\Lumina-Next-T2I\config.json' is not a valid JSON file.

[danieldietzel]

I tried following the instructions here:

https://huggingface.co/Alpha-VLLM/Lumina-Next-T2I

And installed the repo via hugginface CLI and via github, in both cases I get this error:

[rank0]: OSError: It looks like the config file at '........\Lumina-T2X\Lumina-Next-T2I\config.json' is not a valid JSON file.

Using this command:

lumina_next infer -c "lumina_next_t2i/configs/infer/settings.yaml" "a snowman" "./outputs"

in my settings.yaml I've tried the local path, huggingface path, and repo download path.

[PommesPeter]

Hi @danieldietzel ,
Could you provide your yaml config? we will help you check the correctness

[danieldietzel]

Hi @danieldietzel , Could you provide your yaml config? we will help you check the correctness

Hi Pommes,

I ran through the steps again and realized the second item in this config is supposed to be the LLM checkpoint, not the lumina model.

https://github.com/Alpha-VLLM/Lumina-T2X/blob/main/lumina_next_t2i/configs/infer/settings.yaml

Now my Yaml is:

• settings:

  model:
  ckpt: 'C:\models\lumina'
  ckpt_lm: 'C:\models\gemma'
  token: ""

  transport:
  path_type: "Linear" # option: ["Linear", "GVP", "VP"]
  prediction: "velocity" # option: ["velocity", "score", "noise"]
  loss_weight: "velocity" # option: [None, "velocity", "likelihood"]
  sample_eps: 0.1
  train_eps: 0.2

  ode:
  atol: 1e-6 # Absolute tolerance
  rtol: 1e-3 # Relative tolerance
  reverse: false # option: true or false
  likelihood: false # option: true or false

  infer:
  resolution: "1024x1024" # option: ["1024x1024", "512x2048", "2048x512", "(Extrapolation) 1664x1664", "(Extrapolation) 1024x2048", "(Extrapolation) 2048x1024"]
  num_sampling_steps: 60 # range: 1-1000
  cfg_scale: 4. # range: 1-20
  solver: "euler" # option: ["euler", "dopri5", "dopri8"]
  t_shift: 4 # range: 1-20 (int only)
  ntk_scaling: true # option: true or false
  proportional_attn: true # option: true or false
  seed: 0 # rnage: any number

But I get this:
TypeError: NextDiT.forward_with_cfg() got an unexpected keyword argument 'ntk_factor'
[rank0]: AttributeError: 'NoneType' object has no attribute 'float'

I am on Windows by the way if it helps. Had to change all dist.init_process_group("nccl") to dist.init_process_group("gloo") to get this far not sure if that breaks things.

[PommesPeter]

This may not impact performance, but we have not tested whether it can run correctly on the Gloo backend. you could try running the mini version of Lumina-Next-T2I on https://github.com/Alpha-VLLM/Lumina-T2X/tree/main/lumina_next_t2i_mini