Commit
·
1046796
1
Parent(s):
fc20a02
Update README.md
Browse files
README.md
CHANGED
|
@@ -1,3 +1,10 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
# miniG
|
| 2 |
|
| 3 |
A model trained on a synthesis dataset of over 120 million entries, this dataset having been generated through the application of state-of-the-art language models utilizing large context windows, alongside methodologies akin to retrieval-augmented generation and knowledge graph integration, where the data synthesis is conducted within clusters derived from a curated pretraining corpus of 20 billion tokens, with subsequent validation performed by the model itself.
|
|
@@ -24,4 +31,4 @@ Disclaimer: Please note that the model was trained on unfiltered internet data.
|
|
| 24 |
|
| 25 |
推理参数:我们的观察表明,如果想要减少幻觉结果,建议使用top_p=0.8的采样方式,然后设置temperature为0.3,或者使用纯粹的temperature采样,设置为0.2。总体来说,相比类似的模型,该模型需要较低的temperature,我们暂时将其归因于在庞大数据集上的过拟合。
|
| 26 |
|
| 27 |
-
免责声明:请注意,该模型是在未经过滤的互联网数据上训练的。由于我们无法对所有数据进行筛选,仍有可能存在大量不适当的内容——包括从露骨的材料到暴力和攻击性语言的内容——我们无法移除。因此,您必须自行对模型进行安全检查,并在输出中实施关键词过滤。由于计算资源的限制,我们目前无法为伦理和安全考虑进行人类反馈的强化学习(RLHF),也不能对SFT样本进行限制性微调,以限制模型回答某些问题的能力。
|
|
|
|
| 1 |
+
---
|
| 2 |
+
language:
|
| 3 |
+
- en
|
| 4 |
+
- zh
|
| 5 |
+
- ja
|
| 6 |
+
- de
|
| 7 |
+
---
|
| 8 |
# miniG
|
| 9 |
|
| 10 |
A model trained on a synthesis dataset of over 120 million entries, this dataset having been generated through the application of state-of-the-art language models utilizing large context windows, alongside methodologies akin to retrieval-augmented generation and knowledge graph integration, where the data synthesis is conducted within clusters derived from a curated pretraining corpus of 20 billion tokens, with subsequent validation performed by the model itself.
|
|
|
|
| 31 |
|
| 32 |
推理参数:我们的观察表明,如果想要减少幻觉结果,建议使用top_p=0.8的采样方式,然后设置temperature为0.3,或者使用纯粹的temperature采样,设置为0.2。总体来说,相比类似的模型,该模型需要较低的temperature,我们暂时将其归因于在庞大数据集上的过拟合。
|
| 33 |
|
| 34 |
+
免责声明:请注意,该模型是在未经过滤的互联网数据上训练的。由于我们无法对所有数据进行筛选,仍有可能存在大量不适当的内容——包括从露骨的材料到暴力和攻击性语言的内容——我们无法移除。因此,您必须自行对模型进行安全检查,并在输出中实施关键词过滤。由于计算资源的限制,我们目前无法为伦理和安全考虑进行人类反馈的强化学习(RLHF),也不能对SFT样本进行限制性微调,以限制模型回答某些问题的能力。
|