mannix/llava-phi3:iq3

mannix/ llava-phi3:iq3_s

2,014 Downloads Updated 2 years ago

A new small LLaVA model fine-tuned from Phi 3 Mini [I-Quants]

vision

ollama run mannix/llava-phi3:iq3_s

curl http://localhost:11434/api/chat \
  -d '{
    "model": "mannix/llava-phi3:iq3_s",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

from ollama import chat

response = chat(
    model='mannix/llava-phi3:iq3_s',
    messages=[{'role': 'user', 'content': 'Hello!'}],
)
print(response.message.content)

import ollama from 'ollama'

const response = await ollama.chat({
  model: 'mannix/llava-phi3:iq3_s',
  messages: [{role: 'user', content: 'Hello!'}],
})
console.log(response.message.content)

Details

Updated 2 years ago

2 years ago

3b2c62cb62ae · 2.3GB ·

model

archllama

parameters3.82B

quantizationIQ3_S

1.7GB

projector

archclip

parameters303M

quantizationF16

608MB

params

{ "num_ctx": 4096, "num_keep": 4, "stop": [ "<|user|>", "<|assistant|>",

155B

template

{{ if .System }}<|system|> {{ .System }}<|end|> {{ end }}{{ if .Prompt }}<|user|> {{ .Prompt }}<|end

149B

Readme

llava-phi3 is a LLaVA model fine-tuned from Phi 3 Mini 4k, with strong performance benchmarks on par with the original LLaVA model:

References

Hugging Face

GitHub