Nel Nibcord
|
8f0d19e9b0
embed fix
|
7 maanden geleden |
Nel Nibcord
|
ee97563b45
Dummied up an abstact save_checkpoint
|
7 maanden geleden |
Nel Nibcord
|
02281ebe3d
Missed one
|
7 maanden geleden |
Nel Nibcord
|
dffa17b2d1
Slightly simplified waiting for outstanding requests
|
7 maanden geleden |
Nel Nibcord
|
720940d563
Removed statefulModel stuff from mlx impl too
|
7 maanden geleden |
Nel Nibcord
|
bc2812238f
Removed ensure_session to clean stuff up. May revisit later
|
7 maanden geleden |
Nel Nibcord
|
e9971f74ae
Abstract load checkpoint method
|
7 maanden geleden |
Nel Nibcord
|
223d35cea0
circular include lol
|
7 maanden geleden |
Nel Nibcord
|
8dc73074e6
Nodes don't need an abstract base class
|
7 maanden geleden |
Nel Nibcord
|
59af2dd592
Do we need casting here?
|
7 maanden geleden |
Nel Nibcord
|
b22c21ac16
Some session method cleanup
|
7 maanden geleden |
Nel Nibcord
|
98edb393b2
Initialize inference engine session in base class
|
7 maanden geleden |
Nel Nibcord
|
bcf87e79b7
Okay let's turn no_grad back on. We'll worry about that when tinygrad training works
|
7 maanden geleden |
Nel Nibcord
|
b7bbda3348
Removed tinygrad StatefulModel class, as it's no longer used
|
7 maanden geleden |
Nel Nibcord
|
67f5ae25a5
Fixing tinygrad model
|
7 maanden geleden |
Nel Nibcord
|
bfa3b36be5
Fixing tinygrad model
|
7 maanden geleden |
Nel Nibcord
|
37a75d6b96
Fixing tinygrad model
|
7 maanden geleden |
Nel Nibcord
|
0d3abfca95
Made models save properly
|
8 maanden geleden |
Nel Nibcord
|
9283f6d7bd
Correct loss propagation so we can see the actual loss instead of just the requestor shard's loss
|
7 maanden geleden |
Nel Nibcord
|
9eadee310b
Basic model saving
|
8 maanden geleden |
Nel Nibcord
|
38e368f00b
Fixed up the ops so that batches work
|
8 maanden geleden |
Nel Nibcord
|
dd3d99043b
Working distributed training
|
8 maanden geleden |
Nel Nibcord
|
175ebc1c42
Coordination biz
|
8 maanden geleden |
Nel Nibcord
|
3e869051f6
Okay we should probably await the update
|
8 maanden geleden |
Nel Nibcord
|
75c8650f1f
Naive network-propagated loss implementation on MLX
|
7 maanden geleden |
Nel Nibcord
|
836856824e
WIP: Training works on mlx
|
7 maanden geleden |
Nel Nibcord
|
a6fd7a3430
Generalizing some of the dataset biz while also creating uniform batches
|
8 maanden geleden |
Nel Nibcord
|
f5efbe1b8f
Initial distributed evaluation implementation
|
7 maanden geleden |
Alex Cheema
|
db9de97fa6
Merge pull request #549 from exo-explore/fixtokenencode
|
7 maanden geleden |
Alex Cheema
|
c593434808
fix encode endpoint
|
7 maanden geleden |