-
## Decode 128
We are aiming to hit 20 t/s/u end to end for Llama3 decode on t3k. These are the issues left.
![image](https://github.com/user-attachments/assets/ce4a7190-f49d-431b-8f18-79ebb130e36c…
-
- [x] Support measure filters in `metricsview` package
- [x] Deprecate metrics queries in `queries` package
- [x] In `runtime/metricsview`, push unnest handling to dimension expression
- [x] Use `j…
-
Burn graph is super useful, but it'd be even more useful if you could see percentage of a given bar in comparison to clicked bar, currently it just says
net.minecraft.server.MinecraftServer$$Lambda…
-
The Fediverse Instance and the Public Key Directory server could negotiate a shared secret, that only instance administrators possess. Then, every successful BurnDown would require this OTP in additio…
-
This issue tracks the open issues the model team must solve in order to hit Llama2 perf targets.
## Decode 128
We have a new perf target of 20 tok/s at seqlen = 128. This issue lists the problems …
-
We need to find a way to integrate the Burndown Chart into the github ecosystem or use an external tool which would be less ideal.
-
Burndown list for an beta release:
- [x] Fix annoying crosshair/shader interaction - Appears to be an interaction with a certain resource pack
- [ ] Chunk loading order & transparency sorting
- […
-
Track issues which will improve Llama e2e perf
- [ ] (Colman) get all `set_runtime_args` calls below 3 microseconds
## Runtime-specific
`wait_for_tensor_data_populated` in worker threads has …
-
Tracks the open issues for Falcon40b prefill to hit target perf.
Last updated: May 27th
# Prefill
*bfp8*
- Measured May 22nd, main
- 1 GHz
- Perf measurements based on 1 layer perf and ext…
-
### Discussed in https://github.com/GothenburgBitFactory/taskwarrior/discussions/3307
Originally posted by **gour** March 28, 2024
Hello!
While anticipating 3.0 to arrive into my distributi…