admin/LLM-from-scratch
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

25.05.15:update modeling_dolly code and code comments

Browse Source
This commit is contained in:
coffee
2025-05-15 10:46:58 +08:00
parent 48019ee38a
commit d52f33aa4a
1 changed files with 3 additions and 8 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src_llm/modeling_dolly.py
+3 -8
View File
@@ -703,9 +703,7 @@ class DollyForCausalLM(DollyPreTrainedModel, GenerationMixin):
attentions=outputs.attentions, attentions=outputs.attentions,
) )
class DollyForSequenceClassification(DollyPreTrainedModel):
class Qwen3ForSequenceClassification(DollyPreTrainedModel):
def __init__(self, config): def __init__(self, config):
super().__init__(config) super().__init__(config)
self.num_labels = config.num_labels self.num_labels = config.num_labels
@@ -734,7 +732,6 @@ class Qwen3ForSequenceClassification(DollyPreTrainedModel):
output_attentions: Optional[bool] = None, output_attentions: Optional[bool] = None,
output_hidden_states: Optional[bool] = None, output_hidden_states: Optional[bool] = None,
) -> SequenceClassifierOutputWithPast: ) -> SequenceClassifierOutputWithPast:
transformer_outputs: BaseModelOutputWithPast = self.model( transformer_outputs: BaseModelOutputWithPast = self.model(
input_ids, input_ids,
@@ -785,7 +782,7 @@ class Qwen3ForSequenceClassification(DollyPreTrainedModel):
) )
class Qwen3ForTokenClassification(DollyPreTrainedModel): class Dolly3ForTokenClassification(DollyPreTrainedModel):
def __init__(self, config): def __init__(self, config):
super().__init__(config) super().__init__(config)
self.num_labels = config.num_labels self.num_labels = config.num_labels
@@ -822,7 +819,6 @@ class Qwen3ForTokenClassification(DollyPreTrainedModel):
output_hidden_states: Optional[bool] = None, output_hidden_states: Optional[bool] = None,
) -> TokenClassifierOutput: ) -> TokenClassifierOutput:
outputs: BaseModelOutputWithPast = self.model( outputs: BaseModelOutputWithPast = self.model(
input_ids, input_ids,
attention_mask=attention_mask, attention_mask=attention_mask,
@@ -849,7 +845,7 @@ class Qwen3ForTokenClassification(DollyPreTrainedModel):
) )
class Qwen3ForQuestionAnswering(DollyPreTrainedModel): class DollyForQuestionAnswering(DollyPreTrainedModel):
base_model_prefix = "transformer" base_model_prefix = "transformer"
def __init__(self, config): def __init__(self, config):
@@ -892,7 +888,6 @@ class Qwen3ForQuestionAnswering(DollyPreTrainedModel):
) )
sequence_output = outputs.last_hidden_state sequence_output = outputs.last_hidden_state
logits = self.qa_outputs(sequence_output) logits = self.qa_outputs(sequence_output)
start_logits, end_logits = logits.split(1, dim=-1) start_logits, end_logits = logits.split(1, dim=-1)
start_logits = start_logits.squeeze(-1).contiguous() start_logits = start_logits.squeeze(-1).contiguous()