project-llm-trainer 0.4.11__py3-none-any.whl → 0.4.12__py3-none-any.whl

llm_trainer/train_configs.py

@@ -67,7 +67,7 @@ class DsFp16Config:
     initial_scale_power: int = 16
     hysteresis: int = 2
     min_loss_scale: int = 1
-    fp16_opt_level: Optional[str] = '02'
+    fp16_opt_level: Optional[str] = 'O2'
 
 
 @dataclass(kw_only=True)
@@ -77,9 +77,9 @@ class DsBf16Config:
 
 @dataclass(kw_only=True)
 class DsConfig:
-    zero_config: Optional[DsZeROConfig] = DsZero3Config()
-    fp16_config: Optional[DsFp16Config] = DsFp16Config()
-    bf16_config: Optional[DsBf16Config] = DsBf16Config()
+    zero_config: Optional[DsZeROConfig] = field(default_factory=DsZero3Config)
+    fp16_config: Optional[DsFp16Config] = field(default_factory=DsFp16Config)
+    bf16_config: Optional[DsBf16Config] = field(default_factory=DsBf16Config)
     gradient_clipping: Optional[float] = 1.0
     activation_checkpointing: Optional[DsActivationCheckpointingConfig] = None
 
@@ -224,14 +224,14 @@ class TrainConfig:
     model_config: Union[ModelConfig, VLMConfig]
 
     file_dataset: FileDataset
-    data_loader_config: DataLoaderConfig = DataLoaderConfig()
+    data_loader_config: DataLoaderConfig = field(default_factory=DataLoaderConfig)
     image_tags_file_dataset: Optional[FileDataset] = None
 
-    loss_config: LossConfig = LossConfig()
-    lr_config: LrConfig = LrConfig()
+    loss_config: LossConfig = field(default_factory=LossConfig)
+    lr_config: LrConfig = field(default_factory=LrConfig)
 
-    ds_config: DsConfig = DsConfig()
-    fsdp_config: FsdpConfig = FsdpConfig()
+    ds_config: DsConfig = field(default_factory=DsConfig)
+    fsdp_config: FsdpConfig = field(default_factory=FsdpConfig)
 
     kd_config: Optional[KDConfig] = None
     dpo_config: Optional[DPOConfig] = None
@@ -241,7 +241,7 @@ class TrainConfig:
     gradient_accumulation_steps: int = 0
     eval_batch_interval: int = 100
 
-    eval_config: EvalConfig = EvalConfig()
+    eval_config: EvalConfig = field(default_factory=EvalConfig)
     pixel_values_provider: Optional[Callable[[list[str]], torch.Tensor]] = None
 
     init_state_dict: Optional[Mapping[str, Any]] = None

{project_llm_trainer-0.4.11.dist-info → project_llm_trainer-0.4.12.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: project_llm_trainer
-Version: 0.4.11
+Version: 0.4.12
 Summary: LLM and VLM trainer
 Author: qibin
 Author-email: qibin0506@gmail.com

{project_llm_trainer-0.4.11.dist-info → project_llm_trainer-0.4.12.dist-info}/RECORD

@@ -19,17 +19,17 @@ llm_trainer/scheduler.py,sha256=Xz8HhwoRMjRe41sf_NHhpZfkTlEs0I2MYusvMY6hCVw,3531
 llm_trainer/sft_trainer.py,sha256=gxQA7T1o1QGUsHp2CX1Qb_fO5LppBJuNbc0H4ixCYUA,1783
 llm_trainer/tokenizer.py,sha256=A7TYYUbtPf75kjCvWP7yBui4xZBObMk2aPem62YpwpY,6776
 llm_trainer/tools.py,sha256=O45-20wRmh-nyTfU-U-XtjbKAoe7boEIsUvWT_NaKx4,3041
-llm_trainer/train_configs.py,sha256=4sM96SOgwcn6jBGtbG5-qDZbJjiHVB6l7FWqdq7hbj0,7979
+llm_trainer/train_configs.py,sha256=HKzH3nfMT1-SW4Htwa0KqYtMd6FAJcthR5IEo6di8us,8168
 llm_trainer/trainer.py,sha256=pUtJVRosn54j1hn76CFAptJcAsrDo59H6p8NMkg2zt4,25521
 llm_trainer/utils.py,sha256=-ivhMF0d999va13S1wt2uBvtVw8Nvr3uBzhaUFKL04Q,6826
-project_llm_trainer-0.4.11.data/scripts/calc_intermediate_size,sha256=AggpgNHokJiJMbEtVdOnolqr_4bH3i1UYuZNEAzC2Gc,460
-project_llm_trainer-0.4.11.data/scripts/ddp_train,sha256=x81AasaN2-9TwARFFF1l7iV1LmfMQ0bLw0i_CGbOwSw,299
-project_llm_trainer-0.4.11.data/scripts/ds_train,sha256=qL3qc3TcedBCw98UZUjW07ONcErRawLE1HymW2AmscA,265
-project_llm_trainer-0.4.11.data/scripts/plot_loss,sha256=MzFcdJESlVr1srj4Td6-AxPGUKkfB_QEcJwm0Bd-5fU,910
-project_llm_trainer-0.4.11.data/scripts/plot_lr,sha256=w_7XR_x3KYYyboeOVAeu_I4fveLFI-C0wBmRrNlmWUI,894
-project_llm_trainer-0.4.11.data/scripts/py_train,sha256=tOp9TquORQeU8XN5H7OVIk5O0Ypwi34p_GENxTwgwdk,265
-project_llm_trainer-0.4.11.data/scripts/smart_train,sha256=Pmt4Q0to4Hoz82iB9uFPZuz7uahNUbfE7FR1940EBy8,716
-project_llm_trainer-0.4.11.dist-info/METADATA,sha256=JEZo2-np0t_K-J6yapyAXsArpvYTmrSNGDsdy32kWas,196
-project_llm_trainer-0.4.11.dist-info/WHEEL,sha256=Nw36Djuh_5VDukK0H78QzOX-_FQEo6V37m3nkm96gtU,91
-project_llm_trainer-0.4.11.dist-info/top_level.txt,sha256=LtRFg28i0QIG7iBCD2t095oSco99LCtkijibS9cMGik,12
-project_llm_trainer-0.4.11.dist-info/RECORD,,
+project_llm_trainer-0.4.12.data/scripts/calc_intermediate_size,sha256=AggpgNHokJiJMbEtVdOnolqr_4bH3i1UYuZNEAzC2Gc,460
+project_llm_trainer-0.4.12.data/scripts/ddp_train,sha256=x81AasaN2-9TwARFFF1l7iV1LmfMQ0bLw0i_CGbOwSw,299
+project_llm_trainer-0.4.12.data/scripts/ds_train,sha256=qL3qc3TcedBCw98UZUjW07ONcErRawLE1HymW2AmscA,265
+project_llm_trainer-0.4.12.data/scripts/plot_loss,sha256=MzFcdJESlVr1srj4Td6-AxPGUKkfB_QEcJwm0Bd-5fU,910
+project_llm_trainer-0.4.12.data/scripts/plot_lr,sha256=w_7XR_x3KYYyboeOVAeu_I4fveLFI-C0wBmRrNlmWUI,894
+project_llm_trainer-0.4.12.data/scripts/py_train,sha256=tOp9TquORQeU8XN5H7OVIk5O0Ypwi34p_GENxTwgwdk,265
+project_llm_trainer-0.4.12.data/scripts/smart_train,sha256=Pmt4Q0to4Hoz82iB9uFPZuz7uahNUbfE7FR1940EBy8,716
+project_llm_trainer-0.4.12.dist-info/METADATA,sha256=W-HeRGlXi3bFsKIVE1FyQAh4Lcvo0SOXMNu-9YnACKQ,196
+project_llm_trainer-0.4.12.dist-info/WHEEL,sha256=Nw36Djuh_5VDukK0H78QzOX-_FQEo6V37m3nkm96gtU,91
+project_llm_trainer-0.4.12.dist-info/top_level.txt,sha256=LtRFg28i0QIG7iBCD2t095oSco99LCtkijibS9cMGik,12
+project_llm_trainer-0.4.12.dist-info/RECORD,,