Downloads · 30 days
44
40% of all-time downloads
liangyuch/ttcc-sft-qwen25omni-3b-v8
ttcc-sft-qwen25omni-3b-v8 is a machine learning model from liangyuch. Use it for the machine learning task on the model card, and read the license before you ship it in a product. The card lists the license as other.
V8 supervised fine-tune of Qwen/Qwen2.5-Omni-3B with a custom hazard retention head that forecasts second-by-second short-video audience retention curves R(t).
Downloads · 30 days
44
40% of all-time downloads
All-time downloads
111
Public
Parameters
1.3M
9.4 GB on disk
Likes
0
Public
Click a slice to open those files.
.safetensors9.4 GB · 100%
From the Hugging Face model README
V8 supervised fine-tune of Qwen/Qwen2.5-Omni-3B with a custom hazard retention head that forecasts second-by-second short-video audience retention curves R(t).
<cot>...</cot> reasoning (no R decimals); hazard head reads h[</cot>].
Fixes V7's label-in-input leak (see go_viral INCIDENT_2026-05-26).config.json has model_type: qwen2_5_omni_retention, defined by register.py in the ttcc-rl
branch of cliangyu/go_viral. To load:
git clone -b ttcc-rl https://github.com/cliangyu/go_viral--external_plugins examples/custom/qwen2_5_omni_retention/register.py --model_type qwen2_5_omni_retention, env RETENTION_HEAD_TYPE=hazard.Optimizer/RNG/scheduler states are NOT included (inference/init only; not resumable).