me/env-ansible
1
1
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

llm, further reduce contexts based on testing, now using 2.5:14b for Complete successfully alongside the RAM needs of 3:30b on Cline.

Browse Source
This commit is contained in:
me
2026-10-10 16:42:19 -07:00
parent 1f16053042
commit 019066c4dc
1 changed files with 5 additions and 5 deletions
Download Patch File Download Diff File Expand all files Collapse all files
tasks/workstation/linux/software/llm.yml
+5 -5
View File
@@ -62,8 +62,8 @@
# Conversely: # Conversely:
# - qwen3.5:4b would be 3.3GB * 1.3 = 4290, rather than 4000. # - qwen3.5:4b would be 3.3GB * 1.3 = 4290, rather than 4000.
llm_size_coef: 1.3 llm_size_coef: 1.3
# Take the GB of RAM and use it as the MB of context. # Use the MB of RAM as the context, with a maximum to prevent swapping.
llm_context: "{{ [(ansible_memtotal_mb | round(-3, method='ceil')), 32000] | max | int }}" llm_context: "{{ [(ansible_memtotal_mb | round(-3, method='ceil')), 32000] | min | int }}"
- name: Workstation | Linux | Software | LLMs | Facts 02 - name: Workstation | Linux | Software | LLMs | Facts 02
ansible.builtin.set_fact: ansible.builtin.set_fact:
@@ -591,7 +591,7 @@
[providers.models.ollama.{{ llm.provider }}] [providers.models.ollama.{{ llm.provider }}]
model = "{{ llm.model }}" model = "{{ llm.model }}"
timeout_secs = 3600 # 1 hour, default is 120 (2 minutes), issue happens between 5 and 6 minutes though. timeout_secs = 3600 # 1 hour, default is 120 (2 minutes), issue happens between 5 and 6 minutes though.
context_window = {{ [llm_context, llm.context, 24000] | min | int }} context_window = {{ [llm_context, llm.context, 16000] | min | int }}
max_tokens = {{ [llm_tokens, (llm.context / 4), 4000] | min | int }} max_tokens = {{ [llm_tokens, (llm.context / 4), 4000] | min | int }}
{% endif %} {% endif %}
@@ -625,14 +625,14 @@
provider: ollama provider: ollama
model: {{ llm.model }} model: {{ llm.model }}
defaultCompletionOptions: defaultCompletionOptions:
contextLength: {{ [llm_context, llm.context, 32000] | min | int }} contextLength: {{ [llm_context, llm.context, 24000] | min | int }}
maxTokens: {{ [llm_tokens, (llm.context / 4), 8000] | min | int }} maxTokens: {{ [llm_tokens, (llm.context / 4), 8000] | min | int }}
- name: {{ llm.provider }}_autocomplete - name: {{ llm.provider }}_autocomplete
provider: ollama provider: ollama
model: {{ llm.model }} model: {{ llm.model }}
defaultCompletionOptions: defaultCompletionOptions:
contextLength: {{ [llm_context, llm.context, 16000] | min | int }} contextLength: {{ [llm_context, llm.context, 12000] | min | int }}
maxTokens: {{ [llm_tokens, (llm.context / 4), 2000] | min | int }} maxTokens: {{ [llm_tokens, (llm.context / 4), 2000] | min | int }}
roles: roles:
- autocomplete - autocomplete