales of ages past untold,

GGUF Quants: notyet