 justheuristic
|
62d7fde8af
pre-check type
|
3 rokov pred |
|
justheuristic
|
7fba411dff
extended run_serverexample
|
3 rokov pred |
|
justheuristic
|
20497f81d1
switch to hivemind-master
|
3 rokov pred |
|
justheuristic
|
5a15c13ca7
switch to hivemind-master
|
3 rokov pred |
 Pavel Samygin
|
57f4e0a899
add identity path
|
3 rokov pred |
|
justheuristic
|
35310698f0
newer hivemind version
|
3 rokov pred |
|
justheuristic
|
8959727dea
add minimalistic benchmarks
|
3 rokov pred |
|
justheuristic
|
a798ea04a6
add minimalistic benchmarks
|
3 rokov pred |
|
justheuristic
|
e3a7d5af30
inference mode
|
3 rokov pred |
|
justheuristic
|
3b16d6ffdb
black-isort
|
3 rokov pred |
|
justheuristic
|
a28ea0aa6f
Merge pull request #9 from learning-at-home/rpc
|
3 rokov pred |
|
justheuristic
|
14e316b52a
black-isort
|
3 rokov pred |
|
justheuristic
|
eea9287182
RemoteTransformerBlock
|
3 rokov pred |
|
justheuristic
|
3e9fd63a02
RemoteTransformerBlock
|
3 rokov pred |
|
justheuristic
|
1cca611c9f
undo rename
|
3 rokov pred |
|
justheuristic
|
7ce7cd7a97
basic backend
|
3 rokov pred |
|
justheuristic
|
1c49bcb741
basic backend
|
3 rokov pred |
|
justheuristic
|
3215945882
use logger
|
3 rokov pred |
|
justheuristic
|
ce5dedd2c7
rename
|
3 rokov pred |
|
justheuristic
|
8f5d022d18
save hidden size
|
3 rokov pred |
|
justheuristic
|
e2e9d0e94c
memory cache for attention KVs
|
3 rokov pred |
|
justheuristic
|
e5e8c9ed12
expel all bloom-specific files to src.bloom
|
3 rokov pred |
|
justheuristic
|
324ea2dc96
save non-transformer params separately
|
3 rokov pred |
|
justheuristic
|
54cf292374
backend schema
|
3 rokov pred |
|
justheuristic
|
3ccd0b5e2d
account for layer_past in alibi
|
3 rokov pred |
|
justheuristic
|
fb3bfbb78f
optimization TODOs for later
|
3 rokov pred |
|
justheuristic
|
902bf6400a
[temp workaround] create alibi
|
3 rokov pred |
|
justheuristic
|
05faa0b3c8
add quantization script for cpu
|
3 rokov pred |
|
justheuristic
|
ffb56a65ed
keep hidden_size as property
|
3 rokov pred |
|
justheuristic
|
ffce7718ac
layer.py -> block.py
|
3 rokov pred |
