KoshurAI_Tarjuma_v3 API & Inference Endpoint

Model Details

Table

Author	Faizan Iqbal (@Faizaniqbal)
Base model	`Faizaniqbal/KoshurAI_Tarjuma_v2`
Adapter type	LoRA (QLoRA training)
Architecture	Gemma3ForCausalLM + PEFT LoRA
Languages	Kashmiri (ks · kas_Arab), English (en)
License	Apache-2.0
Training data	16,637 curated bidirectional EN↔KS sentence pairs
Training compute	Google Colab GPU

Model Tree

text

google/gemma-3-4b-pt    └─ google/gemma-3-4b-it           └─ sarvamai/sarvam-translate                  └─ Faizaniqbal/KoshurAI_Tarjuma_v2   ← 2.8M Kashmiri pretraining                         └─ Faizaniqbal/KoshurAI_Tarjuma_v3           ← this adapter (SFT)

Quickstart

Install

bash

pip install transformers peft accelerate bitsandbytes sentencepiece

Load & Translate

python

import torchfrom transformers import AutoTokenizer, AutoModelForCausalLM, BitsAndBytesConfigfrom peft import PeftModel BASE_MODEL = "Faizaniqbal/KoshurAI_Tarjuma_v2"ADAPTER    = "Faizaniqbal/KoshurAI_Tarjuma_v3" bnb_cfg = BitsAndBytesConfig(    load_in_4bit=True,    bnb_4bit_use_double_quant=True,    bnb_4bit_quant_type="nf4",    bnb_4bit_compute_dtype=torch.bfloat16,) tok = AutoTokenizer.from_pretrained(BASE_MODEL)tok.pad_token = tok.eos_token base  = AutoModelForCausalLM.from_pretrained(    BASE_MODEL, quantization_config=bnb_cfg, device_map="auto")model = PeftModel.from_pretrained(base, ADAPTER)model.eval() def translate(text, direction="en2ks"):    prefix = "Translate to Kashmiri: " if direction == "en2ks" else "Translate to English: "    prompt = f"<start_of_turn>user\n{prefix}{text}<end_of_turn>\n<start_of_turn>model\n"    inputs = tok(prompt, return_tensors="pt", truncation=True, max_length=512).to("cuda")    with torch.no_grad():        out = model.generate(            **inputs,            max_new_tokens=150,            min_new_tokens=5,            do_sample=False,            repetition_penalty=1.1,        )    return tok.decode(out[0][inputs["input_ids"].shape[1]:], skip_special_tokens=True).strip() print(translate("The dog is sleeping.", "en2ks"))print(translate("ہونٛد چھُ شُنٛگِتھ", "ks2en"))

Training

Stage 1 — Kashmiri Pretraining (base model)

The base model (KoshurAI_Tarjuma_v2) was continually pretrained on 2.8 million tokens of Kashmiri text from publicly available sources (literature, journalism, academic texts, religious scholarship). This gave the model deep Kashmiri language knowledge.

Stage 2 — SFT for Translation (this adapter)

This LoRA adapter was trained on 16,637 curated bidirectional sentence pairs (EN↔KS + KS↔EN) to teach the model explicit translation capability.

Table with columns: Split, Records
Split	Records
Base SFT corpus (v2)	15,527
New pairs (v3)	1,110
Total	16,637

Training Configuration

Table with columns: Hyperparameter, Value
Hyperparameter	Value
Base model	`Faizaniqbal/KoshurAI_Tarjuma_v2`
LoRA rank (r)	16
LoRA alpha	16
LoRA dropout	0.05
LoRA target modules	q_proj, k_proj, v_proj, o_proj, gate_proj, up_proj, down_proj
Quantization	4-bit NF4 (BitsAndBytes)
Compute dtype	bfloat16
Epochs	2

Evaluation — FLORES-200 Devtest (1,012 sentences)

Table with columns: Direction, Model, BLEU, COMET
Direction	Model	BLEU	COMET
KS→EN	KoshurAI v3 (ours)	15.74	0.6982 ✅
KS→EN	NLLB-200 distilled-600M	16.28	0.6741
EN→KS	KoshurAI v3 (ours)	30.37¹	0.6604 ✅
EN→KS	NLLB-200 distilled-600M	39.65¹	0.6431

¹ EN→KS BLEU is character-level (tokenize='char'), standard for Arabic-script output. COMET = Unbabel/wmt22-comet-da system score.

KoshurAI v3 outperforms NLLB-200 on COMET in both directions.

Sample Translations (EN→KS)

Table with columns: English, KoshurAI v3
English	KoshurAI v3
They include the Netherlands, with Anna Jochemsen finishing ninth.	تِیَم چھُ نیدرلینڈس شامِل کَران اَینا جوکیمسن فِنِشِنگ نائنتھ سیتھ
Hershey and Chase used phages, or viruses, to implant their own DNA.	ۂرشے تہٕ چیسن کٔرۍ فیگ تہٕ جَراثیم منٛز پنُن ڈی این اے اَزناوُنہِ خٲطر
They usually have special food, drink and entertainment offers.	تِیَمَن چھُ اکثر خاص کھٮ۪ن، چیٖز تہٕ تفریح پیش کَرنہِ یِوان

Inference Settings

Table with columns: Parameter, Value
Parameter	Value
`do_sample`	`False` (greedy)
`max_new_tokens`	150 (EN→KS) / 200 (KS→EN)
`min_new_tokens`	5
`repetition_penalty`	1.1

Hardware Requirements

Table with columns: Setting, VRAM
Setting	VRAM
4-bit inference (recommended)	~6–8 GB
Colab free tier (T4)	✅ with 4-bit
Colab L4 / A100	✅ comfortable

Limitations

Trained on sentence-level pairs (≤ 512 tokens); long-form translation unsupported.
Performance on technical, legal, or dialectal Kashmiri is unverified.
No human evaluation conducted; COMET and BLEU are automatic metrics only.
4-bit quantization used for inference; full-precision may yield higher scores.

Citation

If you use this model, please cite:

bibtex

@misc{iqbal2026koshurai,  title        = {KoshurAI v3: A Fine-Tuned Neural Machine Translation System                  for Kashmiri--English Bidirectional Translation},  author       = {Iqbal, Faizan},  year         = {2026},  howpublished = {\url{https://huggingface.co/Faizaniqbal/KoshurAI_Tarjuma_v3}},  note         = {LoRA adapter fine-tuned from Faizaniqbal/KoshurAI_Tarjuma_v2}}

This work fine-tunes the model by Malik & Nissar — also cite:

bibtex

@misc{malik2026koshurkouter,  title        = {Koshur Kouter KS-EN v1: A Merged QLoRA Kashmiri--English Translation Model},  author       = {Malik, Haq Nawaz and Nissar, Nahfid},  year         = {2026},  howpublished = {\url{https://huggingface.co/Omarrran/koshur-kouter-ks-en_v1}},  note         = {Fine-tuned from sarvamai/sarvam-translate}}

And the original base model:

bibtex

@misc{sarvam2025translate,  title        = {Sarvam-Translate},  author       = {{Sarvam AI}},  howpublished = {\url{https://huggingface.co/sarvamai/sarvam-translate}}}

Acknowledgements

This model builds on Omarrran/koshur-kouter-ks-en_v1, which was fine-tuned by Haq Nawaz Malik & Nahfid Nissar (2026), itself built on sarvamai/sarvam-translate (Gemma 3, 4.5B) by Sarvam AI. Evaluated on FLORES-200 devtest. COMET scored using Unbabel/wmt22-comet-da.

Author

Faizan Iqbal (@Faizaniqbal)

Base model

Faizaniqbal/KoshurAI_Tarjuma_v2

Adapter type

LoRA (QLoRA training)

Architecture

Gemma3ForCausalLM + PEFT LoRA

Languages

Kashmiri (ks · kas_Arab), English (en)

License

Apache-2.0

Training data

16,637 curated bidirectional EN↔KS sentence pairs

Training compute

Google Colab GPU

google/gemma-3-4b-pt └─ google/gemma-3-4b-it └─ sarvamai/sarvam-translate └─ Faizaniqbal/KoshurAI_Tarjuma_v2 ← 2.8M Kashmiri pretraining └─ Faizaniqbal/KoshurAI_Tarjuma_v3 ← this adapter (SFT)

import torchfrom transformers import AutoTokenizer, AutoModelForCausalLM, BitsAndBytesConfigfrom peft import PeftModel BASE_MODEL = "Faizaniqbal/KoshurAI_Tarjuma_v2"ADAPTER = "Faizaniqbal/KoshurAI_Tarjuma_v3" bnb_cfg = BitsAndBytesConfig( load_in_4bit=True, bnb_4bit_use_double_quant=True, bnb_4bit_quant_type="nf4", bnb_4bit_compute_dtype=torch.bfloat16,) tok = AutoTokenizer.from_pretrained(BASE_MODEL)tok.pad_token = tok.eos_token base = AutoModelForCausalLM.from_pretrained( BASE_MODEL, quantization_config=bnb_cfg, device_map="auto")model = PeftModel.from_pretrained(base, ADAPTER)model.eval() def translate(text, direction="en2ks"): prefix = "Translate to Kashmiri: " if direction == "en2ks" else "Translate to English: " prompt = f"<start_of_turn>user\n{prefix}{text}<end_of_turn>\n<start_of_turn>model\n" inputs = tok(prompt, return_tensors="pt", truncation=True, max_length=512).to("cuda") with torch.no_grad(): out = model.generate( **inputs, max_new_tokens=150, min_new_tokens=5, do_sample=False, repetition_penalty=1.1, ) return tok.decode(out[0][inputs["input_ids"].shape[1]:], skip_special_tokens=True).strip() print(translate("The dog is sleeping.", "en2ks"))print(translate("ہونٛد چھُ شُنٛگِتھ", "ks2en"))

Split

Records

Base SFT corpus (v2)

15,527

New pairs (v3)

1,110

Total

16,637

Hyperparameter

Value

Base model

Faizaniqbal/KoshurAI_Tarjuma_v2

LoRA rank (r)

LoRA alpha

LoRA dropout

0.05

LoRA target modules

q_proj, k_proj, v_proj, o_proj, gate_proj, up_proj, down_proj

Quantization

4-bit NF4 (BitsAndBytes)

Compute dtype

bfloat16

Epochs

Direction

Model

BLEU

COMET

KS→EN

KoshurAI v3 (ours)

15.74

0.6982 ✅

KS→EN

NLLB-200 distilled-600M

16.28

0.6741

EN→KS

KoshurAI v3 (ours)

30.37¹

0.6604 ✅

EN→KS

NLLB-200 distilled-600M

39.65¹

0.6431

English

KoshurAI v3

They include the Netherlands, with Anna Jochemsen finishing ninth.

تِیَم چھُ نیدرلینڈس شامِل کَران اَینا جوکیمسن فِنِشِنگ نائنتھ سیتھ

Hershey and Chase used phages, or viruses, to implant their own DNA.

ۂرشے تہٕ چیسن کٔرۍ فیگ تہٕ جَراثیم منٛز پنُن ڈی این اے اَزناوُنہِ خٲطر

They usually have special food, drink and entertainment offers.

تِیَمَن چھُ اکثر خاص کھٮ۪ن، چیٖز تہٕ تفریح پیش کَرنہِ یِوان

Parameter

Value

do_sample

False (greedy)

max_new_tokens

150 (EN→KS) / 200 (KS→EN)

min_new_tokens

repetition_penalty

1.1

Setting

VRAM

4-bit inference (recommended)

~6–8 GB

Colab free tier (T4)

✅ with 4-bit

Colab L4 / A100

✅ comfortable

@misc{iqbal2026koshurai, title = {KoshurAI v3: A Fine-Tuned Neural Machine Translation System for Kashmiri--English Bidirectional Translation}, author = {Iqbal, Faizan}, year = {2026}, howpublished = {\url{https://huggingface.co/Faizaniqbal/KoshurAI_Tarjuma_v3}}, note = {LoRA adapter fine-tuned from Faizaniqbal/KoshurAI_Tarjuma_v2}}

@misc{malik2026koshurkouter, title = {Koshur Kouter KS-EN v1: A Merged QLoRA Kashmiri--English Translation Model}, author = {Malik, Haq Nawaz and Nissar, Nahfid}, year = {2026}, howpublished = {\url{https://huggingface.co/Omarrran/koshur-kouter-ks-en_v1}}, note = {Fine-tuned from sarvamai/sarvam-translate}}

KoshurAI_Tarjuma_v3

README

Model Details

Model Tree

Quickstart

Install

Load & Translate

Training

Stage 1 — Kashmiri Pretraining (base model)

Stage 2 — SFT for Translation (this adapter)

Training Configuration

Evaluation — FLORES-200 Devtest (1,012 sentences)

Sample Translations (EN→KS)

Inference Settings

Hardware Requirements

Limitations

Citation

Acknowledgements

Explore FriendliAI today

README

Model Details

Model Tree

Quickstart

Install

Load & Translate

Training

Stage 1 — Kashmiri Pretraining (base model)

Stage 2 — SFT for Translation (this adapter)

Training Configuration

Evaluation — FLORES-200 Devtest (1,012 sentences)

Sample Translations (EN→KS)

Inference Settings

Hardware Requirements

Limitations

Citation

Acknowledgements