bge-reranker-v2-minicpm-layerwise 微调出现错误

作者: yvway创建于 2026年5月19日更新于 2026年5月19日

[rank0]: Traceback (most recent call last): [rank0]: File "/home/yw/.conda/envs/bge/lib/python3.10/runpy.py", line 196, in _run_module_as_main [rank0]: return _run_code(code, main_globals, None, [rank0]: File "/home/yw/.conda/envs/bge/lib/python3.10/runpy.py", line 86, in _run_code [rank0]: exec(code, run_globals) [rank0]: File "/home/yw/.conda/envs/bge/lib/python3.10/site-packages/FlagEmbedding/finetune/reranker/decoder_only/layerwise/main.py", line 30, in [rank0]: main() [rank0]: File "/home/yw/.conda/envs/bge/lib/python3.10/site-packages/FlagEmbedding/finetune/reranker/decoder_only/layerwise/main.py", line 21, in main [rank0]: runner = DecoderOnlyRerankerRunner( [rank0]: File "/home/yw/.conda/envs/bge/lib/python3.10/site-packages/FlagEmbedding/finetune/reranker/decoder_only/layerwise/runner.py", line 33, in init [rank0]: super().init(model_args, data_args, training_args) [rank0]: File "/home/yw/.conda/envs/bge/lib/python3.10/site-packages/FlagEmbedding/abc/finetune/reranker/AbsRunner.py", line 73, in init [rank0]: self.tokenizer, self.model = self.load_tokenizer_and_model() [rank0]: File "/home/yw/.conda/envs/bge/lib/python3.10/site-packages/FlagEmbedding/finetune/reranker/decoder_only/layerwise/runner.py", line 68, in load_tokenizer_and_model [rank0]: base_model = get_model(self.model_args, tokenizer('Yes', add_special_tokens=False)['input_ids'][-1]) [rank0]: File "/home/yw/.conda/envs/bge/lib/python3.10/site-packages/FlagEmbedding/finetune/reranker/decoder_only/layerwise/load_model.py", line 131, in get_model [rank0]: model = LayerWiseMiniCPMForCausalLM.from_pretrained( [rank0]: File "/home/yw/.conda/envs/bge/lib/python3.10/site-packages/transformers/modeling_utils.py", line 4072, in from_pretrained [rank0]: model = cls(config, *model_args, **model_kwargs) [rank0]: File "/home/yw/.conda/envs/bge/lib/python3.10/site-packages/FlagEmbedding/finetune/reranker/decoder_only/layerwise/modeling_minicpm_reranker.py", line 1150, in init [rank0]: self.model = LayerWiseMiniCPMModel(config) [rank0]: File "/home/yw/.conda/envs/bge/lib/python3.10/site-packages/FlagEmbedding/finetune/reranker/decoder_only/layerwise/modeling_minicpm_reranker.py", line 973, in init [rank0]: [MiniCPMDecoderLayer(config, layer_idx) for layer_idx in range(config.num_hidden_layers)] [rank0]: File "/home/yw/.conda/envs/bge/lib/python3.10/site-packages/FlagEmbedding/finetune/reranker/decoder_only/layerwise/modeling_minicpm_reranker.py", line 973, in [rank0]: [MiniCPMDecoderLayer(config, layer_idx) for layer_idx in range(config.num_hidden_layers)] [rank0]: File "/home/yw/.conda/envs/bge/lib/python3.10/site-packages/FlagEmbedding/finetune/reranker/decoder_only/layerwise/modeling_minicpm_reranker.py", line 770, in init [rank0]: self.self_attn = MINICPM_ATTENTION_CLASSES[config._attn_implementation](config=config, layer_idx=layer_idx) [rank0]: File

内容来源: FlagOpen/FlagEmbedding