← Docs Home · Model Catalog · HTTP API · Troubleshooting

Chroma1 Base LoRA Inference with Diffusers (AI Toolkit-trained)

API model id: chroma
URL slug: chroma

This page documents the reference Diffusers inference pipeline for chroma (Chroma1 Base). It is designed for running LoRAs trained with ostris/ai-toolkit while minimizing training preview vs inference mismatch. If you are trying to reproduce AI Toolkit sample previews, treat the code linked below as the source of truth (scheduler wiring, resolution snapping, LoRA application, and conditioning).

Run in the cloud (optional): If you want to reproduce the examples on this page in a pinned runtime without local CUDA/driver setup (and reduce preview‑vs‑inference drift), run it via RunComfy’s Cloud AI Toolkit (Train + Inference). 👉 You can open it here: Cloud AI Toolkit (Train + Inference)

Quick facts

Field	Value
Pipeline	src/pipelines/chroma.py
Base checkpoint	lodestones/Chroma1-Base
Defaults	sample_steps=25, guidance_scale=4.0, seed=42
Resolution snapping	Floors width/height to a multiple of 32
Control image	No
LoRA scale behavior	Merged into the transformer before float8 quantization. Scale is fixed after load.
Needs AI Toolkit	Required (needs a local ostris/ai-toolkit checkout via AI_TOOLKIT_PATH)

Reference implementation (source of truth)

• Pipeline implementation: src/pipelines/chroma.py
• Shared behaviors (snapping, seeding, LoRA base logic): src/pipelines/base.py
• Request schema (parameter names): src/schemas/request.py
• Model ids (enum): src/schemas/models.py
• Pipeline registry (model → class): src/pipelines/__init__.py
• Executor (prompt processing + caching): src/tasks/executor.py

Minimal API request

{
  "model": "chroma",
  "trigger_word": "sks",
  "prompts": [
    {
      "prompt": "[trigger] a photo of a person",
      "width": 1024,
      "height": 1024,
      "seed": 42,
      "sample_steps": 25,
      "guidance_scale": 4.0,
      "neg": ""
    }
  ],
  "loras": [
    {
      "path": "my_lora_job/my_lora.safetensors",
      "network_multiplier": 1.0
    }
  ]
}

Pipeline behavior that matters

• Loads a custom Chroma transformer from AI Toolkit and builds an AITK ChromaPipeline wrapper.
• Quantizes the transformer and T5 text encoder to float8 (AI Toolkit torchao path).
• Uses a CustomFlowMatchEulerDiscreteScheduler with dynamic shifting (AI Toolkit sampler).
• Prompt embeddings are computed via T5; the negative_prompt string is not used (unconditional embeds are always empty-string).

Preview-matching notes (training preview vs inference mismatch)

• LoRA scale is baked in: the server merges LoRA weights into the transformer and then quantizes. If you change loras[].network_multiplier, the executor reloads the pipeline for that scale.
• Quantization is part of the reference behavior. If you run Chroma in full precision elsewhere, results can shift even with identical steps/seed.
• The server always uses an empty-string unconditional embedding for the negative path; passing a non-empty neg will not change the unconditional embedding.
• Width/height are floored to a multiple of 32 (resolution bucket divisibility).

What to compare when debugging mismatch

• Confirm the effective width/height after snapping (the server floors to the divisor).
• Match sample_steps and the scheduler family (FlowMatch / UniPC / DDPM differences matter).
• Match guidance_scale semantics (some pipelines map it to a different internal parameter).
• Match loras[].network_multiplier and whether LoRA scale is dynamic vs fused.

Related

• See the Model Catalog for all supported models.