Any-to-Any
MLX
Safetensors
English
gemma4
3-bit
heretic
uncensored
decensored
abliterated
ara
4-bit precision
Instructions to use zecanard/gemma-4-E2B-it-uncensored-heretic-ara-MLX-3bit-mixed_3_6 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- MLX
How to use zecanard/gemma-4-E2B-it-uncensored-heretic-ara-MLX-3bit-mixed_3_6 with MLX:
# Download the model from the Hub pip install huggingface_hub[hf_xet] huggingface-cli download --local-dir gemma-4-E2B-it-uncensored-heretic-ara-MLX-3bit-mixed_3_6 zecanard/gemma-4-E2B-it-uncensored-heretic-ara-MLX-3bit-mixed_3_6
- Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- LM Studio
Jinja Prompt fix for Tool use not working in LM-Studio
#1
by panclamor - opened
Solution to tool call error that tested and works.
Seems the other zecanard gemma prompts are missing the same block.
Received error when trying to use tools:
Failed to regenerate messageError rendering prompt with jinja template: "Cannot call something that is not a function: got UndefinedValue".
Found solution is to update the jinja prompt template with:
{%- macro format_type_argument(type_value) -%}
{%- if type_value is string -%}
{{- type_value | lower -}}
{%- elif type_value is iterable -%}
{{- '[' -}}
{%- for t in type_value -%}
{{- t | lower -}}
{%- if not loop.last %},{% endif -%}
{%- endfor -%}
{{- ']' -}}
{%- else -%}
{{- type_value -}}
{%- endif -%}
{%- endmacro -%}
Needs to be added to jinja prompt before:
{%- macro strip_thinking(text) -%}
Doh! I wish I would have known this before I re-synced all my Gemma 4s with Google's chat template. Thanks for the heads up. I have updated my quants.