Does CogVideoX1.5-5B still not support http-server?

[Chenxwi]

I had no problem running CogVideoX1.5-5B using the example script, So I'm going to run it using http-server, but the error message indicates that it is not supported yet.

When I check the file http-server/host.py, There is no CogVideoX1.5-5B in the pipeline_map.

But I found that there was pipeline_cogvideox.py in xfuser/piplines, Then I manually insert the pipeline_cogvideox.py into pipeline_map in http-service/host.py, like this:

It still didn't work, error message as follows:

TypeError: CogVideoXPipeline.__call__() got an unexpected keyword argument 'use_resolution_binning'

envs info:

CUDA:12.6
GPU: A40
python: 3.11.10
torch: 2.4.0

The config.json

{
    "nproc_per_node": 1,
    "model": "/mnt/pdd/xDiT/models/CogVideoX1.5-5B",
    "height": 768,
    "width": 1360,
    "num_frames": 17,
    "ulysses_degree": 1,
    "ring_degree": 1,
    "pipefusion_parallel_degree": 1,
    "enable_tiling": false
}

Am I running CogVideo inference wrong with http-server? Or is the cogvideo pipeline not finished?

[fielding]

quick workaround or now i you need it @Chenxwi #501