I can't find how starting a IA server

[Marcounet]

Category

Nativ loads an IA module, but the application asks to start the server before to ask a question. But I searched in the app and I didn't found how to make this.

What happened

No description provided.

Diagnostics

Environment

• Nativ: 0.3.8 (202609141816)
• macOS: macOS 27.0 (Build 26A428)
• Chip: Apple M4
• Memory: 32 GB total, 11,47 GB in use
• mlx-vlm: 0.7.1

Model & Server

• Server: stopped
• Selected model: Mia-AiLab/Qwen3.8-27B-EXL3-3.5bpw
• Loaded model: —
• Max output tokens: 2048
• Context window: model default
• KV quantization: off
• Speculative decoding: off
• Prefix caching: off
• Thinking: off
• Launch arguments: --host 127.0.0.1 --port 8080 --max-tokens 2048 --model Mia-AiLab/Qwen3.8-27B-EXL3-3.5bpw

Server output (last 60 lines)

language_model.model.layers.7.self_attn.o_proj.svh,
language_model.model.layers.7.self_attn.o_proj.trellis,
language_model.model.layers.7.self_attn.q_proj.mul1,
language_model.model.layers.7.self_attn.q_proj.suh,
language_model.model.layers.7.self_attn.q_proj.svh,
language_model.model.layers.7.self_attn.q_proj.trellis,
language_model.model.layers.7.self_attn.v_proj.mul1,
language_model.model.layers.7.self_attn.v_proj.suh,
language_model.model.layers.7.self_attn.v_proj.svh,
language_model.model.layers.7.self_attn.v_proj.trellis,
language_model.model.layers.8.linear_attn.in_proj_qkv.mul1,
language_model.model.layers.8.linear_attn.in_proj_qkv.suh,
language_model.model.layers.8.linear_attn.in_proj_qkv.svh,
language_model.model.layers.8.linear_attn.in_proj_qkv.trellis,
language_model.model.layers.8.linear_attn.in_proj_z.mul1,
language_model.model.layers.8.linear_attn.in_proj_z.suh,
language_model.model.layers.8.linear_attn.in_proj_z.svh,
language_model.model.layers.8.linear_attn.in_proj_z.trellis,
language_model.model.layers.8.linear_attn.out_proj.mul1,
language_model.model.layers.8.linear_attn.out_proj.suh,
language_model.model.layers.8.linear_attn.out_proj.svh,
language_model.model.layers.8.linear_attn.out_proj.trellis,
language_model.model.layers.8.mlp.down_proj.mul1,
language_model.model.layers.8.mlp.down_proj.suh,
language_model.model.layers.8.mlp.down_proj.svh,
language_model.model.layers.8.mlp.down_proj.trellis,
language_model.model.layers.8.mlp.gate_proj.mul1,
language_model.model.layers.8.mlp.gate_proj.suh,
language_model.model.layers.8.mlp.gate_proj.svh,
language_model.model.layers.8.mlp.gate_proj.trellis,
language_model.model.layers.8.mlp.up_proj.mul1,
language_model.model.layers.8.mlp.up_proj.suh,
language_model.model.layers.8.mlp.up_proj.svh,
language_model.model.layers.8.mlp.up_proj.trellis,
language_model.model.layers.9.linear_attn.in_proj_qkv.mul1,
language_model.model.layers.9.linear_attn.in_proj_qkv.suh,
language_model.model.layers.9.linear_attn.in_proj_qkv.svh,
language_model.model.layers.9.linear_attn.in_proj_qkv.trellis,
language_model.model.layers.9.linear_attn.in_proj_z.mul1,
language_model.model.layers.9.linear_attn.in_proj_z.suh,
language_model.model.layers.9.linear_attn.in_proj_z.svh,
language_model.model.layers.9.linear_attn.in_proj_z.trellis,
language_model.model.layers.9.linear_attn.out_proj.mul1,
language_model.model.layers.9.linear_attn.out_proj.suh,
language_model.model.layers.9.linear_attn.out_proj.svh,
language_model.model.layers.9.linear_attn.out_proj.trellis,
language_model.model.layers.9.mlp.down_proj.mul1,
language_model.model.layers.9.mlp.down_proj.suh,
language_model.model.layers.9.mlp.down_proj.svh,
language_model.model.layers.9.mlp.down_proj.trellis,
language_model.model.layers.9.mlp.gate_proj.mul1,
language_model.model.layers.9.mlp.gate_proj.suh,
language_model.model.layers.9.mlp.gate_proj.svh,
language_model.model.layers.9.mlp.gate_proj.trellis,
language_model.model.layers.9.mlp.up_proj.mul1,
language_model.model.layers.9.mlp.up_proj.suh,
language_model.model.layers.9.mlp.up_proj.svh,
language_model.model.layers.9.mlp.up_proj.trellis.
ERROR:    Application startup failed. Exiting.
mlx-vlm-server stopped unexpectedly (status 3; Nativ did not request a stop)

[linear-code]

NEY-193

[Lazarus-931]

hey @Marcounet,

I believe you mean starting the server so models can be loaded, simply click the "play" button (second to right) on the bottom left of nav-bar:

As well as the top:

[Marcounet]

Thank you, but I have another problem: as soon as the server activates, it stops automatically, blocking any attempt to write a prompt for the AI. Best regards, Marc Blériot.

…

---===***===--- Marc Blériot – e-mail: ***@***.*** – mobile: (+33) 7 49 38 52 04

Le 16 sept. 2026 à 14:11, Alazer Manakelew ***@***.***> a écrit : Lazarus-931 left a comment (Blaizzy/nativ#571) <#571 (comment)> hey @Marcounet <https://github.com/Marcounet>, I believe you mean starting the server so models can be loaded, simply click the "play" button (second to right) on the bottom left of nav-bar: Image: Image (view on web) <https://github.com/user-attachments/assets/c647cab4-52df-493f-8aab-76b689f0c34b> As well as the top: Image: Image (view on web) <https://github.com/user-attachments/assets/7a52e51b-60b1-4202-83e7-b72a79a889d9> — Reply to this email directly, view it on GitHub <#571?email_source=notifications&email_token=ABSRMFKV56VDPIF42MCH3ID5PJ7N3A5CNFSNUABFM5UWIORPF5TWS5BNNB2WEL2JONZXKZKDN5WW2ZLOOQXTKNRZG4YTSMJVGMZ2M4TFMFZW63VHNVSW45DJN5XKKZLWMVXHJLDGN5XXIZLSL5RWY2LDNM#issuecomment-5697191533>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ABSRMFPAFSD5RSTWU2KWGJL5PJ7N3AVCNFSNUABGKJSXA33TNF2G64TZHMYTGMBWG42TQMRYHE5US43TOVSTWNJUG4ZTGOBSGQ4DTILWAI>. Triage notifications, keep track of coding agent tasks and review pull requests on the go with GitHub Mobile for iOS <https://github.com/notifications/mobile/ios/ABSRMFMVRF6BXWBQ5WC6UXT5PJ7N3A5CNFSNUABFM5UWIORPF5TWS5BNNB2WEL2JONZXKZKDN5WW2ZLOOQXTKNRZG4YTSMJVGMZ2M4TFMFZW63VHNVSW45DJN5XKKZLWMVXHJKTGN5XXIZLSL5UW64Y> and Android <https://github.com/notifications/mobile/android/ABSRMFPBSVPOHWXIWBH2H2T5PJ7N3A5CNFSNUABFM5UWIORPF5TWS5BNNB2WEL2JONZXKZKDN5WW2ZLOOQXTKNRZG4YTSMJVGMZ2M4TFMFZW63VHNVSW45DJN5XKKZLWMVXHJLTGN5XXIZLSL5QW4ZDSN5UWI>. Download it today! You are receiving this because you were mentioned.

[Lazarus-931]

@Marcounet if you're running the Mia-AiLab/Qwen3.8-27B-EXL3-3.5bpw model, then it won't work, it requires ExLlamaV3.

it uses "quant_method": "exl3", which mlx doesn't support

this is only if your issue is when loading this specific model :)

[Marcounet]

Ocky Rocky Best regards, Marc Blériot.

…

---===***===--- Marc Blériot – e-mail: ***@***.*** – mobile: (+33) 7 49 38 52 04

Le 16 sept. 2026 à 19:05, Alazer Manakelew ***@***.***> a écrit : Lazarus-931 left a comment (Blaizzy/nativ#571) <#571 (comment)> @Marcounet <https://github.com/Marcounet> if you're running the Mia-AiLab/Qwen3.8-27B-EXL3-3.5bpw model, then it won't work, it requires ExLlamaV3. it uses "quant_method": "exl3", which mlx doesn't support this is only if your issue is when loading this specific model :) — Reply to this email directly, view it on GitHub <#571?email_source=notifications&email_token=ABSRMFLQGRY4SJXXQRGCTI35PLB6XA5CNFSNUABFM5UWIORPF5TWS5BNNB2WEL2JONZXKZKDN5WW2ZLOOQXTKNZQGE2DANZSGUY2M4TFMFZW63VHNVSW45DJN5XKKZLWMVXHJLDGN5XXIZLSL5RWY2LDNM#issuecomment-5701407251>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ABSRMFL4EW4T6ZNTGCKGGTD5PLB6XAVCNFSNUABGKJSXA33TNF2G64TZHMYTGMBWG42TQMRYHE5US43TOVSTWNJUG4ZTGOBSGQ4DTILWAI>. Triage notifications, keep track of coding agent tasks and review pull requests on the go with GitHub Mobile for iOS <https://github.com/notifications/mobile/ios/ABSRMFJHQY24OH4JSQJ5NA35PLB6XA5CNFSNUABFM5UWIORPF5TWS5BNNB2WEL2JONZXKZKDN5WW2ZLOOQXTKNZQGE2DANZSGUY2M4TFMFZW63VHNVSW45DJN5XKKZLWMVXHJKTGN5XXIZLSL5UW64Y> and Android <https://github.com/notifications/mobile/android/ABSRMFPCS5ABWVD2DKSRGCL5PLB6XA5CNFSNUABFM5UWIORPF5TWS5BNNB2WEL2JONZXKZKDN5WW2ZLOOQXTKNZQGE2DANZSGUY2M4TFMFZW63VHNVSW45DJN5XKKZLWMVXHJLTGN5XXIZLSL5QW4ZDSN5UWI>. Download it today! You are receiving this because you were mentioned.