AI-Hypercomputer / AI-Hypercomputer/maxtext
DEFAULT_MASK_VALUE causes gradient explosion and nan loss on deep models
- 主要语言
- Python
- 星标
- 2.4k
- 派生
- 607
- 平均合并
- 2 天 19 小时
- 30 天内合并 PR
- 158
描述
I was training a llama model on GPU, with a custom embedding. It worked fine with 12 layers, dim 1024, seq length 256, but loss would become nan after the first step if setting num_layers to more than 17. I debugged the gradients, and found after each layer their magnitude would increase by around 100x, until they hit float32_max at around the 18th layer and became inf, leading to nan loss.
The gradient explosion seemed to be coming from
`local_exps = jnp.exp(attn_weights - local_max)`
in attentions.py.
Changing
`DEFAULT_MASK_VALUE = -0.7 * float(jnp.finfo(jnp.dtype("float32")).max)`
to
`DEFAULT_MASK_VALUE = -jnp.inf`
fixed the issue, and the gradients' magnitude stopped increasing after each level.
Presumably the issue wasn't noticed during TPU training as that uses a separate codepath.
贡献指南
评估
这个 Issue 还没有评估数据。