support for fp8

src/axolotl/core/trainer_builder.py

@@ -483,6 +483,10 @@ def build(self, total_num_steps):
         training_arguments_kwargs["fp16"] = (
             self.cfg.fp16 and not self.cfg.bf16
         ) or False
+        if self.cfg.fp8:
+            training_arguments_kwargs["fp16"] = False
+            training_arguments_kwargs["bf16"] = False
+
         training_arguments_kwargs["tf32"] = self.cfg.tf32
         training_arguments_kwargs["warmup_steps"] = warmup_steps
         training_arguments_kwargs["logging_steps"] = logging_steps

src/axolotl/utils/config.py

@@ -70,7 +70,9 @@ def normalize_config(cfg):
     else:
         torch.backends.cuda.matmul.allow_tf32 = cfg.tf32 or False
 
-    if cfg.bf16 or cfg.bfloat16:
+    if cfg.fp8:
+        cfg.torch_dtype = torch.bfloat16
+    elif cfg.bf16 or cfg.bfloat16:
         cfg.torch_dtype = torch.bfloat16
     elif cfg.load_in_8bit or cfg.fp16 or cfg.float16:
         cfg.torch_dtype = torch.float16

src/axolotl/utils/trainer.py

@@ -268,6 +268,8 @@ def setup_trainer(cfg, train_dataset, eval_dataset, model, tokenizer, total_num_
         setup_fsdp_envs(cfg)
     elif cfg.deepspeed:
         os.environ["ACCELERATE_USE_DEEPSPEED"] = "true"
+    if cfg.fp8:
+        os.environ["ACCELERATE_MIXED_PRECISION"] = "fp8"
 
     trainer_builder = HFCausalTrainerBuilder(cfg, model, tokenizer)
     trainer_builder.train_dataset = train_dataset