Full fine-tuning recipe: SFT (supervised fine-tuning) on Llama 3.1 70B via OpenRLHF, targeting 4x A100 40GB, with data mix and eval plan.
Full fine-tuning recipe: SFT (supervised fine-tuning) on Mistral Small 3 via DeepSpeed, targeting 8x H100, with data mix and eval plan.
Full fine-tuning recipe: SFT (supervised fine-tuning) on Qwen 2.5 7B via Unsloth, targeting 8x H100, with data mix and eval plan.
Full fine-tuning recipe: SFT (supervised fine-tuning) on Qwen 2.5 32B via OpenRLHF, targeting 8x H100, with data mix and eval plan.
Full fine-tuning recipe: SFT (supervised fine-tuning) on Gemma 2 9B via DeepSpeed, targeting single RTX 4090 (24GB), with data mix and eval plan.
Full fine-tuning recipe: SFT (supervised fine-tuning) on Gemma 2 27B via Hugging Face TRL, targeting single RTX 4090 (24GB), with data mix and eval plan.
Full fine-tuning recipe: SFT (supervised fine-tuning) on Phi-4 via Megatron-LM, targeting single RTX 3090 (24GB), with data mix and eval plan.
Full fine-tuning recipe: SFT (supervised fine-tuning) on DeepSeek-V3 base via FSDP, targeting single RTX 3090 (24GB), with data mix and eval plan.
Full fine-tuning recipe: SFT (supervised fine-tuning) on Mixtral 8x7B via Hugging Face TRL, targeting single RTX 3090 (24GB), with data mix and eval plan.
Full fine-tuning recipe: SFT (supervised fine-tuning) on Yi 1.5 34B via Megatron-LM, targeting 2x RTX 4090, with data mix and eval plan.
Full fine-tuning recipe: SFT (supervised fine-tuning) on Llama 3.3 70B via FSDP, targeting 2x RTX 4090, with data mix and eval plan.
Full fine-tuning recipe: SFT (supervised fine-tuning) on Llama 3.1 70B via LLaMA-Factory, targeting AWS g5.12xlarge, with data mix and eval plan.