allow the optimizer prune ratio for ReLoRA to be configurable (#1287)

* allow the optimizer prune ration for relora to be configurable

* update docs for relora

* prevent circular imports

README.md

@@ -734,6 +734,8 @@ peft:
 # Must use either 'lora' or 'qlora' adapter, and does not support fsdp or deepspeed
 relora_steps: # Number of steps per ReLoRA restart
 relora_warmup_steps: # Number of per-restart warmup steps
+relora_anneal_steps: # Number of anneal steps for each relora cycle
+relora_prune_ratio: # threshold for optimizer magnitude when pruning
 relora_cpu_offload: # True to perform lora weight merges on cpu during restarts, for modest gpu memory savings
 
 # wandb configuration if you're using it

src/axolotl/core/trainer_builder.py

@@ -131,6 +131,10 @@ class AxolotlTrainingArguments(TrainingArguments):
         default=None,
         metadata={"help": "how many warmup steps to take after reset for ReLoRA"},
     )
+    relora_prune_ratio: Optional[float] = field(
+        default=0.9,
+        metadata={"help": "prune ratio for magnitude pruning of the optimizer"},
+    )
     bench_split: Optional[str] = field(
         default="eval", metadata={"help": "The benchmark split to run on"}
     )
@@ -900,9 +904,20 @@ class HFCausalTrainerBuilder(TrainerBuilderBase):
         training_arguments_kwargs[
             "sample_packing_seq_len_multiplier"
         ] = self.cfg.micro_batch_size
-        training_arguments_kwargs["relora_steps"] = self.cfg.relora_steps
-        training_arguments_kwargs["relora_warmup_steps"] = self.cfg.relora_warmup_steps
-        training_arguments_kwargs["relora_anneal_steps"] = self.cfg.relora_anneal_steps
+        if self.cfg.relora_steps:
+            training_arguments_kwargs["relora_steps"] = self.cfg.relora_steps
+            training_arguments_kwargs[
+                "relora_warmup_steps"
+            ] = self.cfg.relora_warmup_steps
+            if self.cfg.relora_anneal_steps:
+                training_arguments_kwargs[
+                    "relora_anneal_steps"
+                ] = self.cfg.relora_anneal_steps
+            if self.cfg.relora_prune_ratio:
+                training_arguments_kwargs[
+                    "relora_prune_ratio"
+                ] = self.cfg.relora_prune_ratio
+
         training_arguments_kwargs = self.hook_pre_create_training_args(
             training_arguments_kwargs
         )

src/axolotl/monkeypatch/relora.py

@@ -46,8 +46,9 @@ def reset_optimizer(
     *,
     reset_params: list[str],  # where str is the key to a torch.nn.Parameter
     optimizer_state_keys: list[str],
+    prune_ratio: float = 0.9,
 ):
-    pruning_fn = partial(magnitude_pruning_, prune_ratio=0.9)
+    pruning_fn = partial(magnitude_pruning_, prune_ratio=prune_ratio)
     n_zeros = 0
     n_total = 0
 
@@ -159,6 +160,7 @@ class ReLoRACallback(TrainerCallback):
                     optimizer,
                     reset_params=lora_params,
                     optimizer_state_keys=optimizer_state_keys,
+                    prune_ratio=args.relora_prune_ratio,
                 )
 
             if self.quantized: