Beta
Explore
Marketplace
Neural Labs
Chat
Wallet
Docs
gemma-2-2b-legal – AI Model by sudhisrk1982 | AlphaNeural AI
You can deploy this model and start earning money today!
sudhisrk1982
/
gemma-2-2b-legal
like
0
safetensors
gemma2
text-generation
conversational
google/gemma-2-2b-it
finetune
gemma
us
Views
No views yet
Model card
Files and Versions
Community
API
Deploy
gemma-2-2b-legal
QLoRA fine-tune of
google/gemma-2-2b-it
on 11,891 legal/financial Q&A + summarization pairs generated from a case-law + SEC corpus.
Training
Method
: QLoRA (4-bit NF4 base, LoRA r=16 alpha=32, ~20.8M trainable = 1.28%)
Data
: 11,891 pairs (9,071 grounded QA + 2,820 summarization), answer-only loss
1 epoch
, lr 2e-4 cosine, bf16 adapter, 1xA100
Best val loss
: 1.577 (perplexity 4.84)
Adapter merged into bf16 weights; serve as an ordinary model.
Limitations
2B parameters, ~12k fine-tuning pairs, single domain (US law + SEC filings). Verify anything load-bearing. Not legal advice.
