Skip to content

Conversation

@jamesruio
Copy link
Contributor

No description provided.

Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

这个文件在perf里面需要改动吗?能否直接使用FlagScale仓库的pretrain_llama

chip_tps = whole_tps / (args.nproc_per_node * args.nnodes)
print("System tokens per second: ", whole_tps)
print("Tokens/p/s: ", chip_tps)
print("MFU: ", chip_tps * 7000000000.0 * 6 / theoryflops) No newline at end of file
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

这个MFU计算的正确吗?没数错的话这个是7B吧


# Hosts to run the benchmark. Each item is an IP address or a hostname.
HOSTS = ["10.1.2.2", "10.1.2.3", "10.1.2.4"]
HOSTS = ["192.2.32.13", "192.2.32.14", "192.2.32.2", "192.2.32.4"]
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

这个不需要更改

@@ -0,0 +1,5 @@
#!/bin/bash
# using github mirrors to avoid github TTL
git clone https://githubfast.com/FlagOpen/FlagScale
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

这里需要指定一下版本,可以指定commit

@upvenly upvenly merged commit ebe6776 into flagos-ai:main Jan 15, 2024
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants