project-llm-trainer 0.7.5__py3-none-any.whl → 0.7.7__py3-none-any.whl

project_llm_trainer-0.7.7.data/scripts/ddp_train

@@ -0,0 +1,24 @@
+#!python
+
+if __name__ == '__main__':
+    import os, sys
+    arguments = sys.argv[1:]
+    # file_name
+    run_file_name = arguments[0]
+
+    # cuda_visible_devive
+    if len(arguments) > 1:
+        # 0,1,2,3
+        cuda_visible_devive = arguments[1]
+    else:
+        cuda_visible_devive = None
+
+    os.environ['PARALLEL_TYPE'] = 'ddp'
+
+    if cuda_visible_devive:
+        os.environ['CUDA_VISIBLE_DEVICES'] = cuda_visible_devive
+
+    command = f'torchrun --standalone --nproc_per_node=gpu {run_file_name}'
+
+    print(f'run command {command}')
+    os.system(command)

project_llm_trainer-0.7.7.data/scripts/ds_train

@@ -0,0 +1,29 @@
+#!python
+
+if __name__ == '__main__':
+    import os, sys
+    arguments = sys.argv[1:]
+    # file_name
+    run_file_name = arguments[0]
+
+    # cuda_visible_devive
+    if len(arguments) > 1:
+        # 0,1,2,3
+        cuda_visible_devive = arguments[1]
+    else:
+        cuda_visible_devive = None
+
+    # cuda location
+    if len(arguments) > 2:
+        cuda_loc = arguments[2]
+    else:
+        cuda_loc = 'localhost'
+
+    os.environ['PARALLEL_TYPE'] = 'ds'
+
+    cuda_ctrl = f' --include {cuda_loc}:{cuda_visible_devive}' if cuda_visible_devive else ''
+
+    command = f'deepspeed{cuda_ctrl} {run_file_name}'
+
+    print(f'run command {command}')
+    os.system(command)

{project_llm_trainer-0.7.5.data → project_llm_trainer-0.7.7.data}/scripts/smart_train

@@ -2,9 +2,24 @@
 
 if __name__ == '__main__':
     import os, sys, torch
+
     arguments = sys.argv[1:]
+    # file name
     run_file_name = arguments[0]
 
+    # cuda_visible_devive
+    if len(arguments) > 1:
+        # 0,1,2,3
+        cuda_visible_devive = arguments[1]
+    else:
+        cuda_visible_devive = None
+
+    # cuda location
+    if len(arguments) > 2:
+        cuda_loc = arguments[2]
+    else:
+        cuda_loc = 'localhost'
+
     try:
         import deepspeed
         parallel_type = 'ds'
@@ -18,11 +33,14 @@ if __name__ == '__main__':
     os.environ['PARALLEL_TYPE'] = parallel_type
 
     if parallel_type == 'ds':
-        command = f'deepspeed {run_file_name}'
+        cuda_ctrl = f' --include {cuda_loc}:{cuda_visible_devive}' if cuda_visible_devive else ''
+        command = f'deepspeed{cuda_ctrl} {run_file_name}'
     elif parallel_type == 'ddp':
+        if cuda_visible_devive:
+            os.environ['CUDA_VISIBLE_DEVICES'] = cuda_visible_devive
         command = f'torchrun --standalone --nproc_per_node=gpu {run_file_name}'
     else:
         command = f'python3 {run_file_name}'
 
-    print(f'real command is {command}')
+    print(f'run command {command}')
     os.system(command)

{project_llm_trainer-0.7.5.dist-info → project_llm_trainer-0.7.7.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: project_llm_trainer
-Version: 0.7.5
+Version: 0.7.7
 Summary: LLM and VLM trainer
 Author: qibin
 Author-email: qibin0506@gmail.com

{project_llm_trainer-0.7.5.dist-info → project_llm_trainer-0.7.7.dist-info}/RECORD

@@ -20,14 +20,14 @@ llm_trainer/tools.py,sha256=5op5qrjjkK-Lr9oes5VxIVnOVYOYGoAdlIJq9mPUf64,2637
 llm_trainer/train_configs.py,sha256=N3ykM1uaLHcSNRC8ErYIxp9VYhSP7voJyAP-2D4ZJe0,7574
 llm_trainer/trainer.py,sha256=jS31zEXIIj9BoPTPlmaGYq61x72HGCjKfS2u3_gOkDk,27924
 llm_trainer/utils.py,sha256=xcdzpvPvXRKqsOK2yB7PZ9GmOvZMDFcglDPUZY2hJTY,11484
-project_llm_trainer-0.7.5.data/scripts/calc_intermediate_size,sha256=AggpgNHokJiJMbEtVdOnolqr_4bH3i1UYuZNEAzC2Gc,460
-project_llm_trainer-0.7.5.data/scripts/ddp_train,sha256=x81AasaN2-9TwARFFF1l7iV1LmfMQ0bLw0i_CGbOwSw,299
-project_llm_trainer-0.7.5.data/scripts/ds_train,sha256=qL3qc3TcedBCw98UZUjW07ONcErRawLE1HymW2AmscA,265
-project_llm_trainer-0.7.5.data/scripts/plot_loss,sha256=MzFcdJESlVr1srj4Td6-AxPGUKkfB_QEcJwm0Bd-5fU,910
-project_llm_trainer-0.7.5.data/scripts/plot_lr,sha256=w_7XR_x3KYYyboeOVAeu_I4fveLFI-C0wBmRrNlmWUI,894
-project_llm_trainer-0.7.5.data/scripts/py_train,sha256=tOp9TquORQeU8XN5H7OVIk5O0Ypwi34p_GENxTwgwdk,265
-project_llm_trainer-0.7.5.data/scripts/smart_train,sha256=Pmt4Q0to4Hoz82iB9uFPZuz7uahNUbfE7FR1940EBy8,716
-project_llm_trainer-0.7.5.dist-info/METADATA,sha256=9DcoFVuXDrhxZOVWF1Ouzk7NF6NTEnpBTkg1n6bMCYQ,195
-project_llm_trainer-0.7.5.dist-info/WHEEL,sha256=Nw36Djuh_5VDukK0H78QzOX-_FQEo6V37m3nkm96gtU,91
-project_llm_trainer-0.7.5.dist-info/top_level.txt,sha256=LtRFg28i0QIG7iBCD2t095oSco99LCtkijibS9cMGik,12
-project_llm_trainer-0.7.5.dist-info/RECORD,,
+project_llm_trainer-0.7.7.data/scripts/calc_intermediate_size,sha256=AggpgNHokJiJMbEtVdOnolqr_4bH3i1UYuZNEAzC2Gc,460
+project_llm_trainer-0.7.7.data/scripts/ddp_train,sha256=Z-309mM56CN0m3bxoeC5us4LUuwuNnoiOm3-fDdLMjQ,566
+project_llm_trainer-0.7.7.data/scripts/ds_train,sha256=3nXNNKmYI7miqyBdf-Ijl_rW1cGIKrAMZ1CSswN_gGo,665
+project_llm_trainer-0.7.7.data/scripts/plot_loss,sha256=MzFcdJESlVr1srj4Td6-AxPGUKkfB_QEcJwm0Bd-5fU,910
+project_llm_trainer-0.7.7.data/scripts/plot_lr,sha256=w_7XR_x3KYYyboeOVAeu_I4fveLFI-C0wBmRrNlmWUI,894
+project_llm_trainer-0.7.7.data/scripts/py_train,sha256=tOp9TquORQeU8XN5H7OVIk5O0Ypwi34p_GENxTwgwdk,265
+project_llm_trainer-0.7.7.data/scripts/smart_train,sha256=3oLIDuuqb4U4TU1lXy9V8lw_0gIf7i8tGsxlQ_s6bro,1220
+project_llm_trainer-0.7.7.dist-info/METADATA,sha256=1O3xW3QM5aJgk1EESixX5DjxQ5ReX_pikUjI5x2qOvk,195
+project_llm_trainer-0.7.7.dist-info/WHEEL,sha256=Nw36Djuh_5VDukK0H78QzOX-_FQEo6V37m3nkm96gtU,91
+project_llm_trainer-0.7.7.dist-info/top_level.txt,sha256=LtRFg28i0QIG7iBCD2t095oSco99LCtkijibS9cMGik,12
+project_llm_trainer-0.7.7.dist-info/RECORD,,

project_llm_trainer-0.7.5.data/scripts/ddp_train

@@ -1,12 +0,0 @@
-#!python
-
-if __name__ == '__main__':
-    import os, sys
-    arguments = sys.argv[1:]
-    run_file_name = arguments[0]
-
-    os.environ['PARALLEL_TYPE'] = 'ddp'
-    command = f'torchrun --standalone --nproc_per_node=gpu {run_file_name}'
-
-    print(f'real command is {command}')
-    os.system(command)

project_llm_trainer-0.7.5.data/scripts/ds_train

@@ -1,12 +0,0 @@
-#!python
-
-if __name__ == '__main__':
-    import os, sys
-    arguments = sys.argv[1:]
-    run_file_name = arguments[0]
-
-    os.environ['PARALLEL_TYPE'] = 'ds'
-    command = f'deepspeed {run_file_name}'
-
-    print(f'real command is {command}')
-    os.system(command)