github/ai-toolkit
1
0
Fork 0
Code Issues Pull Requests Actions 1 Packages Projects Releases Wiki Activity

Add support for audo frame count so datasets can have varrying length videos. Varous ltx 2.3 VAE optimizations such as removing tiling articacts, and doing frame split encoding to reduce vram on encoding/decoding.

Browse Source
This commit is contained in:
Jaret Burkett
2026-03-24 12:20:09 -06:00
parent 4ad14d211a
commit 7f3309b291
10 changed files with 132 additions and 36 deletions
Download Patch File Download Diff File Expand all files Collapse all files

10
ui/src/app/jobs/new/SimpleJob.tsx
View File

@@ -846,7 +846,7 @@ export default function SimpleJob({
min={0}
required
/>
{modelArch?.additionalSections?.includes('datasets.num_frames') && (
{modelArch?.additionalSections?.includes('datasets.num_frames') && !dataset.auto_frame_count && (
<NumberInput
label="Num Frames"
className="pt-2"
@@ -873,6 +873,14 @@ export default function SimpleJob({
checked={dataset.is_reg || false}
onChange={value => setJobConfig(value, `config.process[0].datasets[${i}].is_reg`)}
/>
{modelArch?.additionalSections?.includes('datasets.auto_frame_count') && (
<Checkbox
label="Auto Frame Count"
checked={dataset.auto_frame_count || false}
onChange={value => setJobConfig(value, `config.process[0].datasets[${i}].auto_frame_count`)}
docKey="datasets.auto_frame_count"
/>
)}
{modelArch?.additionalSections?.includes('datasets.do_i2v') && (
<Checkbox
label="Do I2V"

7
ui/src/app/jobs/new/options.ts
View File

@@ -20,6 +20,7 @@ type AdditionalSections =
| 'datasets.do_audio'
| 'datasets.audio_normalize'
| 'datasets.audio_preserve_pitch'
| 'datasets.auto_frame_count'
| 'sample.ctrl_img'
| 'sample.multi_ctrl_imgs'
| 'train.audio_loss_multiplier'
@@ -664,9 +665,10 @@ export const modelArchs: ModelArch[] = [
'config.process[0].datasets[x].do_i2v': [false, undefined],
'config.process[0].datasets[x].do_audio': [true, undefined],
'config.process[0].datasets[x].fps': [24, undefined],
'config.process[0].datasets[x].auto_frame_count': [false, undefined],
},
disableSections: ['network.conv'],
additionalSections: ['sample.ctrl_img', 'datasets.num_frames', 'model.layer_offloading', 'model.low_vram', 'datasets.do_audio', 'datasets.audio_normalize', 'datasets.audio_preserve_pitch', 'datasets.do_i2v', 'train.audio_loss_multiplier'],
additionalSections: ['sample.ctrl_img', 'datasets.num_frames', 'model.layer_offloading', 'model.low_vram', 'datasets.do_audio', 'datasets.audio_normalize', 'datasets.audio_preserve_pitch', 'datasets.do_i2v', 'train.audio_loss_multiplier', 'datasets.auto_frame_count'],
},
{
name: 'ltx2.3',
@@ -691,9 +693,10 @@ export const modelArchs: ModelArch[] = [
'config.process[0].datasets[x].do_i2v': [false, undefined],
'config.process[0].datasets[x].do_audio': [true, undefined],
'config.process[0].datasets[x].fps': [24, undefined],
'config.process[0].datasets[x].auto_frame_count': [false, undefined],
},
disableSections: ['network.conv'],
additionalSections: ['sample.ctrl_img', 'datasets.num_frames', 'model.layer_offloading', 'model.low_vram', 'datasets.do_audio', 'datasets.audio_normalize', 'datasets.audio_preserve_pitch', 'datasets.do_i2v', 'train.audio_loss_multiplier'],
additionalSections: ['sample.ctrl_img', 'datasets.num_frames', 'model.layer_offloading', 'model.low_vram', 'datasets.do_audio', 'datasets.audio_normalize', 'datasets.audio_preserve_pitch', 'datasets.do_i2v', 'train.audio_loss_multiplier', 'datasets.auto_frame_count'],
},
{
name: 'flux2_klein_4b',