zhangyiss/mlx-examples
1
0
Fork 0
mirror of https://github.com/ml-explore/mlx-examples.git synced 2025-07-19 00:41:12 +08:00
Code Issues Actions Packages Projects Releases Wiki Activity

nice metric printing in testing

Browse Source
This commit is contained in:
Goekdeniz-Guelmez 2025-02-04 11:21:52 +01:00
parent 43940ec673
commit 56712664f6
1 changed files with 5 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
llms/mlx_lm/lora.py
View File

@ -287,7 +287,7 @@ def evaluate_model(args, model: nn.Module, tokenizer: TokenizerWrapper, test_set
model.eval() model.eval()
if args.training_mode == "orpo": if args.training_mode == "orpo":
test_loss, test_rewards, _, _ = evaluate_orpo( test_loss, test_rewards, _, test_metrics = evaluate_orpo(
model=model, model=model,
dataset=test_set, dataset=test_set,
batch_size=args.batch_size, batch_size=args.batch_size,
@ -297,6 +297,10 @@ def evaluate_model(args, model: nn.Module, tokenizer: TokenizerWrapper, test_set
) )
test_ppl = math.exp(test_loss) test_ppl = math.exp(test_loss)
print(f"Test loss {test_loss:.3f}, Test ppl {test_ppl:.3f}, Rewards: {test_rewards[0]:.3f}, {test_rewards[1]:.3f}") print(f"Test loss {test_loss:.3f}, Test ppl {test_ppl:.3f}, Rewards: {test_rewards[0]:.3f}, {test_rewards[1]:.3f}")
print("ORPO Test Metrics:")
for metric_name, metric_value in test_metrics.items():
print(f" {metric_name}: {float(metric_value):.3f}")
else: else:
test_loss = evaluate( test_loss = evaluate(
model=model, model=model,