LTX-2

​

1. Model introduction

• Native model package, not diffusers. miles/backends/fsdp_utils/models/ltx/ supplies loading, modeling, attention, and the FSDP plan directly against ltx_core; FSDP wraps BasicAVTransformerBlock. Requires ltx-core (pinned in requirements.txt, baked into the Docker image).
• Unguided training. supports_cfg_training = False: any --diffusion-guidance-scale other than 1.0, or a --diffusion-negative-prompt, is a parse-time error. The training forward is a single velocity pass — there is no negative branch to combine.
• Video branch only. The audio stream is loaded but never trained (optimizer_state_allowed_missing = ["audio"]), and --rollout-patch-group ltx disables audio-video cross-attention on the engine to match.
• No sequence parallelism. sequence_parallel_plan raises NotImplementedError; leave --sequence-parallel-size at 1.

​

2. Supported variants

​

3. Family config

​

4. Launch

python3 scripts/run_diffusion_grpo_ltx23_sglang.py

​

5. Recipe notes

• The step strategy is epoch_global_random_choice: 3 SDE steps drawn once per epoch from candidates 0-9 and shared by every sample — with only 64 samples per rollout, per-request randomization would add variance the batch is too small to average out.
• The clip range is 1e-5, an order of magnitude tighter than the image recipes’ 1e-4.
• --pickscore-num-frames 3 scores three evenly-spaced frames per video; scoring all 57 would dominate rollout time for a reward signal that barely changes.
• Health checks are relaxed (--rollout-health-check-interval 120, router failure threshold 30): a 57-frame request takes minutes, and the image-recipe defaults would kill healthy engines mid-generation.
• --rollout-parser-num-workers 8: 57-frame trajectory tensors are large enough that one deserializer actor starves the engines.

​

6. Reference results

​

7. Pairs well with

• Dtype Control — the boundary-dtype policy is why LTX has one.
• Deterministic Training — sdpa_math is the deterministic-safe backend this recipe relies on.