add helper to verify the correct model output file exists (#2245)

* add helper to verify the correct model output file exists

* more checks using helper

* chore: lint

* fix import and relora model check

* workaround for trl trainer saves

* remove stray print

tests/e2e/patched/test_4d_multipack_llama.py

@@ -5,7 +5,6 @@ E2E tests for multipack fft llama using 4d attention masks
 import logging
 import os
 import unittest
-from pathlib import Path
 
 from axolotl.cli import load_datasets
 from axolotl.common.cli import TrainerCliArgs
@@ -13,7 +12,7 @@ from axolotl.train import train
 from axolotl.utils.config import normalize_config
 from axolotl.utils.dict import DictDefault
 
-from ..utils import require_torch_2_3_1, with_temp_dir
+from ..utils import check_model_output_exists, require_torch_2_3_1, with_temp_dir
 
 LOG = logging.getLogger("axolotl.tests.e2e")
 os.environ["WANDB_DISABLED"] = "true"
@@ -67,7 +66,7 @@ class Test4dMultipackLlama(unittest.TestCase):
         dataset_meta = load_datasets(cfg=cfg, cli_args=cli_args)
 
         train(cfg=cfg, cli_args=cli_args, dataset_meta=dataset_meta)
-        assert (Path(temp_dir) / "adapter_model.bin").exists()
+        check_model_output_exists(temp_dir, cfg)
 
     @with_temp_dir
     def test_torch_lora_packing(self, temp_dir):
@@ -111,4 +110,4 @@ class Test4dMultipackLlama(unittest.TestCase):
         dataset_meta = load_datasets(cfg=cfg, cli_args=cli_args)
 
         train(cfg=cfg, cli_args=cli_args, dataset_meta=dataset_meta)
-        assert (Path(temp_dir) / "adapter_model.bin").exists()
+        check_model_output_exists(temp_dir, cfg)

tests/e2e/patched/test_fa_xentropy.py

@@ -4,7 +4,6 @@ E2E tests for lora llama
 
 import logging
 import os
-from pathlib import Path
 
 import pytest
 from transformers.utils import is_torch_bf16_gpu_available
@@ -15,7 +14,7 @@ from axolotl.train import train
 from axolotl.utils.config import normalize_config
 from axolotl.utils.dict import DictDefault
 
-from ..utils import check_tensorboard
+from ..utils import check_model_output_exists, check_tensorboard
 
 LOG = logging.getLogger("axolotl.tests.e2e")
 os.environ["WANDB_DISABLED"] = "true"
@@ -82,7 +81,7 @@ class TestFAXentropyLlama:
         dataset_meta = load_datasets(cfg=cfg, cli_args=cli_args)
 
         train(cfg=cfg, cli_args=cli_args, dataset_meta=dataset_meta)
-        assert (Path(temp_dir) / "adapter_model.bin").exists()
+        check_model_output_exists(temp_dir, cfg)
 
         check_tensorboard(
             temp_dir + "/runs", "train/train_loss", 1.5, "Train Loss is too high"

tests/e2e/patched/test_falcon_samplepack.py

@@ -5,7 +5,6 @@ E2E tests for falcon
 import logging
 import os
 import unittest
-from pathlib import Path
 
 from axolotl.cli import load_datasets
 from axolotl.common.cli import TrainerCliArgs
@@ -13,7 +12,7 @@ from axolotl.train import train
 from axolotl.utils.config import normalize_config
 from axolotl.utils.dict import DictDefault
 
-from ..utils import with_temp_dir
+from ..utils import check_model_output_exists, with_temp_dir
 
 LOG = logging.getLogger("axolotl.tests.e2e")
 os.environ["WANDB_DISABLED"] = "true"
@@ -69,7 +68,7 @@ class TestFalconPatched(unittest.TestCase):
         dataset_meta = load_datasets(cfg=cfg, cli_args=cli_args)
 
         train(cfg=cfg, cli_args=cli_args, dataset_meta=dataset_meta)
-        assert (Path(temp_dir) / "adapter_model.bin").exists()
+        check_model_output_exists(temp_dir, cfg)
 
     @with_temp_dir
     def test_ft(self, temp_dir):
@@ -109,4 +108,4 @@ class TestFalconPatched(unittest.TestCase):
         dataset_meta = load_datasets(cfg=cfg, cli_args=cli_args)
 
         train(cfg=cfg, cli_args=cli_args, dataset_meta=dataset_meta)
-        assert (Path(temp_dir) / "pytorch_model.bin").exists()
+        check_model_output_exists(temp_dir, cfg)

tests/e2e/patched/test_fused_llama.py

@@ -5,7 +5,6 @@ E2E tests for lora llama
 import logging
 import os
 import unittest
-from pathlib import Path
 
 import pytest
 from transformers.utils import is_torch_bf16_gpu_available
@@ -16,7 +15,7 @@ from axolotl.train import train
 from axolotl.utils.config import normalize_config
 from axolotl.utils.dict import DictDefault
 
-from ..utils import with_temp_dir
+from ..utils import check_model_output_exists, with_temp_dir
 
 LOG = logging.getLogger("axolotl.tests.e2e")
 os.environ["WANDB_DISABLED"] = "true"
@@ -73,4 +72,4 @@ class TestFusedLlama(unittest.TestCase):
         dataset_meta = load_datasets(cfg=cfg, cli_args=cli_args)
 
         train(cfg=cfg, cli_args=cli_args, dataset_meta=dataset_meta)
-        assert (Path(temp_dir) / "pytorch_model.bin").exists()
+        check_model_output_exists(temp_dir, cfg)

tests/e2e/patched/test_llama_s2_attention.py

@@ -5,7 +5,6 @@ E2E tests for llama w/ S2 attn
 import logging
 import os
 import unittest
-from pathlib import Path
 
 import pytest
 
@@ -15,7 +14,7 @@ from axolotl.train import train
 from axolotl.utils.config import normalize_config
 from axolotl.utils.dict import DictDefault
 
-from ..utils import with_temp_dir
+from ..utils import check_model_output_exists, with_temp_dir
 
 LOG = logging.getLogger("axolotl.tests.e2e")
 os.environ["WANDB_DISABLED"] = "true"
@@ -71,7 +70,7 @@ class TestLlamaShiftedSparseAttention(unittest.TestCase):
         dataset_meta = load_datasets(cfg=cfg, cli_args=cli_args)
 
         train(cfg=cfg, cli_args=cli_args, dataset_meta=dataset_meta)
-        assert (Path(temp_dir) / "adapter_model.bin").exists()
+        check_model_output_exists(temp_dir, cfg)
 
     @with_temp_dir
     def test_fft_s2_attn(self, temp_dir):
@@ -111,4 +110,4 @@ class TestLlamaShiftedSparseAttention(unittest.TestCase):
         dataset_meta = load_datasets(cfg=cfg, cli_args=cli_args)
 
         train(cfg=cfg, cli_args=cli_args, dataset_meta=dataset_meta)
-        assert (Path(temp_dir) / "pytorch_model.bin").exists()
+        check_model_output_exists(temp_dir, cfg)

tests/e2e/patched/test_lora_llama_multipack.py

@@ -5,7 +5,6 @@ E2E tests for lora llama
 import logging
 import os
 import unittest
-from pathlib import Path
 
 import pytest
 from transformers.utils import is_auto_gptq_available, is_torch_bf16_gpu_available
@@ -16,7 +15,7 @@ from axolotl.train import train
 from axolotl.utils.config import normalize_config
 from axolotl.utils.dict import DictDefault
 
-from ..utils import with_temp_dir
+from ..utils import check_model_output_exists, with_temp_dir
 
 LOG = logging.getLogger("axolotl.tests.e2e")
 os.environ["WANDB_DISABLED"] = "true"
@@ -76,7 +75,7 @@ class TestLoraLlama(unittest.TestCase):
         dataset_meta = load_datasets(cfg=cfg, cli_args=cli_args)
 
         train(cfg=cfg, cli_args=cli_args, dataset_meta=dataset_meta)
-        assert (Path(temp_dir) / "adapter_model.bin").exists()
+        check_model_output_exists(temp_dir, cfg)
 
     @pytest.mark.skipif(not is_auto_gptq_available(), reason="auto-gptq not available")
     @with_temp_dir
@@ -126,4 +125,4 @@ class TestLoraLlama(unittest.TestCase):
         dataset_meta = load_datasets(cfg=cfg, cli_args=cli_args)
 
         train(cfg=cfg, cli_args=cli_args, dataset_meta=dataset_meta)
-        assert (Path(temp_dir) / "adapter_model.bin").exists()
+        check_model_output_exists(temp_dir, cfg)

tests/e2e/patched/test_mistral_samplepack.py

@@ -5,7 +5,6 @@ E2E tests for lora llama
 import logging
 import os
 import unittest
-from pathlib import Path
 
 from axolotl.cli import load_datasets
 from axolotl.common.cli import TrainerCliArgs
@@ -13,7 +12,7 @@ from axolotl.train import train
 from axolotl.utils.config import normalize_config
 from axolotl.utils.dict import DictDefault
 
-from ..utils import with_temp_dir
+from ..utils import check_model_output_exists, with_temp_dir
 
 LOG = logging.getLogger("axolotl.tests.e2e")
 os.environ["WANDB_DISABLED"] = "true"
@@ -69,7 +68,7 @@ class TestMistral(unittest.TestCase):
         dataset_meta = load_datasets(cfg=cfg, cli_args=cli_args)
 
         train(cfg=cfg, cli_args=cli_args, dataset_meta=dataset_meta)
-        assert (Path(temp_dir) / "adapter_model.bin").exists()
+        check_model_output_exists(temp_dir, cfg)
 
     @with_temp_dir
     def test_ft_packing(self, temp_dir):
@@ -110,4 +109,4 @@ class TestMistral(unittest.TestCase):
         dataset_meta = load_datasets(cfg=cfg, cli_args=cli_args)
 
         train(cfg=cfg, cli_args=cli_args, dataset_meta=dataset_meta)
-        assert (Path(temp_dir) / "pytorch_model.bin").exists()
+        check_model_output_exists(temp_dir, cfg)

tests/e2e/patched/test_mixtral_samplepack.py

@@ -5,7 +5,6 @@ E2E tests for mixtral
 import logging
 import os
 import unittest
-from pathlib import Path
 
 from axolotl.cli import load_datasets
 from axolotl.common.cli import TrainerCliArgs
@@ -13,7 +12,7 @@ from axolotl.train import train
 from axolotl.utils.config import normalize_config
 from axolotl.utils.dict import DictDefault
 
-from ..utils import with_temp_dir
+from ..utils import check_model_output_exists, with_temp_dir
 
 LOG = logging.getLogger("axolotl.tests.e2e")
 os.environ["WANDB_DISABLED"] = "true"
@@ -66,7 +65,7 @@ class TestMixtral(unittest.TestCase):
         dataset_meta = load_datasets(cfg=cfg, cli_args=cli_args)
 
         train(cfg=cfg, cli_args=cli_args, dataset_meta=dataset_meta)
-        assert (Path(temp_dir) / "adapter_model.bin").exists()
+        check_model_output_exists(temp_dir, cfg)
 
     @with_temp_dir
     def test_ft(self, temp_dir):
@@ -108,4 +107,4 @@ class TestMixtral(unittest.TestCase):
             "MixtralFlashAttention2"
             in model.model.layers[0].self_attn.__class__.__name__
         )
-        assert (Path(temp_dir) / "pytorch_model.bin").exists()
+        check_model_output_exists(temp_dir, cfg)

tests/e2e/patched/test_phi_multipack.py

@@ -5,7 +5,6 @@ E2E tests for lora llama
 import logging
 import os
 import unittest
-from pathlib import Path
 
 from axolotl.cli import load_datasets
 from axolotl.common.cli import TrainerCliArgs
@@ -13,7 +12,7 @@ from axolotl.train import train
 from axolotl.utils.config import normalize_config
 from axolotl.utils.dict import DictDefault
 
-from ..utils import with_temp_dir
+from ..utils import check_model_output_exists, with_temp_dir
 
 LOG = logging.getLogger("axolotl.tests.e2e")
 os.environ["WANDB_DISABLED"] = "true"
@@ -69,7 +68,7 @@ class TestPhiMultipack(unittest.TestCase):
         dataset_meta = load_datasets(cfg=cfg, cli_args=cli_args)
 
         train(cfg=cfg, cli_args=cli_args, dataset_meta=dataset_meta)
-        assert (Path(temp_dir) / "pytorch_model.bin").exists()
+        check_model_output_exists(temp_dir, cfg)
 
     @with_temp_dir
     def test_qlora_packed(self, temp_dir):
@@ -120,4 +119,4 @@ class TestPhiMultipack(unittest.TestCase):
         dataset_meta = load_datasets(cfg=cfg, cli_args=cli_args)
 
         train(cfg=cfg, cli_args=cli_args, dataset_meta=dataset_meta)
-        assert (Path(temp_dir) / "adapter_model.bin").exists()
+        check_model_output_exists(temp_dir, cfg)

tests/e2e/patched/test_resume.py

@@ -6,7 +6,6 @@ import logging
 import os
 import re
 import subprocess
-from pathlib import Path
 
 from transformers.utils import is_torch_bf16_gpu_available
 
@@ -16,7 +15,7 @@ from axolotl.train import train
 from axolotl.utils.config import normalize_config
 from axolotl.utils.dict import DictDefault
 
-from ..utils import most_recent_subdir
+from ..utils import check_model_output_exists, most_recent_subdir
 
 LOG = logging.getLogger("axolotl.tests.e2e")
 os.environ["WANDB_DISABLED"] = "true"
@@ -83,7 +82,7 @@ class TestResumeLlama:
         cli_args = TrainerCliArgs()
 
         train(cfg=resume_cfg, cli_args=cli_args, dataset_meta=dataset_meta)
-        assert (Path(temp_dir) / "adapter_model.bin").exists()
+        check_model_output_exists(temp_dir, cfg)
 
         tb_log_path_1 = most_recent_subdir(temp_dir + "/runs")
         cmd = f"tensorboard --inspect  --logdir {tb_log_path_1}"

tests/e2e/patched/test_unsloth_qlora.py

@@ -3,7 +3,6 @@ e2e tests for unsloth qlora
 """
 import logging
 import os
-from pathlib import Path
 
 import pytest
 
@@ -13,7 +12,7 @@ from axolotl.train import train
 from axolotl.utils.config import normalize_config
 from axolotl.utils.dict import DictDefault
 
-from ..utils import check_tensorboard
+from ..utils import check_model_output_exists, check_tensorboard
 
 LOG = logging.getLogger("axolotl.tests.e2e")
 os.environ["WANDB_DISABLED"] = "true"
@@ -77,7 +76,7 @@ class TestUnslothQLoRA:
         dataset_meta = load_datasets(cfg=cfg, cli_args=cli_args)
 
         train(cfg=cfg, cli_args=cli_args, dataset_meta=dataset_meta)
-        assert (Path(temp_dir) / "adapter_model.bin").exists()
+        check_model_output_exists(temp_dir, cfg)
 
         check_tensorboard(
             temp_dir + "/runs", "train/train_loss", 2.0, "Train Loss is too high"
@@ -127,7 +126,7 @@ class TestUnslothQLoRA:
         dataset_meta = load_datasets(cfg=cfg, cli_args=cli_args)
 
         train(cfg=cfg, cli_args=cli_args, dataset_meta=dataset_meta)
-        assert (Path(temp_dir) / "adapter_model.bin").exists()
+        check_model_output_exists(temp_dir, cfg)
 
         check_tensorboard(
             temp_dir + "/runs", "train/train_loss", 2.0, "Train Loss is too high"
@@ -182,7 +181,7 @@ class TestUnslothQLoRA:
         dataset_meta = load_datasets(cfg=cfg, cli_args=cli_args)
 
         train(cfg=cfg, cli_args=cli_args, dataset_meta=dataset_meta)
-        assert (Path(temp_dir) / "adapter_model.bin").exists()
+        check_model_output_exists(temp_dir, cfg)
 
         check_tensorboard(
             temp_dir + "/runs", "train/train_loss", 2.0, "Train Loss is too high"