Hi, thanks for this exciting repo!
I switched from using qwen3.6-35B-A3-4bit served by LM Studio on mac, to serving the same model using dflash-mlx.
However, I get output like this:
<model_settings>
<request_snapshot>false</request_snapshot>
<skip_nonessential_tool_output>false</skip_nonessential_tool_output>
</model_settings><|im_end|>
And this:
<model_settings>
<always_apply_skills>
</always_apply_skills>
</model_settings><|im_end|>
Is this due to tool calls not working, or something else?
The previous setup (hermes + LM studio) with the same model has no such issues.
TIA.
Hi, thanks for this exciting repo!
I switched from using qwen3.6-35B-A3-4bit served by LM Studio on mac, to serving the same model using dflash-mlx.
However, I get output like this:
<model_settings>
<request_snapshot>false</request_snapshot>
<skip_nonessential_tool_output>false</skip_nonessential_tool_output>
</model_settings><|im_end|>
And this:
<model_settings>
<always_apply_skills>
</always_apply_skills>
</model_settings><|im_end|>
Is this due to tool calls not working, or something else?
The previous setup (hermes + LM studio) with the same model has no such issues.
TIA.