Modelpublic

RichardErkhov/mhenrichsen_-_hestenettetLM-gguf

sourceHugging Faceupdated 2y agoView on Hugging Face

0likes80downloads

Model Card

Quantization made by Richard Erkhov.

Github

Discord

Request more models

hestenettetLM - GGUF

—Model creator: https://huggingface.co/mhenrichsen/
—Original model: https://huggingface.co/mhenrichsen/hestenettetLM/

Name	Quant method	Size
hestenettetLM.Q2_K.gguf	Q2_K	2.09GB
hestenettetLM.IQ3_XS.gguf	IQ3_XS	2.81GB
hestenettetLM.IQ3_S.gguf	IQ3_S	2.96GB
hestenettetLM.Q3_K_S.gguf	Q3KS	2.95GB
hestenettetLM.IQ3_M.gguf	IQ3_M	3.06GB
hestenettetLM.Q3_K.gguf	Q3_K	3.28GB
hestenettetLM.Q3_K_M.gguf	Q3KM	3.28GB
hestenettetLM.Q3_K_L.gguf	Q3KL	3.56GB
hestenettetLM.IQ4_XS.gguf	IQ4_XS	3.67GB
hestenettetLM.Q4_0.gguf	Q4_0	3.83GB
hestenettetLM.IQ4_NL.gguf	IQ4_NL	3.87GB
hestenettetLM.Q4_K_S.gguf	Q4KS	2.02GB
hestenettetLM.Q4_K.gguf	Q4_K	4.07GB
hestenettetLM.Q4_K_M.gguf	Q4KM	4.07GB
hestenettetLM.Q4_1.gguf	Q4_1	4.24GB
hestenettetLM.Q5_0.gguf	Q5_0	4.65GB
hestenettetLM.Q5_K_S.gguf	Q5KS	4.65GB
hestenettetLM.Q5_K.gguf	Q5_K	4.78GB
hestenettetLM.Q5_K_M.gguf	Q5KM	4.78GB
hestenettetLM.Q5_1.gguf	Q5_1	5.07GB
hestenettetLM.Q6_K.gguf	Q6_K	5.53GB
hestenettetLM.Q8_0.gguf	Q8_0	7.17GB

Original model description: --- license: mit datasets:

—mhenrichsen/hestenettet language:
—da ---

HestenettetLM

En dansk LLM trænet på hele hestenettet over 3 epoker.

Modellen er baseret på Mistral 7b, og har et kontekstvindue på 8k.

python

from transformers import AutoTokenizer, TextStreamer, AutoModelForCausalLM

model = AutoModelForCausalLM.from_pretrained("mhenrichsen/hestenettetLM")
tokenizer = AutoTokenizer.from_pretrained("mhenrichsen/hestenettetLM")
streamer = TextStreamer(tokenizer, skip_special_tokens=True)


tokens = tokenizer(
    "Den bedste hest er en ", 
    return_tensors='pt'
)['input_ids']

# Generate output
generation_output = model.generate(
    tokens,
    streamer=streamer,
    max_length = 8194,
)

Eksempel: "Den bedste hest er en " bliver til: "Den bedste hest er en veltrænet hest."