nordavind-7b-instruct-adapter

Nordavind-7B

A model trained on top of normistral-7b (warm) on various Norwegian instruct data with a context length of 2048. The name is derived from Mistral.

Trained on a V100 32GB. See example outputs below!

Prompt format:

<s>{system_prompt} [INST] {inst} [/INST] \\n {res} </s>

Inference:

<s>{system_prompt} [INST] {inst} [/INST] \\n *YOUR_PROMPT*

Running the model

Suggested (PEFT adapter):

import torch
from transformers import AutoModelForCausalLM, AutoTokenizer
BASE_MODEL = "norallm/normistral-7b-warm"
ADAPTER = "tollefj/nordavind-7b-instruct-adapter"
model = AutoModelForCausalLM.from_pretrained(
    ADAPTER,
    device_map="auto",
    return_dict=True,
    torch_dtype=torch.float16,
)
tokenizer = AutoTokenizer.from_pretrained(BASE_MODEL)

or the full model (will be updated less frequently):

from transformers import pipeline
pipe = pipeline("text-generation", model="tollefj/nordavind-7b-instruct-warm")

See train.py for training details and util.py for the prompt format. Here you can import the generate(...) function for inference. So far, the model has seen better results with greedy sampling.

TODO:

• Define more tasks/templates for generation

Things it can do...

• Translation:
  • "Oversett fra engelsk til norsk"
  • "Oversett frasen ''tekst'' til språk
• Listing:
  • Lag en liste over ...
  • Ranger en liste med fire elementer fra lavest til høyest...
• Explanations:
  • Beskriv en av fordelene ved...
  • Forklar hvordan ...
  • Kan du tenke på ...
  • Beskrive konseptet ...
  • Hvordan kan ...
• Tasks:
  • Du vil få en oppgave. Du må generere et detaljert og langt svar. ...
  • Bruker vil du gi deg en oppgave. Målet ditt er å fullføre oppgaven så trofast du kan. Mens du utfører oppgaven, tenk trinn for trinn og begrunn trinnene dine.
• Generation
  • Finn ...
  • Generer ...
• Summarization
  • Oppsummer ...
  • Gi en kort oppsummering av ...
  • Lag en kort oppsummering av teksten: ...

Examples

#https://www.nrk.no/nordland/dayo-i-bodo-fekk-jobb-i-siste-augeblikk-1.16753304
text = "Dayo måtte ut av landet om han ikkje fekk rett jobb på tre veker. Advokat meiner få utanlandske studentar forstår kor viktig val av studie er."

generate(
    model,
    tokenizer,
    text,
    task="create-title",
    tokens=50,
    first_sent=True
)

Utenlandske studenter må velge riktig studie for å få jobb i Norge

generate("Lag en skikkelig norsk setning", tokens=50, first_sent=True)

Jeg liker å gå på tur i skogen.

generate("Lag en liste med tre forskjellige bruksområder (inkludert eksempler) for kunstig intelligens.")

1. Automatisering av oppgaver: AI kan brukes til å automatisere oppgaver som å svare på kundehenvendelser, behandle medisinske bilder og diagnostisere sykdommer, og til og med kjøre biler.
2. Forbedre kundeservice: AI kan brukes for å forbedre kundeservicen ved å gi personlige anbefalinger, gi raskere svar og gi mer nøyaktige svar.
3. Forbedre sikkerhet: AI kan brukes i sikkerhetssystemer for å oppdage og forhindre cyberangrep, oppdage svindel og oppdage mistenkelige aktiviteter.

generate("Hvem tror du vinner valget i Bergen?", tokens=60)

Det er vanskelig å forutsi hvem som vinner valget i Bergen, men det er sannsynlig at enten Arbeiderpartiet eller Høyre vil vinne. Arbeiderpartiet har hatt makten i Bergen i mange år, men Høyre har hatt en sterk tilstedeværelse i byen de siste årene.

generate("Forklar dyp læring til en femteklassing")

Dyp læring er en type maskinlæring som bruker nevrale nettverk for å lære komplekse mønstre i data. Det innebærer å trene en datamaskin til å gjenkjenne mønstre i data og ta beslutninger basert på disse mønstrene.