Add CAME Optimizer (axolotl-ai-cloud#2385)

Author: xzuyn

docs/config.qmd

@@ -612,6 +612,7 @@ lr_div_factor: # Learning rate div factor
 # - optimi_adamw
 # - ao_adamw_8bit
 # - ao_adamw_fp8
+# - came_pytorch
 optimizer:
 # Dictionary of arguments to pass to the optimizer
 optim_args:

setup.py

@@ -142,6 +142,7 @@ def get_package_version():
         "apollo-torch",
         "lomo-optim==0.1.1",
         "torch-optimi==0.2.1",
+        "came_pytorch==0.1.3",
     ],
     "ray": [
         "ray[train]",

src/axolotl/core/trainer_builder.py

@@ -708,6 +708,20 @@ def build(self, total_num_steps):
                 optimizer_cls = ADOPT
                 adam_kwargs["decouple"] = True
                 optimizer_kwargs.update(adam_kwargs)
+            elif self.cfg.optimizer == "came_pytorch":
+                from came_pytorch import CAME
+
+                optimizer_cls = CAME
+
+                beta1 = training_arguments_kwargs.get("adam_beta1", 0.9)
+                beta2 = training_arguments_kwargs.get("adam_beta2", 0.999)
+                beta3 = training_arguments_kwargs.get("adam_beta2", 0.9999)
+                eps1 = training_arguments_kwargs.get("adam_epsilon", 1e-30)
+                eps2 = training_arguments_kwargs.get("adam_epsilon2", 1e-16)
+                adam_kwargs["betas"] = (beta1, beta2, beta3)
+                adam_kwargs["eps"] = (eps1, eps2)
+
+                optimizer_kwargs.update(adam_kwargs)
 
             # Parse any additional optimizer args from config
             if self.cfg.optim_args:

src/axolotl/utils/schemas/enums.py

@@ -53,4 +53,5 @@ class CustomSupportedOptimizers(str, Enum):
     ao_adamw_8bit = "ao_adamw_8bit"  # pylint: disable=invalid-name
     ao_adamw_fp8 = "ao_adamw_fp8"  # pylint: disable=invalid-name
     adopt_adamw = "adopt_adamw"  # pylint: disable=invalid-name
+    came_pytorch = "came_pytorch"  # pylint: disable=invalid-name
     muon = "muon"  # pylint: disable=invalid-name

src/axolotl/utils/schemas/training.py

@@ -75,8 +75,10 @@ class HyperparametersConfig(BaseModel):
     lr_groups: list[LrGroup] | None = None
 
     adam_epsilon: float | None = None
+    adam_epsilon2: float | None = None
     adam_beta1: float | None = None
     adam_beta2: float | None = None
+    adam_beta3: float | None = None
     max_grad_norm: float | None = None
     num_epochs: float = Field(default=1.0)
 

tests/e2e/test_optimizers.py

@@ -199,3 +199,50 @@ def test_fft_schedule_free_adamw(self, temp_dir):
 
         train(cfg=cfg, dataset_meta=dataset_meta)
         check_model_output_exists(temp_dir, cfg)
+
+    @with_temp_dir
+    def test_came_pytorch(self, temp_dir):
+        # pylint: disable=duplicate-code
+        cfg = DictDefault(
+            {
+                "base_model": "JackFram/llama-68m",
+                "tokenizer_type": "LlamaTokenizer",
+                "sequence_len": 1024,
+                "load_in_8bit": True,
+                "adapter": "lora",
+                "lora_r": 8,
+                "lora_alpha": 16,
+                "lora_dropout": 0.05,
+                "lora_target_linear": True,
+                "val_set_size": 0.1,
+                "special_tokens": {
+                    "unk_token": "<unk>",
+                    "bos_token": "<s>",
+                    "eos_token": "</s>",
+                },
+                "datasets": [
+                    {
+                        "path": "mhenrichsen/alpaca_2k_test",
+                        "type": "alpaca",
+                    },
+                ],
+                "num_epochs": 1,
+                "micro_batch_size": 8,
+                "gradient_accumulation_steps": 1,
+                "output_dir": temp_dir,
+                "learning_rate": 0.00001,
+                "optimizer": "came_pytorch",
+                "adam_beta3": 0.9999,
+                "adam_epsilon2": 1e-16,
+                "max_steps": 5,
+                "lr_scheduler": "cosine",
+            }
+        )
+
+        cfg = validate_config(cfg)
+        normalize_config(cfg)
+        cli_args = TrainerCliArgs()
+        dataset_meta = load_datasets(cfg=cfg, cli_args=cli_args)
+
+        train(cfg=cfg, dataset_meta=dataset_meta)
+        check_model_output_exists(temp_dir, cfg)