初始化项目，由ModelHub XC社区提供模型

Model: jackf857/qwen3-8b-base-beta-dpo-hh-helpful-4xh200-batch-64-20260424-013732 Source: Original Platform
2026-05-10 12:41:10 +08:00
commit d493a9b916
23 changed files with 163691 additions and 0 deletions
--- a/README.md
+++ b/README.md
@@ -0,0 +1,79 @@
+---
+library_name: transformers
+base_model: jackf857/qwen3-8b-base-sft-hh-helpful-4xh200-batch-64-20260417-214452
+tags:
+- alignment-handbook
+- beta-dpo
+- generated_from_trainer
+datasets:
+- Anthropic/hh-rlhf
+model-index:
+- name: qwen3-8b-base-beta-dpo-hh-helpful-4xh200-batch-64-20260424-013732
+  results: []
+---
+
+<!-- This model card has been generated automatically according to the information the Trainer had access to. You
+should probably proofread and complete it, then remove this comment. -->
+
+# qwen3-8b-base-beta-dpo-hh-helpful-4xh200-batch-64-20260424-013732
+
+This model is a fine-tuned version of [jackf857/qwen3-8b-base-sft-hh-helpful-4xh200-batch-64-20260417-214452](https://huggingface.co/jackf857/qwen3-8b-base-sft-hh-helpful-4xh200-batch-64-20260417-214452) on the Anthropic/hh-rlhf dataset.
+It achieves the following results on the evaluation set:
+- Loss: 0.6505
+- Beta Dpo/gap Mean: 25.7183
+- Beta Dpo/gap Std: 26.1829
+- Beta Dpo/beta Used Raw: -0.2537
+- Beta Dpo/beta Used: 0.0212
+- Beta Dpo/mask Keep Frac: 1.0
+- Logits/chosen: -1.6123
+- Logits/rejected: -1.2796
+
+## 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: 5e-07
+- train_batch_size: 8
+- eval_batch_size: 8
+- seed: 42
+- distributed_type: multi-GPU
+- num_devices: 4
+- gradient_accumulation_steps: 2
+- total_train_batch_size: 64
+- total_eval_batch_size: 32
+- optimizer: Use OptimizerNames.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: 1
+
+### Training results
+
+| Training Loss | Epoch  | Step | Validation Loss | Beta Dpo/gap Mean | Beta Dpo/gap Std | Beta Dpo/beta Used Raw | Beta Dpo/beta Used | Beta Dpo/mask Keep Frac | Logits/chosen | Logits/rejected |
+|:-------------:|:------:|:----:|:---------------:|:-----------------:|:----------------:|:----------------------:|:------------------:|:-----------------------:|:-------------:|:---------------:|
+| 1.1246        | 0.1468 | 100  | 0.6499          | 4.6173            | 6.2353           | -0.0000                | 0.0293             | 1.0                     | -0.1879       | 0.0487          |
+| 0.313         | 0.2937 | 200  | 0.6346          | 12.9809           | 14.5981          | -0.0999                | 0.0229             | 1.0                     | -1.1708       | -0.8922         |
+| 0.6858        | 0.4405 | 300  | 0.6556          | 19.2389           | 20.8328          | -0.2038                | 0.0146             | 1.0                     | -1.4133       | -1.1126         |
+| 0.7246        | 0.5874 | 400  | 0.6605          | 25.6196           | 25.0584          | -0.3654                | 0.0094             | 1.0                     | -1.5698       | -1.2495         |
+| 0.9705        | 0.7342 | 500  | 0.6340          | 24.0392           | 26.5263          | -0.1783                | 0.0322             | 1.0                     | -1.7312       | -1.4078         |
+| 0.8177        | 0.8811 | 600  | 0.6505          | 25.7183           | 26.1829          | -0.2537                | 0.0212             | 1.0                     | -1.6123       | -1.2796         |
+
+
+### Framework versions
+
+- Transformers 4.51.0
+- Pytorch 2.3.1+cu121
+- Datasets 2.21.0
+- Tokenizers 0.21.4