LightX2V/configs · li/LightX2V

GGitHubupdate offline quantization for realtimewam (#1605)

文件	最后提交记录	最后更新时间
attentions	refactor: simplify inference configs and example scripts(v1) (#1560)	22 天前
bagel	refactor: simplify inference configs and example scripts(v1) (#1560)	22 天前
bench	refactor: simplify inference configs and example scripts(v1) (#1560)	22 天前
caching	refactor: simplify inference configs and example scripts(v1) (#1560)	22 天前
changing_resolution	refactor: simplify inference configs and example scripts(v1) (#1560)	22 天前
cosmos3	refactor: simplify inference configs and example scripts(v1) (#1560)	22 天前
deploy	refactor: simplify inference configs and example scripts(v2) (#1580)	8 天前
disagg	refactor: simplify inference configs and example scripts(v2) (#1580)	8 天前
dist_infer	refactor: simplify inference configs and example scripts(v2) (#1580)	8 天前
distill	refactor: simplify config loading and Wan/MiniMax-H3 examples (#1604)	4 天前
dreamzero	refactor: simplify inference configs and example scripts(v2) (#1580)	8 天前
ernie_image	SKILL-guided, Codex-generated update: add native ERNIE-Image / ERNIE-Image-Turbo support to LightX2V inference (#1133) --------- Co-authored-by: helloyongyang <yongyang1030@163.com>	4 个月前
fastwam	refactor: simplify inference configs and example scripts(v2) (#1580)	8 天前
flux2	refactor: simplify inference configs and example scripts(v2) (#1580)	8 天前
hidream_o1_image	refactor: simplify inference configs and example scripts(v2) (#1580)	8 天前
hunyuan3d	Optimize hunyuan3d (#1240) Improve Hunyuan3D inference performance with reduced transformer sync overhead, FlashInfer MoE, fused QK RMSNorm, FP8 DiT support, and fused self-attention QKV projection. MoE FP8 remains disabled to avoid accuracy risk.	2 个月前
hunyuan_image3	refactor: simplify inference configs and example scripts(v2) (#1580)	8 天前
hunyuan_video_15	refactor: simplify inference configs and example scripts(v2) (#1580)	8 天前
infinitetalk	refactor(api)!: unify inference parameters and remove RIFE (#1511)	1 个月前
lingbot	refactor: simplify inference configs and example scripts(v2) (#1580)	8 天前
lingbot_fast	refactor: simplify inference configs and example scripts(v2) (#1580)	8 天前
lingbot_va	refactor: simplify inference configs and example scripts(v2) (#1580)	8 天前
lingbot_video	refactor: simplify inference configs and example scripts(v2) (#1580)	8 天前
longcat_image	refactor: simplify inference configs and example scripts(v2) (#1580)	8 天前
ltx2	refactor: simplify inference configs and example scripts(v3) (#1592)	6 天前
matrix_game2	refactor: simplify inference configs and example scripts(v3) (#1592)	6 天前
matrix_game3	refactor: simplify inference configs and example scripts(v3) (#1592)	6 天前
meanflow	refactor: simplify inference configs and example scripts(v3) (#1592)	6 天前
minimax_h3	refactor: simplify config loading and Wan/MiniMax-H3 examples (#1604)	4 天前
minimax_h3_causal	refactor: simplify inference configs and example scripts(v3) (#1592)	6 天前
motus	refactor: simplify inference configs and example scripts(v3) (#1592)	6 天前
neopp	refactor(neopp): make KV replay inputs explicit (#1494)	1 个月前
offload	refactor: simplify config loading and Wan/MiniMax-H3 examples (#1604)	4 天前
platforms	Support qwen-image-2.1-turbo for macbook mps (#1602)	5 天前
quantization	refactor(api)!: unify inference parameters and remove RIFE (#1511)	1 个月前
qwen_image	refactor: simplify config loading and Wan/MiniMax-H3 examples (#1604)	4 天前
qwen_image_21	Support Qwen-Image-2.1-Turbo (#1601)	5 天前
realtimewam	update offline quantization for realtimewam (#1605)	4 天前
seedvr	refactor: simplify inference configs and example scripts(v3) (#1592)	6 天前
seko_talk	refactor(sp): finalize explicit sequence-parallel attention API (#1531) ## Summary Finalize the public Ulysses/Ring attention API and migrate all reachable consumers to the explicit main/aux contract. ## Changes - Replace the legacy `apply` API with the finalized explicit contract. - Cover main-only, full-QKV, Q-only auxiliary-KV, and dynamic layouts. - Preserve model-specific RoPE, token ordering, cache, and output reconstruction. - Unify FP8/FP4 communication configuration under `seq_p_quant_scheme`.	28 天前
self_forcing	refactor: simplify inference configs and example scripts(v3) (#1592)	6 天前
sparse_attn	refactor: simplify inference configs and example scripts(v3) (#1592)	6 天前
swiftvr	perf(swiftvr): reduce ReAE memory usage and add deployment guides (#1547)	25 天前
wan	refactor: simplify config loading and Wan/MiniMax-H3 examples (#1604)	4 天前
wan22	refactor: simplify inference configs and example scripts(v3) (#1592)	6 天前
wan22_vace	refactor: simplify inference configs and example scripts(v3) (#1592)	6 天前
wan_dancer	refactor(api)!: unify inference parameters and remove RIFE (#1511)	1 个月前
worldmirror	refactor: simplify inference configs and example scripts(v3) (#1592)	6 天前
worldplay	refactor: simplify inference configs and example scripts(v3) (#1592)	6 天前
z_image	refactor: simplify inference configs and example scripts(v3) (#1592)	6 天前
model_pipeline.json	refactor: simplify inference configs and example scripts(v3) (#1592)	6 天前
mooncake_config.json	Add disaggregation feature (#893) # Enable Disaggregation Feature ## Summary This PR introduces a disaggregation architecture to LightX2V, enabling distributed deployment of the video generation pipeline across multiple devices or machines. ## What's New ### Core Functionality - Service Decoupling: Separate encoder and transformer services that can run independently - High-Performance Communication: ZeroMQ and RDMA-based messaging with Mooncake transfer engine - Flexible Deployment: Support for single-machine multi-GPU and cross-machine distributed setups ### New Components - `lightx2v/disagg/`: Complete disaggregation package - `conn.py`: Data connection and management - `services/encoder.py`: Encoder service implementation - `services/transformer.py`: Transformer service implementation - `examples/`: Usage examples for WAN I2V and T2V models ## Key Benefits 1. Resource Flexibility: Distribute compute-intensive tasks across multiple devices 2. Scalability: Easy horizontal scaling for production deployments 3. Memory Efficiency: Run large models on hardware-constrained environments 4. Service-Oriented: Build microservice-based video generation systems ## Usage Example ```shell python3 lightx2v/disagg/examples/wan_t2v_service.py ``` See `lightx2v/disagg/examples/` for complete working examples. ## Backward Compatibility ✅ This is an optional feature that doesn't affect existing functionality: - Default mode preserves current behavior - All existing APIs remain unchanged - Users can opt-in to use disaggregation when needed ## Testing - ✅ Tested with WAN I2V and T2V models - ✅ Verified cross-device communication stability - ✅ Validated accuracy matches single-machine mode ## Files Changed - Added: `lightx2v/disagg/` package with all disaggregation modules - Modified: None (purely additive) ## Future Enhancements - Automatic service discovery - Load balancing across multiple workers - Enhanced monitoring and health checks --- Type: Feature Breaking Changes: None Documentation: Included in `lightx2v/disagg/examples/` --------- Co-authored-by: jasonzhang517 <yzhang298@e.ntu.edu.sg> Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com> Co-authored-by: helloyongyang <yongyang1030@163.com>	7 个月前
volcengine_voices_list.json	Update frontend (#436) 修复了一些问题、流程优化以及页面设计 --------- Co-authored-by: XHPlus <xhplus@163.com> Co-authored-by: qinxinyi <qxy118045534@163.com>	11 个月前