Spaces:

haiyangliu0
/

FloodDiffusion-Streaming

Running on T4

App Files Files Community

FloodDiffusion-Streaming

Commit History

Oct 25 Yesterday

06:34
Update account and repository links to haiyangliu0 Running

haiyangliu0

Verified signature
c68e928

Apr 9

08:46
Update README.md

Haiyang Liu

Verified signature
8a89963

Mar 16

12:28
feat: add Bo Zheng homepage link

haiyangliu0

c8826ac
12:27
feat: add author homepage links for Yiyi Cai, Yuhan Wu, Kunhang Li

haiyangliu0

7fe2985
12:22
fix: correct author order per paper

haiyangliu0

a30e265
12:20
feat: add FloodDiffusion title, CVPR 2026, authors, paper/code badges

haiyangliu0

b2734f9
12:16
feat: spectators see 'Take Over' button to seamlessly take control

haiyangliu0

4da9546
12:11
fix: syntax error in get_frame else branch indentation

haiyangliu0

5f2882e
12:04
feat: spectator mode - all visitors see the same streaming generation

haiyangliu0

b1e1a05
11:50
fix: move k_lens to GPU in SDPA fallback (tested locally)

haiyangliu0

7d73321
11:44
fix: build proper attention mask in SDPA fallback for text cross-attention

haiyangliu0

6c6483b
11:37
perf: increase server frame buffer target to 16 for batch fetching

haiyangliu0

7237651
11:34
revert: remove float16 override, bfloat16 is safer (float16 overflow risk)

haiyangliu0

2cefd84
11:25
perf: batch frame fetching to overcome HTTP round-trip latency

haiyangliu0

a3de023
11:20
perf: use float16 instead of bfloat16 in SDPA for T4 tensor core acceleration

haiyangliu0

80c3e53
11:14
fix: use model_manager.get_buffer_status() for config endpoint instead of model attrs

haiyangliu0

3ab7701
10:47
fix: adapt model_manager to HF model API (no schedule_config/cfg_config dicts)

haiyangliu0

47551f4
10:39
fix: cast SDPA output back to original dtype + idempotent patching

haiyangliu0

e843211
10:34
fix: patch flash_attention with SDPA fallback for T4 (no flash-attn)

haiyangliu0

bb7e158
10:22
fix: correct CUDA base image tag to 12.4.1

haiyangliu0

3042363
10:19
fix: upgrade to CUDA 12.4 + torch>=2.6 for CVE-2025-32434 compat

haiyangliu0

c5fa989
10:14
fix: pin transformers<5 for google/umt5-base compat (no model_type in config)

haiyangliu0

975c258
10:08
fix: remove pre-download step, model downloads at runtime startup

haiyangliu0

da265c6
10:04
fix: add diffusers, ftfy, regex deps from model dependency tree

haiyangliu0

e153101
10:00
fix: add omegaconf and safetensors deps required by model

haiyangliu0

21a7038
09:57
fix: add lightning dependency required by model

haiyangliu0

ff0b1d9
09:53
fix: split pip install to avoid index-url conflict for non-torch packages

haiyangliu0

c62140b
09:45
feat: initial FloodDiffusion streaming demo for HF Space⁠

haiyangliu0 +1: Claude Opus 4.6 (1M context)

a4f8eb3
09:45
initial commit

Haiyang Liu

Verified signature
384dead

Commit History

Oct 25 Yesterday

Update account and repository links to haiyangliu0 Running

Apr 9

Update README.md

Mar 16

feat: add Bo Zheng homepage link

feat: add author homepage links for Yiyi Cai, Yuhan Wu, Kunhang Li

fix: correct author order per paper

feat: add FloodDiffusion title, CVPR 2026, authors, paper/code badges

feat: spectators see 'Take Over' button to seamlessly take control

fix: syntax error in get_frame else branch indentation

feat: spectator mode - all visitors see the same streaming generation

fix: move k_lens to GPU in SDPA fallback (tested locally)

fix: build proper attention mask in SDPA fallback for text cross-attention

perf: increase server frame buffer target to 16 for batch fetching

revert: remove float16 override, bfloat16 is safer (float16 overflow risk)

perf: batch frame fetching to overcome HTTP round-trip latency

perf: use float16 instead of bfloat16 in SDPA for T4 tensor core acceleration

fix: use model_manager.get_buffer_status() for config endpoint instead of model attrs

fix: adapt model_manager to HF model API (no schedule_config/cfg_config dicts)

fix: cast SDPA output back to original dtype + idempotent patching

fix: patch flash_attention with SDPA fallback for T4 (no flash-attn)

fix: correct CUDA base image tag to 12.4.1

fix: upgrade to CUDA 12.4 + torch>=2.6 for CVE-2025-32434 compat

fix: pin transformers<5 for google/umt5-base compat (no model_type in config)

fix: remove pre-download step, model downloads at runtime startup

fix: add diffusers, ftfy, regex deps from model dependency tree

fix: add omegaconf and safetensors deps required by model

fix: add lightning dependency required by model

fix: split pip install to avoid index-url conflict for non-torch packages

feat: initial FloodDiffusion streaming demo for HF Space⁠

initial commit