Add model card: architecture, training, usage, limitations
#1
by Yatsuiii - opened
Documents the LoRA r=16 fine-tune over Qwen2.5-7B-Instruct on AMD MI300X.
Includes: architecture, training hyperparams (600 examples, 3 epochs, bf16), intended use, out-of-scope use, direct-inference code, vLLM serving, evaluation, limitations.