rhundt/GLM-4-0414-32b-128k-Q4_K

rhundt/ GLM-4-0414-32b-128k-Q4_K_M:latest

1,073 Downloads Updated 1 year ago

GLM-4-0414 32B with 128k context (YaRN RoPE scaling). Needs ollama 0.6.6

tools

ollama run rhundt/GLM-4-0414-32b-128k-Q4_K_M

curl http://localhost:11434/api/chat \
  -d '{
    "model": "rhundt/GLM-4-0414-32b-128k-Q4_K_M",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

from ollama import chat

response = chat(
    model='rhundt/GLM-4-0414-32b-128k-Q4_K_M',
    messages=[{'role': 'user', 'content': 'Hello!'}],
)
print(response.message.content)

import ollama from 'ollama'

const response = await ollama.chat({
  model: 'rhundt/GLM-4-0414-32b-128k-Q4_K_M',
  messages: [{role: 'user', content: 'Hello!'}],
})
console.log(response.message.content)

Details

Updated 1 year ago

1 year ago

67a2a027906e · 20GB ·

model

archglm4

parameters32.6B

quantizationQ4_K_M

20GB

template

[gMASK]<sop>{{- /* ---------- tools section ---------- */}} {{- if .Tools }} <|system|> # Available

1.0kB

params

{ "num_ctx": 64000, "stop": [ "<|system|>", "<|user|>", "<|assistant

115B

Readme

Quantized with YaRN RoPE scaling to 128k context (factor 4). This needs Ollama >=0.6.6 to run. The num_ctx in the Modelfile defaults to 64k just because I don’t have gobs of VRAM.