You can deploy this model and start earning money today!

pankajpandey-dev/
qwen3.5-9b-hindi-instruct

Name: qwen3.5-9b-hindi-instruct
Rating: 2.5 (38 reviews)

Views

No views yet

Qwen3.5-9B Hindi Instruct 🇮🇳

Text-only fine-tune of a multimodal base — use for text generation as shown below (ignore the auto-suggested image examples).

Qwen3.5-9B fine-tuned to answer directly in fluent, native Hindi — no English code-switching, no English "thinking" detours. Runs on a laptop CPU via the GGUF version (~5.7 GB).

Why this exists

Base Qwen3.5-9B knows Hindi, but ask it a Hindi question and it burns hundreds of tokens thinking in English inside its think block before answering — and its answers code-switch. This fine-tune closes the think block immediately and replies in pure Devanagari, following instructions tightly.

Examples (from the release eval)

प्रश्न: अपनी माँ के लिए धन्यवाद व्यक्त करते हुए चार पंक्तियाँ लिखिए।

उत्तर: माँ, शब्दों में क्या कहूँ, आपने मेरे लिए सब कुछ किया है। आप मेरी बुद्धि, मेरी ताकत, मेरा आशीर्वाद और मेरी प्रेरणा हैं। आपका प्यार मुझे हर दिन बढ़ने और आगे बढ़ने में मदद करता है। आपके बिना मैं कुछ नहीं हूँ।

प्रश्न: चाय बनाने की विधि बताइए।

उत्तर: पानी के बर्तन में पानी उबालें और चाय के पत्ते डाल दें। 5 से 7 मिनट तक उबलने दें, फिर चाय को छान लें। एक कप या चम्मच चीनी या मिश्रण जोड़ें और चाय का आनंद लें।

प्रश्न: भारत की राजधानी क्या है? एक वाक्य में उत्तर दीजिए।

उत्तर: भारत की राजधानी नई दिल्ली है।

Quickstart

python

from transformers import AutoModelForCausalLM, AutoTokenizer
m = AutoModelForCausalLM.from_pretrained("pankajpandey-dev/qwen3.5-9b-hindi-instruct", torch_dtype="bfloat16", device_map="auto")
t = AutoTokenizer.from_pretrained("pankajpandey-dev/qwen3.5-9b-hindi-instruct")
msgs = [{"role": "user", "content": "जल संरक्षण के पाँच तरीके बताइए।"}]
text = t.apply_chat_template(msgs, add_generation_prompt=True, tokenize=False)
out = m.generate(**t(text=text, return_tensors="pt").to(m.device),
                 max_new_tokens=400, repetition_penalty=1.1)
print(t.decode(out[0], skip_special_tokens=True))

Use repetition_penalty=1.1 (set in this repo's generation_config) — long letter/essay outputs can loop without it. The chat template opens a think block; this model closes it immediately — strip <think>...</think> from output if present.

Training recipe (fully reproducible)


Base	unsloth/Qwen3.5-9B (bf16)
Method	LoRA r=16, alpha=16, response-only loss (Unsloth)
Data	12,912 Hindi pairs — anudesh 5,000 · dolly-hi 4,000 (chrF++ >= 55) · wikiHow-hi 3,000 · Aya-hi 912
Schedule	2 epochs, LR 1e-4 cosine, effective batch 16, seq 2048
Hardware	1x NVIDIA L40S (48 GB), ~135 min
Final train loss	0.938

Data deduplicated across sources, filtered for length and Latin-script ratio. wikiHow adds long-form step-by-step answers; Aya adds human-written originals.

Limitations

Parts of the data are machine-translated (dolly) or model-generated (anudesh via Llama-2-70B), so occasional unnatural phrasing or factual errors occur. Long formal-letter outputs may repeat without repetition_penalty=1.1. No additional safety tuning. Knowledge cutoff follows base Qwen3.5.

Licensing

Weights: Apache 2.0. Data licenses apply to data, not weights: dolly-hi CC-BY-SA 3.0-derived; anudesh generated by Llama-2-70B (Llama 2 license); Aya Apache 2.0.

🇮🇳 About the Hindi LLM Series

Weekly open releases making small LLMs speak fluent, native Hindi — trained on free/low-cost GPUs, shipped as GGUF for laptops and edge devices. Built by pankajpandey-dev (contact links on profile).

This release: Model · GGUF · LoRA · Series: 🇮🇳 Hindi LLM Collection