Quiet Lantern 58

Native Diffusers export of a trained FLUX.2 Klein Base 4B checkpoint.

Provenance

Source experiment: flux2_klein_base_4b_diffusionnft_dvreward_prompt_rubric_v4_16prompts_group14_7train_1dvreward_tp1_2node_512px_20step_cfg4_nrt
Formal run: flux2-v4-direct-formal-00d00171
Source checkpoint: models/step_500.pt
Global step: 500
Export weights: EMA
Source checkpoint SHA-256: a17d2bc35cb4cae69329aabe6a1fe4c34880acc6513b95652b1622cfcc46d62f
Base model revision: a3b4f4849157f664bdbc776fd7453c2783562f4d
LoRA rank / alpha: 32 / 64
Training profile: prompt-rubric v4 DVReward, 512px, 20 rollout steps, CFG 4
Export format: diffusers.Flux2KleinPipeline, BF16, max shard size 1GB

Verification

The exporter matched and merged 60 default LoRA pairs. Relative to the Base transformer, 60 tensors and 2,426,981,339 elements changed; maximum absolute parameter delta was 0.0089111328125. Offline reload reproduced the same deltas. The fixed-seed 512px smoke image differed from Base in 772,373 channel values. The checkpoint's 120 _diffusionnft_old_policy_#### rollout buffers were recorded and excluded because they are training-only state rather than Diffusers transformer weights.

See export_manifest.json, verification.json, and artifact_checksums.sha256 for machine-readable provenance and integrity data.

Loading

import torch
from diffusers import Flux2KleinPipeline

pipe = Flux2KleinPipeline.from_pretrained(
    "kimi000/quiet-lantern-58",
    torch_dtype=torch.bfloat16,
)
pipe.to("cuda")
image = pipe(
    "A cinematic photograph of a red fox walking through a snowy forest",
    num_inference_steps=20,
    guidance_scale=4.0,
).images[0]

Downloads last month: 8

Safetensors

Model size

4B params

Tensor type

BF16