Hugging Face's logo
Pankaj Pandey's picture
šŸ—ļø Building on HF

Pankaj Pandey

pankajpandey-dev
3 3 16
DARKSAIDX's profile picture Tralalabs's profile picture keke-420's profile picture
Ā·

AI & ML interests

Natural Language Processing, Text Generation, Large Language Models, Quantization, Fine-Tuning, RLHF, Model Merging.

Recent Activity

new activity about 17 hours ago
huggingface/InferenceSupport:pankajpandey-dev/gemma-4-e4b-hindi-instruct-GGUF
repliedto their post 1 day ago
šŸ‡®šŸ‡³ New in my Hindi LLM Series: Gemma-4 E4B, fine-tuned for Hindi ā€” and it runs on your laptop's CPU. I fine-tuned Google's new Gemma-4 E4B on ~10k Hindi instruction pairs (AI4Bharat: anudesh + dolly) using Unsloth + LoRA, on a single L4 GPU. Then I ran an honest side-by-side eval: base Gemma-4 vs my fine-tune, across 25 Hindi prompts. The results were interesting šŸ‘‡ āœ… My fine-tune is more concise ā€” ask for "3 tips" and it gives exactly 3. Base writes a 1,200-character essay. āœ… Pure native Hindi ā€” base keeps slipping into English ("ą¤øą¤‚ą¤¤ą„ą¤²ą¤æą¤¤ ą¤†ą¤¹ą¤¾ą¤° (Eat a Balanced Diet)", "ą¤¤ą¤¾ą¤°ą¤¾ (Star)"). My fine-tune stays in clean Hindi. āœ… Tighter instruction-following ā€” ask for a "short message" and it gives one, not a menu of options. āš–ļø And to be honest: base Gemma-4 is more detailed and comprehensive. I didn't build a "smarter" model ā€” I built a focused, Hindi-native, edge-friendly one that runs as a 5GB GGUF (Q4) on CPU. šŸ”— Try it: Live demo (CPU): https://huggingface.co/spaces/pankajpandey-dev/gemma-4-e4b-hindi-demo GGUF (Ollama/llama.cpp): https://huggingface.co/pankajpandey-dev/gemma-4-e4b-hindi-instruct-GGUF 16-bit model: https://huggingface.co/pankajpandey-dev/gemma-4-e4b-hindi-instruct Built with @unsloth Ā· Data by @ai4bharat šŸ™ #Hindi #LLM #Gemma #Unsloth #IndicNLP #GGUF
reacted to theirpost with ā¤ļø 2 days ago
šŸ‡®šŸ‡³ New in my Hindi LLM Series: Gemma-4 E4B, fine-tuned for Hindi ā€” and it runs on your laptop's CPU. I fine-tuned Google's new Gemma-4 E4B on ~10k Hindi instruction pairs (AI4Bharat: anudesh + dolly) using Unsloth + LoRA, on a single L4 GPU. Then I ran an honest side-by-side eval: base Gemma-4 vs my fine-tune, across 25 Hindi prompts. The results were interesting šŸ‘‡ āœ… My fine-tune is more concise ā€” ask for "3 tips" and it gives exactly 3. Base writes a 1,200-character essay. āœ… Pure native Hindi ā€” base keeps slipping into English ("ą¤øą¤‚ą¤¤ą„ą¤²ą¤æą¤¤ ą¤†ą¤¹ą¤¾ą¤° (Eat a Balanced Diet)", "ą¤¤ą¤¾ą¤°ą¤¾ (Star)"). My fine-tune stays in clean Hindi. āœ… Tighter instruction-following ā€” ask for a "short message" and it gives one, not a menu of options. āš–ļø And to be honest: base Gemma-4 is more detailed and comprehensive. I didn't build a "smarter" model ā€” I built a focused, Hindi-native, edge-friendly one that runs as a 5GB GGUF (Q4) on CPU. šŸ”— Try it: Live demo (CPU): https://huggingface.co/spaces/pankajpandey-dev/gemma-4-e4b-hindi-demo GGUF (Ollama/llama.cpp): https://huggingface.co/pankajpandey-dev/gemma-4-e4b-hindi-instruct-GGUF 16-bit model: https://huggingface.co/pankajpandey-dev/gemma-4-e4b-hindi-instruct Built with @unsloth Ā· Data by @ai4bharat šŸ™ #Hindi #LLM #Gemma #Unsloth #IndicNLP #GGUF
View all activity

Organizations

LocalLLaMA's profile picture Hugging Science's profile picture nanochat students's profile picture