afm-crag-movie-sft-v2
This model is a fine-tuned version of Qwen/Qwen3-8B on the afm_movie_all_trajectories_v2_include_max_turns_train dataset. It achieves the following results on the evaluation set:
- Loss: 0.0388
Model description
More information needed
Intended uses & limitations
More information needed
Training and evaluation data
More information needed
Training procedure
Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 1e-05
- train_batch_size: 1
- eval_batch_size: 1
- seed: 42
- distributed_type: multi-GPU
- num_devices: 3
- gradient_accumulation_steps: 10
- total_train_batch_size: 30
- total_eval_batch_size: 3
- optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
- lr_scheduler_type: cosine
- lr_scheduler_warmup_ratio: 0.1
- num_epochs: 2.0
Training results
| Training Loss | Epoch | Step | Validation Loss |
|---|---|---|---|
| 2.1968 | 0.0395 | 10 | 1.2296 |
| 0.5746 | 0.0789 | 20 | 0.1444 |
| 0.1503 | 0.1184 | 30 | 0.0836 |
| 0.1081 | 0.1579 | 40 | 0.0652 |
| 0.0813 | 0.1973 | 50 | 0.0631 |
| 0.0785 | 0.2368 | 60 | 0.0618 |
| 0.0915 | 0.2762 | 70 | 0.0662 |
| 0.0803 | 0.3157 | 80 | 0.0569 |
| 0.0633 | 0.3552 | 90 | 0.0569 |
| 0.0913 | 0.3946 | 100 | 0.0559 |
| 0.0746 | 0.4341 | 110 | 0.0529 |
| 0.084 | 0.4736 | 120 | 0.0518 |
| 0.0607 | 0.5130 | 130 | 0.0509 |
| 0.0654 | 0.5525 | 140 | 0.0521 |
| 0.059 | 0.5919 | 150 | 0.0496 |
| 0.0558 | 0.6314 | 160 | 0.0487 |
| 0.0697 | 0.6709 | 170 | 0.0476 |
| 0.0564 | 0.7103 | 180 | 0.0467 |
| 0.0595 | 0.7498 | 190 | 0.0474 |
| 0.0507 | 0.7893 | 200 | 0.0465 |
| 0.0758 | 0.8287 | 210 | 0.0463 |
| 0.0603 | 0.8682 | 220 | 0.0458 |
| 0.0495 | 0.9077 | 230 | 0.0444 |
| 0.0531 | 0.9471 | 240 | 0.0441 |
| 0.0541 | 0.9866 | 250 | 0.0446 |
| 0.0419 | 1.0237 | 260 | 0.0443 |
| 0.0415 | 1.0631 | 270 | 0.0435 |
| 0.0518 | 1.1026 | 280 | 0.0435 |
| 0.034 | 1.1421 | 290 | 0.0432 |
| 0.0247 | 1.1815 | 300 | 0.0430 |
| 0.0408 | 1.2210 | 310 | 0.0423 |
| 0.0289 | 1.2605 | 320 | 0.0416 |
| 0.0413 | 1.2999 | 330 | 0.0422 |
| 0.0376 | 1.3394 | 340 | 0.0417 |
| 0.0323 | 1.3788 | 350 | 0.0408 |
| 0.0311 | 1.4183 | 360 | 0.0402 |
| 0.036 | 1.4578 | 370 | 0.0400 |
| 0.0312 | 1.4972 | 380 | 0.0402 |
| 0.0332 | 1.5367 | 390 | 0.0407 |
| 0.0329 | 1.5762 | 400 | 0.0404 |
| 0.0476 | 1.6156 | 410 | 0.0399 |
| 0.0268 | 1.6551 | 420 | 0.0395 |
| 0.0316 | 1.6946 | 430 | 0.0395 |
| 0.0313 | 1.7340 | 440 | 0.0393 |
| 0.0339 | 1.7735 | 450 | 0.0391 |
| 0.0293 | 1.8129 | 460 | 0.0389 |
| 0.0412 | 1.8524 | 470 | 0.0388 |
| 0.0472 | 1.8919 | 480 | 0.0388 |
| 0.0274 | 1.9313 | 490 | 0.0388 |
| 0.0402 | 1.9708 | 500 | 0.0387 |
Framework versions
- Transformers 4.51.3
- Pytorch 2.10.0+cu128
- Datasets 3.5.0
- Tokenizers 0.21.1