feat(transformers/models): add phi (#1073)

* phi

phi

phi

phi

fix

* fix

* fix

* fix

* fix

* fix

Author: LiTingyu1997

mindone/transformers/__init__.py

@@ -225,6 +225,13 @@
     PersimmonModel,
     PersimmonPreTrainedModel,
 )
+from .models.phi import (
+    PhiForCausalLM,
+    PhiForSequenceClassification,
+    PhiForTokenClassification,
+    PhiModel,
+    PhiPreTrainedModel,
+)
 from .models.phi3 import (
     Phi3ForCausalLM,
     Phi3ForSequenceClassification,

mindone/transformers/models/__init__.py

@@ -31,6 +31,7 @@
     mixtral,
     mobilebert,
     persimmon,
+    phi,
     phi3,
     qwen2,
     qwen2_5_omni,

mindone/transformers/models/auto/configuration_auto.py

@@ -63,6 +63,7 @@
         ("mt5", "MT5Config"),
         ("megatron-bert", "MegatronBertConfig"),
         ("mixtral", "MixtralConfig"),
+        ("phi", "PhiConfig"),
         ("phi3", "Phi3Config"),
         ("qwen2", "Qwen2Config"),
         ("qwen2_5_vl", "Qwen2_5_VLConfig"),
@@ -130,6 +131,7 @@
         ("mt5", "MT5"),
         ("megatron-bert", "Megatron-BERT"),
         ("mixtral", "Mixtral"),
+        ("phi", "Phi"),
         ("phi3", "Phi3"),
         ("qwen2", "Qwen2"),
         ("qwen2_5_vl", "Qwen2_5_VL"),

mindone/transformers/models/auto/modeling_auto.py

@@ -64,6 +64,7 @@
         ("mt5", "MT5Model"),
         ("megatron-bert", "MegatronBertModel"),
         ("mixtral", "MixtralModel"),
+        ("phi", "PhiModel"),
         ("phi3", "Phi3Model"),
         ("qwen2", "Qwen2Model"),
         ("qwen2_5_vl", "Qwen2_5_VLModel"),
@@ -145,6 +146,7 @@
         ("granitemoeshared", "GraniteMoeSharedForCausalLM"),
         ("llama", "LlamaForCausalLM"),
         ("megatron-bert", "MegatronBertForCausalLM"),
+        ("phi", "PhiForCausalLM"),
         ("phi3", "Phi3ForCausalLM"),
         ("mixtral", "MixtralForCausalLM"),
         ("qwen2", "Qwen2ForCausalLM"),
@@ -316,6 +318,7 @@
         ("mt5", "MT5ForSequenceClassification"),
         ("megatron-bert", "MegatronBertForSequenceClassification"),
         ("mixtral", "MixtralForSequenceClassification"),
+        ("phi", "PhiForSequenceClassification"),
         ("phi3", "Phi3ForSequenceClassification"),
         ("qwen2", "Qwen2ForSequenceClassification"),
         ("qwen3", "Qwen3ForSequenceClassification"),
@@ -368,6 +371,7 @@
         ("persimmon", "PersimmonForTokenClassification"),
         ("megatron-bert", "MegatronBertForTokenClassification"),
         ("mixtral", "MixtralForTokenClassification"),
+        ("phi", "PhiForTokenClassification"),
         ("phi3", "Phi3ForTokenClassification"),
         ("qwen2", "Qwen2ForTokenClassification"),
         ("qwen3", "Qwen3ForTokenClassification"),

mindone/transformers/models/phi/__init__.py

@@ -0,0 +1,7 @@
+from .modeling_phi import (
+    PhiForCausalLM,
+    PhiForSequenceClassification,
+    PhiForTokenClassification,
+    PhiModel,
+    PhiPreTrainedModel,
+)