RichardErkhov/Xenon1_-_MetaModel_moex8-gguf

Quantization made by Richard Erkhov.

Github

Discord

Request more models

MetaModel_moex8 - GGUF

• Model creator: https://huggingface.co/Xenon1/
• Original model: https://huggingface.co/Xenon1/MetaModel_moex8/

Name	Quant method	Size
MetaModel_moex8.Q2_K.gguf	Q2_K	24.11GB
MetaModel_moex8.IQ3_XS.gguf	IQ3_XS	26.95GB
MetaModel_moex8.IQ3_S.gguf	IQ3_S	28.46GB
MetaModel_moex8.Q3_K_S.gguf	Q3_K_S	28.46GB
MetaModel_moex8.IQ3_M.gguf	IQ3_M	29.86GB
MetaModel_moex8.Q3_K.gguf	Q3_K	31.42GB
MetaModel_moex8.Q3_K_M.gguf	Q3_K_M	31.42GB
MetaModel_moex8.Q3_K_L.gguf	Q3_K_L	33.69GB
MetaModel_moex8.IQ4_XS.gguf	IQ4_XS	19.56GB
MetaModel_moex8.Q4_0.gguf	Q4_0	36.85GB
MetaModel_moex8.IQ4_NL.gguf	IQ4_NL	37.28GB
MetaModel_moex8.Q4_K_S.gguf	Q4_K_S	37.28GB
MetaModel_moex8.Q4_K.gguf	Q4_K	11.76GB
MetaModel_moex8.Q4_K_M.gguf	Q4_K_M	6.62GB
MetaModel_moex8.Q4_1.gguf	Q4_1	32.46GB
MetaModel_moex8.Q5_0.gguf	Q5_0	44.93GB
MetaModel_moex8.Q5_K_S.gguf	Q5_K_S	44.93GB
MetaModel_moex8.Q5_K.gguf	Q5_K	46.33GB
MetaModel_moex8.Q5_K_M.gguf	Q5_K_M	46.33GB
MetaModel_moex8.Q5_1.gguf	Q5_1	48.97GB
MetaModel_moex8.Q6_K.gguf	Q6_K	53.51GB
MetaModel_moex8.Q8_0.gguf	Q8_0	69.19GB

Original model description:

license: apache-2.0 tags: - moe - mergekit - merge - chinese - arabic - english - multilingual - german - french - gagan3012/MetaModel - jeonsworld/CarbonVillain-en-10.7B-v2 - jeonsworld/CarbonVillain-en-10.7B-v4 - TomGrc/FusionNet_linear - DopeorNope/SOLARC-M-10.7B - VAGOsolutions/SauerkrautLM-SOLAR-Instruct - upstage/SOLAR-10.7B-Instruct-v1.0 - fblgit/UNA-SOLAR-10.7B-Instruct-v1.0

MetaModel_moex8

This model is a Mixure of Experts (MoE) made with mergekit (mixtral branch). It uses the following base models:

• gagan3012/MetaModel
• jeonsworld/CarbonVillain-en-10.7B-v2
• jeonsworld/CarbonVillain-en-10.7B-v4
• TomGrc/FusionNet_linear
• DopeorNope/SOLARC-M-10.7B
• VAGOsolutions/SauerkrautLM-SOLAR-Instruct
• upstage/SOLAR-10.7B-Instruct-v1.0
• fblgit/UNA-SOLAR-10.7B-Instruct-v1.0

🧩 Configuration

dtype: bfloat16
experts:
- positive_prompts:
  - ''
  source_model: gagan3012/MetaModel
- positive_prompts:
  - ''
  source_model: jeonsworld/CarbonVillain-en-10.7B-v2
- positive_prompts:
  - ''
  source_model: jeonsworld/CarbonVillain-en-10.7B-v4
- positive_prompts:
  - ''
  source_model: TomGrc/FusionNet_linear
- positive_prompts:
  - ''
  source_model: DopeorNope/SOLARC-M-10.7B
- positive_prompts:
  - ''
  source_model: VAGOsolutions/SauerkrautLM-SOLAR-Instruct
- positive_prompts:
  - ''
  source_model: upstage/SOLAR-10.7B-Instruct-v1.0
- positive_prompts:
  - ''
  source_model: fblgit/UNA-SOLAR-10.7B-Instruct-v1.0
gate_mode: hidden

💻 Usage

!pip install -qU transformers bitsandbytes accelerate

from transformers import AutoTokenizer
import transformers
import torch

model = "gagan3012/MetaModel_moex8"

tokenizer = AutoTokenizer.from_pretrained(model)
pipeline = transformers.pipeline(
    "text-generation",
    model=model,
    model_kwargs={"torch_dtype": torch.float16, "load_in_4bit": True},
)

messages = [{"role": "user", "content": "Explain what a Mixture of Experts is in less than 100 words."}]
prompt = pipeline.tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
outputs = pipeline(prompt, max_new_tokens=256, do_sample=True, temperature=0.7, top_k=50, top_p=0.95)
print(outputs[0]["generated_text"])