Phoenix - Your AI AssistantQLoRA for 70B Models — Fine-Tune Open Models on Multi-GPU
Delivery in
4 days
- Views 3
Amount of days required to complete work for this Offer as set by the freelancer.
Rating of the Offer as calculated from other buyers' reviews.
Average time for the freelancer to first reply on the workstream after purchase or contact on this Offer.
What you get with this Offer
I will run QLoRA fine-tuning of a 70B parameter model on your multi-GPU infrastructure — configuring model sharding across GPUs with DeepSpeed ZeRO or FSDP, 4-bit quantisation for memory efficiency, LoRA adapter configuration targeting the layers with highest return for your task type, and training execution with monitoring. Fine-tuning 70B models with QLoRA requires more infrastructure configuration than smaller models — model sharding across multiple GPUs with appropriate pipeline parallelism, quantised model loading that distributes correctly across devices, and gradient communication optimisation are all requirements that single-GPU QLoRA setups don't encounter.
The fine-tuning covers multi-GPU sharding configuration, 4-bit quantisation distribution, LoRA adapter targeting for 70B architecture, training execution with loss monitoring, evaluation, and adapter delivery.
The fine-tuning covers multi-GPU sharding configuration, 4-bit quantisation distribution, LoRA adapter targeting for 70B architecture, training execution with loss monitoring, evaluation, and adapter delivery.
What the Freelancer needs to start the work
Please share your training dataset, your target 70B model, your multi-GPU infrastructure (number of GPUs, VRAM each), your task and quality requirements, and your training timeline.