Changes / Feeds / meta-llama/Llama-3.1-405B-FP8
meta-llama/Llama-3.1-405B-FP8
6 changes to a tracked key since 2024-07-20, newest first. Head as last read: cb4739810a51170037906b5daa2bb78ed13cd43c.
| Date | Advisory | What changed, and what it does |
|---|---|---|
| 2024-12-06 | x_NXMC-2024-0149 | pad token unset to "<|finetune_right_pad_id|>"; batched training masks a different id (2024-12-06).d79e67b5ee26 -> cb4739810a51 |
| 2024-08-14 | x_NXMC-2024-0148 breaking | KV heads 16 to 8; derived parameters 410.08B to 405.85B (-4.23B, -8.46 GB of bf16 weights) (2024-08-14).60db52617fa6 -> d79e67b5ee26 |
| 2024-07-24 | x_NXMC-2024-0147 breaking | quantization_config changed; the stored format of the weights is different (2024-07-24).b46ddc84f9bb -> 60db52617fa6 |
| 2024-07-23 | x_NXMC-2024-0146 breaking | rope_scaling added ({"factor":8,"high_freq_factor":4,"low_freq_factor":1,"ori...); positions are now scaled (2024-07-23).f2528c44b8a3 -> b46ddc84f9bb |
| 2024-07-22 | x_NXMC-2024-0145 | generation defaults changed (do_sample unset to true, temperature unset to 0.6, top_p unset to 0.9); a call that does not pass these explicitly now generates differently (2024-07-22).1965e1c19271 -> f2528c44b8a3 |
| 2024-07-21 | x_NXMC-2024-0144 | declared dtype unset to "bfloat16" (2024-07-21).e981b1a79707 -> 79d637ba1eea |
Pin it, and let a bot move the pin
from_pretrained("meta-llama/Llama-3.1-405B-FP8", revision="cb4739810a51170037906b5daa2bb78ed13cd43c")With the Renovate preset, a pull request moves this pin when the head moves, and links back here.
[](https://www.neurarch.com/changes-feeds/models/meta-llama/Llama-3.1-405B-FP8.html)