🔄 In a Training Loop

RACER IS OP

saidutta69

5 4 24

https://sdad.pro

instax-dutta

AI & ML interests

On-device ML · Abliteration · Distillation · Odia NLP · Security models · Running everything locally on TFLOPS and vibes

Recent Activity

updated a dataset 1 day ago

saidutta69/PhishTrap

updated a collection 1 day ago

RACER IS OP — Heretic Models

updated a model 2 days ago

saidutta69/granite-4.2-3b-heretic

View all activity

Organizations

New activity in Qwen/Qwen2.5-1.5B-Instruct 8 days ago

Qwen2.5-1.5B-Instruct under parallel constrained decoding: 0.605 over 1,240 typed decisions, with the cost that implies on CPU

#29 opened 8 days ago by

saidutta69

New activity in convaiinnovations/laya 8 days ago

Independent head-to-head on a sealed 1,190-case manifest: laya 0.686 vs Jev 1.13.0 0.907 vs Qwen2.5-1.5B PCD 0.605, with paired intervals, plus three adapter findings

🔥 2

#20 opened 8 days ago by

saidutta69

New activity in convaiinnovations/laya 12 days ago

Independent Laya vs Jev head-to-head on identical inputs

👍🔥 3

#6 opened 13 days ago by

saidutta69

New activity in harshatheg/Qwen-2.5-1B-RLCD 12 days ago

Independent 4-way benchmark incl. this engine (751 identical states)

#2 opened 12 days ago by

saidutta69

New activity in huggingface/InferenceSupport 3 months ago

saidutta69/Qwen2.5-Coder-3B-Instruct-heretic

🤗 1

#10810 opened 3 months ago by

saidutta69

RACER IS OP

AI & ML interests

Recent Activity

Organizations

saidutta69's activity

Qwen2.5-1.5B-Instruct under parallel constrained decoding: 0.605 over 1,240 typed decisions, with the cost that implies on CPU

Independent head-to-head on a sealed 1,190-case manifest: laya 0.686 vs Jev 1.13.0 0.907 vs Qwen2.5-1.5B PCD 0.605, with paired intervals, plus three adapter findings

Independent Laya vs Jev head-to-head on identical inputs

Independent 4-way benchmark incl. this engine (751 identical states)

saidutta69/Qwen2.5-Coder-3B-Instruct-heretic