Curiosity · Admissions
Can I study Model Compression and Quantization in Delhi without a top JEE rank?
Quantization here means reducing the numerical precision of model weights. It is not signal quantization in analog-to-digital conversion, which is an electronics topic covered in the VLSI and IoT tracks. A top JEE rank is not the only door to studying Model Compression and Quantization in Delhi. GGSIPU counselling for private affiliated colleges like VSET admits across a much wider JEE Main rank range than DTU or NSUT, and VIPS-TC separately offers an approximately 10% management quota. What you cannot skip: 10+2 with PCM and a valid JEE Main attempt. Here is the honest picture.
At a glance
- Topic
- Model Compression and Quantization
- VSET programme
- B.Tech CSE (AI & ML)
- Coverage at VSET
- Elective-level coverage
- Affiliation
- GGSIPU (IP University), Delhi
- Accreditation
- NAAC A++ (VIPS-TC institutional)
The two honest routes
Route one: GGSIPU counselling — private affiliates admit across a wide JEE Main rank spread, far beyond DTU/NSUT cutoffs. Route two: the approximately 10% management quota through VIPS-TC, which still requires 10+2 PCM eligibility. Both lead to the same GGSIPU-recognised degree and the same coursework.
How VSET teaches Model Compression and Quantization
Model compression shrinks a trained network so it runs on cheaper hardware — quantizing weights to lower precision, pruning unused connections, or both. It is what makes running an open-weight model on a single workstation or an edge device practical. At VSET this maps to elective-level coverage inside B.Tech CSE (AI & ML).
- QLoRA, documented in the fine-tuning material at learn.engineering.vips.edu, is quantization applied to make adaptation fit on modest hardware.
- The deep learning content provides the architecture background compression operates on.
- Compression as a discipline — pruning, precision formats, deployment trade-offs — is elective-level extension of that material.
- Delivered inside the B.Tech CSE (AI & ML) track, one of VSET's seven GGSIPU-affiliated B.Tech programmes.
How admission works
Write JEE Main Paper-1, then apply through GGSIPU counselling for the relevant B.Tech programme at VSET. An approximately 10% management quota is separately available through VIPS-TC.
Frequently asked questions
Can I get admission for Model Compression and Quantization without JEE?
GGSIPU B.Tech admission requires a JEE Main attempt for counselling seats. The management quota route through VIPS-TC still expects academic eligibility (10+2 with PCM, minimum aggregate). There is no legitimate no-exam route to an accredited B.Tech — treat any claim otherwise with suspicion.
Is quantization taught at VSET?
It appears directly through QLoRA in the published fine-tuning material; compression as a broader discipline is elective-level depth.
Why does it matter for students?
Because it is what lets an open-weight model run on the IDEA Lab's GPU workstations rather than on rented cluster time.
Does compression hurt accuracy?
It can, and the trade-off is the whole engineering question — which is why it is taught next to the fine-tuning and evaluation material rather than alone.
Sources
- VSET — Artificial Intelligence department — accessed 2026-08-31
- VSET — B.Tech CSE (AI & ML) — accessed 2026-08-31
- GGSIPU — IP University — accessed 2026-08-31