Repository Issue Activity (beta)

#123 MAGI-1.1 24B base checkpoint appears incompatible with public inference code

#122 DreamX-World: A General-Purpose Interactive World Model

#89 Request for fine-tuning / LoRA code

#119 TypeError: flex_flash_attn_func() got an unexpected keyword argument 'max_seqlen_q'

#120 Does Magi-1 support batch inference?

#118 pip install --no-build-isolation . error

#117 Request for MAGI 4.5B Model's I2V Physical IQ Test Results

#116 Cannot see services.

#113 Audio-Driven / Lip-Synced Video Generation Support (as available on magi.sand.ai)

#112 luva

#111 luva

#110 LUVA HELT

#109 Does Magi-1 24B require H100 x 8 only for inference?

#106 About Multiframe V2V Generation

#108 24B_distill_quant: ZeroDivisionError: integer division or modulo by zero

#107 8 x 4090 run MAGI-1-24B-distill+fp8_quant: CUDA error: too many resources requested for launch

#104 Beija-flor

#103 Why sample-then-blend in `tiled_encode`?

#55 What is the difference between 24B_base and 24B_distill?

#102 KV_Range length

#96 ComfyUI import error

#101 Is the lipsync feature from the website supported by the open release?

#100 1080p possible on single H100 (80gb VRAM)?

#99 Reconstruction using MAGI-1-VAE

#98 need help - ComfyUI with MAGI-1