量化qwen2-audio-7b 报错

[Liufeiran123]

详细报错信息 Traceback (most recent call last): File "/mnt/d/ai/swift/swift/cli/export.py", line 5, in export_main() File "/mnt/d/ai/swift/swift/utils/run_utils.py", line 32, in x_main result = llm_x(args, *kwargs) File "/mnt/d/ai/swift/swift/llm/export.py", line 263, in llm_export gptq_quantizer = gptq_model_quantize(model, template.tokenizer, args.quant_batch_size) File "/mnt/d/ai/swift/swift/llm/export.py", line 162, in gptq_model_quantize gptq_quantizer.quantize_model(model, tokenizer) File "/mnt/d/envs_swift/lib/python3.10/site-packages/torch/utils/_contextlib.py", line 116, in decorate_context return func(args, **kwargs) File "/mnt/d/envs_swift/lib/python3.10/site-packages/optimum/gptq/quantizer.py", line 570, in quantize_model self.pack_model(model=model, quantizers=quantizers) File "/mnt/d/envs_swift/lib/python3.10/site-packages/optimum/gptq/quantizer.py", line 641, in pack_model self._replace_by_quant_layers(model, quantizers) File "/mnt/d/envs_swift/lib/python3.10/site-packages/optimum/gptq/quantizer.py", line 299, in _replace_by_quant_layers self._replace_by_quant_layers(child, names, name + "." + name1 if name != "" else name1) File "/mnt/d/envs_swift/lib/python3.10/site-packages/optimum/gptq/quantizer.py", line 267, in _replace_by_quant_layers layer = getattr(module, attr) File "/mnt/d/envs_swift/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1729, in getattr raise AttributeError(f"'{type(self).name}' object has no attribute '{name}'") AttributeError: 'Qwen2AudioEncoder' object has no attribute '_supports_sdpa'

[slin000111]

关于此报错的命令行有吗？

[Liufeiran123]

swift export --model_type qwen2-audio-7b-instruct --quant_bits 4 --dataset alpaca-zh --quant_method gptq

[Jintao-Huang]

我这里测试一切正常啊 你的transformers版本是?

[Liufeiran123]

那很奇怪了，我的 transformers 版本为 4.45.2

[Liufeiran123]

optimum 1.23.1