Conferred with PPO- Fine-tuned 16B code-gen LLM (base and chat) for an enterprise language on text2code, code2text, and autocomplete. - Attained remarkable scores of 0.5 and 0.7 in zero-shot and few-shot eval on self-developed BLEU k-pass benchmarks. - Leveraged advanced techniques such as Low Rank Adaptation (LoRA) from Parameter Efficient Fine-Tuning (PEFT) library, gradient accumulation and 8-bit quantization for efficient computation on a low-resource multi-GPU setup. -… Show more Conferred with PPO- Fine-tuned 16B code-gen LLM (base and chat) for an enterprise language on text2code, code2text, and autocomplete. - Attained remarkable scores of 0.5 and 0.7 in zero-shot and few-shot eval on self-developed BLEU k-pass benchmarks. - Leveraged advanced techniques such as Low Rank Adaptation (LoRA) from Parameter Efficient Fine-Tuning (PEFT) library, gradient accumulation and 8-bit quantization for efficient computation on a low-resource multi-GPU setup. - Curated a high-quality dataset employing web scraping/crawling, PII, near deduplication and prompt engineering. Show less