fix bug when resuming generation with OpenAI models #328

Job	Run time
unit-tests-gpu (gpu-runner, 3.10, gpt2gpu) / unit-tests-on-gpu	14m 30s
unit-tests-gpu (gpu-runner, 3.10, hfllama_7b_gpu) / unit-tests-on-gpu	15m 30s
unit-tests-gpu (gpu-runner, 3.11, gpt2gpu) / unit-tests-on-gpu	14m 51s
unit-tests-gpu (gpu-runner, 3.10, phi2gpu) / unit-tests-on-gpu	14m 25s
unit-tests-gpu (gpu-runner, 3.11, hfllama_7b_gpu) / unit-tests-on-gpu	15m 37s
unit-tests-gpu (gpu-runner, 3.12, hfllama_7b_gpu) / unit-tests-on-gpu	16m 51s
unit-tests-gpu (gpu-runner, 3.11, phi2gpu) / unit-tests-on-gpu	14m 13s
unit-tests-gpu (gpu-runner, 3.12, gpt2gpu) / unit-tests-on-gpu	15m 37s
unit-tests-gpu (gpu-runner, 3.12, phi2gpu) / unit-tests-on-gpu	15m 6s
unit-tests-gpu (gpu-runner, 3.8, gpt2gpu) / unit-tests-on-gpu	14m 44s
unit-tests-gpu (gpu-runner, 3.8, hfllama_7b_gpu) / unit-tests-on-gpu	15m 23s
unit-tests-gpu (gpu-runner, 3.8, phi2gpu) / unit-tests-on-gpu	14m 37s
unit-tests-gpu (gpu-runner, 3.9, gpt2gpu) / unit-tests-on-gpu	14m 47s
unit-tests-gpu (gpu-runner, 3.9, hfllama_7b_gpu) / unit-tests-on-gpu	15m 42s
unit-tests-gpu (gpu-runner, 3.9, phi2gpu) / unit-tests-on-gpu	14m 23s
	3h 46m 16s

Provide feedback