add e2e for flattening

axolotl-ai-cloud · Dec 17, 2024 · 503ead8 · 503ead8
1 parent add3a94
commit 503ead8
Showing 1 changed file with 42 additions and 6 deletions.
diff --git a/tests/e2e/test_llama.py b/tests/e2e/test_llama.py
@@ -4,7 +4,6 @@
 
 import logging
 import os
-import unittest
 from pathlib import Path
 
 from axolotl.cli import load_datasets
@@ -13,18 +12,15 @@
 from axolotl.utils.config import normalize_config
 from axolotl.utils.dict import DictDefault
 
-from .utils import with_temp_dir
-
 LOG = logging.getLogger("axolotl.tests.e2e")
 os.environ["WANDB_DISABLED"] = "true"
 
 
-class TestLlama(unittest.TestCase):
+class TestLlama:
     """
     Test case for Llama models
     """
 
-    @with_temp_dir
     def test_fft_trust_remote_code(self, temp_dir):
         # pylint: disable=duplicate-code
         cfg = DictDefault(
@@ -46,7 +42,8 @@ def test_fft_trust_remote_code(self, temp_dir):
                     },
                 ],
                 "num_epochs": 1,
-                "micro_batch_size": 8,
+                "max_steps": 5,
+                "micro_batch_size": 2,
                 "gradient_accumulation_steps": 1,
                 "output_dir": temp_dir,
                 "learning_rate": 0.00001,
@@ -64,3 +61,42 @@ def test_fft_trust_remote_code(self, temp_dir):
 
         train(cfg=cfg, cli_args=cli_args, dataset_meta=dataset_meta)
         assert (Path(temp_dir) / "model.safetensors").exists()
+
+    def test_batch_flattening(self, temp_dir):
+        # pylint: disable=duplicate-code
+        cfg = DictDefault(
+            {
+                "base_model": "HuggingFaceTB/SmolLM2-135M",
+                "trust_remote_code": True,
+                "sequence_len": 512,
+                "val_set_size": 0.01,
+                "special_tokens": {
+                    "pad_token": "<|endoftext|>",
+                },
+                "datasets": [
+                    {
+                        "path": "mhenrichsen/alpaca_2k_test",
+                        "type": "alpaca",
+                    },
+                ],
+                "num_epochs": 1,
+                "max_steps": 5,
+                "micro_batch_size": 4,
+                "gradient_accumulation_steps": 1,
+                "output_dir": temp_dir,
+                "learning_rate": 0.00001,
+                "optimizer": "adamw_8bit",
+                "lr_scheduler": "cosine",
+                "flash_attention": True,
+                "sample_packing": False,
+                "batch_flattening": True,
+                "bf16": True,
+                "save_safetensors": True,
+            }
+        )
+        normalize_config(cfg)
+        cli_args = TrainerCliArgs()
+        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) / "model.safetensors").exists()