37 lines
1.3 KiB
Plaintext
37 lines
1.3 KiB
Plaintext
# Ollama allows cross-origin requests from 127.0.0.1 and 0.0.0.0 by default.
|
|
# Additional origins can be configured with OLLAMA_ORIGINS.
|
|
# export OLLAMA_ORIGINS="<ip>"
|
|
# Allow connections from any origin (or specify exact domains/IPs like "http://192.168.1.50:*")
|
|
# export OLLAMA_ORIGINS="*"
|
|
# Explicitly bind to all network interfaces (needed if running behind a proxy or remote API access)
|
|
# export OLLAMA_HOST="0.0.0.0"
|
|
|
|
# log to syslog
|
|
# output_logger="logger -t \"$RC_SVCNAME\" -p daemon.info"
|
|
# error_logger="logger -t \"$RC_SVCNAME\" -p daemon.err"
|
|
|
|
# log to file
|
|
output_log="/var/log/ollama/ollama.log"
|
|
error_log="/var/log/ollama/ollama.log"
|
|
|
|
# Set the default context size - default is 4096 token
|
|
# export OLLAMA_CONTEXT_LENGTH=8192
|
|
# export OLLAMA_CONTEXT_LENGTH=16384
|
|
# export OLLAMA_CONTEXT_LENGTH=32768
|
|
# export OLLAMA_CONTEXT_LENGTH=131072
|
|
# export OLLAMA_CONTEXT_LENGTH=262144
|
|
|
|
# Optimizations
|
|
#
|
|
# Enable Flash Attention (drastically reduces VRAM usage for long contexts)
|
|
# export OLLAMA_FLASH_ATTENTION=true
|
|
# Compress the Key-Value cache (saves VRAM at a tiny precision cost) default is f16
|
|
# export OLLAMA_KV_CACHE_TYPE=q4_0
|
|
#
|
|
# Number of parallel model requests.
|
|
# export OLLAMA_NUM_PARALLEL=1
|
|
#
|
|
# export OLLAMA_VULKAN=false
|
|
# export OLLAMA_DEBUG=INFO
|
|
# export OLLAMA_DEBUG=1
|