TypeError: Gemma2DecoderLayer.forward() missing 1 required positional argument: 'position_embeddings'

[Charimanhua]

请问有人遇到过这个问题么？在运行BAAI/bge-reranker-v2.5-gemma2-lightweight示例代码：
`from FlagEmbedding import LightWeightFlagLLMReranker
reranker = LightWeightFlagLLMReranker('BAAI/bge-reranker-v2.5-gemma2-lightweight', use_fp16=True) # Setting use_fp16 to True speeds up computation with a slight performance degradation

score = reranker.compute_score(['query', 'passage'], cutoff_layers=[28], compress_ratio=2, compress_layer=[24, 40]) # Adjusting 'cutoff_layers' to pick which layers are used for computing the score.
print(score)

scores = reranker.compute_score([['what is panda?', 'hi'], ['what is panda?', 'The giant panda (Ailuropoda melanoleuca), sometimes called a panda bear or simply panda, is a bear species endemic to China.']], cutoff_layers=[28], compress_ratio=2, compress_layer=[24, 40])
print(scores)
报错：TypeError Traceback (most recent call last)
Cell In[1], line 4
1 from FlagEmbedding import LightWeightFlagLLMReranker
2 reranker = LightWeightFlagLLMReranker('bge-reranker-v2.5-gemma2-lightweight', use_fp16=True) # Setting use_fp16 to True speeds up computation with a slight performance degradation
----> 4 score = reranker.compute_score(['query', 'passage'], cutoff_layers=[28], compress_ratio=2, compress_layer=[24, 40]) # Adjusting 'cutoff_layers' to pick which layers are used for computing the score.
5 print(score)
7 scores = reranker.compute_score([['what is panda?', 'hi'], ['what is panda?', 'The giant panda (Ailuropoda melanoleuca), sometimes called a panda bear or simply panda, is a bear species endemic to China.']], cutoff_layers=[28], compress_ratio=2, compress_layer=[24, 40])

File ~/autodl-tmp/FlagEmbedding/FlagEmbedding/abc/inference/AbsReranker.py:217, in AbsReranker.compute_score(self, sentence_pairs, **kwargs)
214 sentence_pairs = self.get_detailed_inputs(sentence_pairs)
216 if isinstance(sentence_pairs, str) or len(self.target_devices) == 1:
--> 217 return self.compute_score_single_gpu(
218 sentence_pairs,
219 device=self.target_devices[0],
220 **kwargs
221 )
223 if self.pool is None:
224 self.pool = self.start_multi_process_pool()

File ~/miniconda3/lib/python3.12/site-packages/torch/utils/_contextlib.py:115, in context_decorator..decorate_context(*args, **kwargs)
112 @functools.wraps(func)
113 def decorate_context(*args, **kwargs):
114 with ctx_factory():
--> 115 return func(*args, **kwargs)

File ~/autodl-tmp/FlagEmbedding/FlagEmbedding/inference/reranker/decoder_only/lightweight.py:359, in LightweightLLMReranker.compute_score_single_gpu(self, sentence_pairs, batch_size, query_max_length, max_length, cutoff_layers, compress_layer, compress_layers, compress_ratio, prompt, normalize, device, **kwargs)
348 batch_inputs = collater_instance([
349 [{
350 'input_ids': item['input_ids'],
(...)
354 prompt_lengths
355 ])[0]
357 batch_inputs = {key: val.to(device) for key, val in batch_inputs.items()}
--> 359 self.model(
360 **batch_inputs,
361 output_hidden_states=True,
362 compress_layer=compress_layers,
363 compress_ratio=compress_ratio,
364 query_lengths=query_lengths,
365 prompt_lengths=prompt_lengths,
366 cutoff_layers=cutoff_layers
367 )
368 flag = True
369 except RuntimeError as e:

File ~/miniconda3/lib/python3.12/site-packages/torch/nn/modules/module.py:1532, in Module._wrapped_call_impl(self, *args, **kwargs)
1530 return self._compiled_call_impl(*args, **kwargs) # type: ignore[misc]
1531 else:
-> 1532 return self._call_impl(*args, **kwargs)

File ~/miniconda3/lib/python3.12/site-packages/torch/nn/modules/module.py:1541, in Module._call_impl(self, *args, **kwargs)
1536 # If we don't have any hooks, we want to skip the rest of the logic in
1537 # this function, and just call forward.
1538 if not (self._backward_hooks or self._backward_pre_hooks or self._forward_hooks or self._forward_pre_hooks
1539 or _global_backward_pre_hooks or _global_backward_hooks
1540 or _global_forward_hooks or _global_forward_pre_hooks):
-> 1541 return forward_call(*args, **kwargs)
1543 try:
1544 result = None

File ~/autodl-tmp/FlagEmbedding/FlagEmbedding/inference/reranker/decoder_only/models/gemma_model.py:599, in CostWiseGemmaForCausalLM.forward(self, input_ids, attention_mask, position_ids, past_key_values, inputs_embeds, labels, use_cache, output_attentions, output_hidden_states, return_dict, cache_position, compress_layer, compress_ratio, cutoff_layers, query_lengths, prompt_lengths)
596 raise ValueError(f"Your cutoff layers must in [{self.config.start_layer}, {self.config.num_hidden_layers}]")
598 # decoder outputs consists of (dec_features, layer_state, dec_hidden, dec_attn)
--> 599 outputs = self.model(
600 input_ids=input_ids,
601 attention_mask=attention_mask,
602 position_ids=position_ids,
603 past_key_values=past_key_values,
604 inputs_embeds=inputs_embeds,
605 use_cache=use_cache,
606 output_attentions=output_attentions,
607 output_hidden_states=output_hidden_states,
608 return_dict=return_dict,
609 cache_position=cache_position,
610 compress_layer=compress_layer,
611 compress_ratio=compress_ratio,
612 query_lengths=query_lengths,
613 prompt_lengths=prompt_lengths,
614 cutoff_layers=cutoff_layers,
615 )
617 if not self.config.layer_wise:
618 hidden_states = outputs[0]

File ~/miniconda3/lib/python3.12/site-packages/torch/nn/modules/module.py:1532, in Module._wrapped_call_impl(self, *args, **kwargs)
1530 return self._compiled_call_impl(*args, **kwargs) # type: ignore[misc]
1531 else:
-> 1532 return self._call_impl(*args, **kwargs)

File ~/miniconda3/lib/python3.12/site-packages/torch/nn/modules/module.py:1541, in Module._call_impl(self, *args, **kwargs)
1536 # If we don't have any hooks, we want to skip the rest of the logic in
1537 # this function, and just call forward.
1538 if not (self._backward_hooks or self._backward_pre_hooks or self._forward_hooks or self._forward_pre_hooks
1539 or _global_backward_pre_hooks or _global_backward_hooks
1540 or _global_forward_hooks or _global_forward_pre_hooks):
-> 1541 return forward_call(*args, **kwargs)
1543 try:
1544 result = None

File ~/autodl-tmp/FlagEmbedding/FlagEmbedding/inference/reranker/decoder_only/models/gemma_model.py:395, in CostWiseGemmaModel.forward(self, input_ids, attention_mask, position_ids, past_key_values, inputs_embeds, use_cache, output_attentions, output_hidden_states, return_dict, cache_position, compress_layer, compress_ratio, cutoff_layers, query_lengths, prompt_lengths)
384 layer_outputs = self._gradient_checkpointing_func(
385 decoder_layer.call,
386 hidden_states,
(...)
392 cache_position,
393 )
394 else:
--> 395 layer_outputs = decoder_layer(
396 hidden_states,
397 attention_mask=causal_mask,
398 position_ids=position_ids,
399 past_key_value=past_key_values,
400 output_attentions=output_attentions,
401 use_cache=use_cache,
402 cache_position=cache_position,
403 )
405 hidden_states = layer_outputs[0]
407 if output_attentions:

File ~/miniconda3/lib/python3.12/site-packages/torch/nn/modules/module.py:1532, in Module._wrapped_call_impl(self, *args, **kwargs)
1530 return self._compiled_call_impl(*args, **kwargs) # type: ignore[misc]
1531 else:
-> 1532 return self._call_impl(*args, **kwargs)

File ~/miniconda3/lib/python3.12/site-packages/torch/nn/modules/module.py:1541, in Module._call_impl(self, *args, **kwargs)
1536 # If we don't have any hooks, we want to skip the rest of the logic in
1537 # this function, and just call forward.
1538 if not (self._backward_hooks or self._backward_pre_hooks or self._forward_hooks or self._forward_pre_hooks
1539 or _global_backward_pre_hooks or _global_backward_hooks
1540 or _global_forward_hooks or _global_forward_pre_hooks):
-> 1541 return forward_call(*args, **kwargs)
1543 try:
1544 result = None

TypeError: Gemma2DecoderLayer.forward() missing 1 required positional argument: 'position_embeddings'`