Spaces:
Runtime error
Runtime error
Commit History
-
fix compatability issues with python 3.9
af41e07
-
Bugfix: pass device type for autocast.
daniel shalemd675fab
-
vae: fix attention blocks and timestep conditioning
bd9f5fd
-
VAE: Check for timesteps parameter in decoder before calling
8920e2d -
VAE: Add timestep conditioning
65dad79
-
Merge pull request #34 from LightricksResearch/add_atten_to_decoder
Unverified signatureDecoder: add AttentionResBlocks block.f63ea56 -
Decoder: add AttentionResBlocks block
1. Support attention block after residual block. 2. Add flash attention support.07ddecf -
Merge pull request #33 from LightricksResearch/compress-all-half-channels
Guy ShiranUnverified signatureVAE: support channel reduction in depth to space blocks427926d
-
causal_video_autoencoder: add option to half channels in depth to space upsample block
d5e984f
-
Merge pull request #29 from LightricksResearch/scale_vae_decoder_v1
Unverified signatureVAE Decoder: Inject noise between conv layers.c4b2a35 -
VAE Decoder: Inject noise between conv layers.
1.Add inject_noise flag to res_x, rex_x_y blocks. 2.Init noise to zero in ResnetBlock3D constructor. 2.Add _feed_spatial_noise method to inject noise between conv layers.1f21780 -
Merge pull request #26 from LightricksResearch/cuda_optional
Sapir WeissbuchUnverified signatureCPU runtime & multiple video save on inference.py7f1b6d2
-
ran the ruff formatter
c042515 -
Merge pull request #30 from LightricksResearch/fix-no-flash-attention
Sapir WeissbuchUnverified signaturemodel: fix flash attention enabling - do not check device type at this point05cb3e4 -
model: fix flash attention enabling - do not check device type at this point (can be CPU)
5940103
-
Added NIE on xla device
f5895e7 -
added cuda as optional
85a3cf8
-
README: added inference + installation guidelines, inference clearer.
b6c994f
-
Feature: Add full bfloat16 support.
daniel shalem6a9d9a1
-
Merge pull request #19 from LightricksResearch/bfloat16-inference
Sapir WeissbuchUnverified signatureFeature: Add mixed precision support and direct bfloat16 support.91602f9 -
Feature: Add mixed precision support and direct bfloat16 support.
daniel shalem1940326 -
VAE: Support different latent_var_log options when returning intermediate features for 3D perceptual loss
7d89bb0
-
VAE: Support retuning intermediate features for 3d perceptual loss
PR fix lint028b6a1
-
Remove the word "pixart" from code.
d504563
-
VAE: Support more configurations for Encoder and Decoder blocks
VAE: Define encoder compress-all block with channel multiplier VAE: Support residual connection in the decoder VAE: Refactor CausalConv3d parameters lint43d3c68
-
Lint: added ruff.
325137b
-
Image to video script: make determinist by random seed.
4bb89c5
-
Image to video: updated script.
4535a03 -
VAE encode bugfix: remove bad dist_util.
eec4cb2
-
Examples: update and fix scripts.
e46ff5e -
CausalVideoAutoencoder: made neater load_ckpt.
d699d2b
-
Ckpt conversion: script + usage examples updated.
bebbcd0
-
Merge pull request #7 from LightricksResearch/feature/fix-transformer-init-bug
Dudu MosheUnverified signaturetransformer3d: init mode xora never happens because lower case needed.fc02e02 -
transformer3d: init mode xora never happens because lower case needed.
a3498bb
-
import fixes.
b30014f
-
Added tpu flash attention.
4f52f00 -
working.
86b1a7e
-
refactor
c811a04