These 32 commits are when the Protocol Buffers files have changed:
| Commit: | 9954ce8 | |
|---|---|---|
| Author: | Alex Cheema | |
fix treating token as a list
The documentation is generated from this commit.
| Commit: | 218c1e7 | |
|---|---|---|
| Author: | Alex Cheema | |
Merge branch 'main' into runners2
| Commit: | 5460529 | |
|---|---|---|
| Author: | Pranav Veldurthi | |
Merge Latest
| Commit: | db010d5 | |
|---|---|---|
| Author: | Alex Cheema | |
distributed tracing
| Commit: | c9ded9b | |
|---|---|---|
| Author: | Alex Cheema | |
| Committer: | Alex Cheema | |
optimise networking, remove bloat
| Commit: | f55a53a | |
|---|---|---|
| Author: | Alex Cheema | |
| Committer: | Alex Cheema | |
one token at a time
| Commit: | f5efbe1 | |
|---|---|---|
| Author: | Nel Nibcord | |
| Committer: | Nel Nibcord | |
Initial distributed evaluation implementation
| Commit: | 9283f6d | |
|---|---|---|
| Author: | Nel Nibcord | |
| Committer: | Nel Nibcord | |
Correct loss propagation so we can see the actual loss instead of just the requestor shard's loss
| Commit: | 75c8650 | |
|---|---|---|
| Author: | Nel Nibcord | |
| Committer: | Nel Nibcord | |
Naive network-propagated loss implementation on MLX
| Commit: | 8368568 | |
|---|---|---|
| Author: | Nel Nibcord | |
| Committer: | Nel Nibcord | |
WIP: Training works on mlx Still debugging some tinygrad stuff, and fixing comms
| Commit: | 5534424 | |
|---|---|---|
| Author: | Alex Cheema | |
use double for flops protobuf
| Commit: | b8c4b46 | |
|---|---|---|
| Author: | Alex Cheema | |
prioritise network interfaces and display in the tui
| Commit: | ca0caad | |
|---|---|---|
| Author: | Pranav Veldurthi | |
Image to image generation
| Commit: | 6b28ef0 | |
|---|---|---|
| Author: | Pranav Veldurthi | |
Stable stable diffusion mlx
| Commit: | 8b71d57 | |
|---|---|---|
| Author: | Nel Nibcord | |
Removed inference state entirely For tinygrad, this means caching start_pos in StatefulModel
| Commit: | 82cce44 | |
|---|---|---|
| Author: | Nel Nibcord | |
| Committer: | Nel Nibcord | |
Some initial inference engine refactors for enabling training Only on MLX for now, breaks Tinygrad (doesn't fulfill interface yet)
| Commit: | de2f6d2 | |
|---|---|---|
| Author: | Alex Cheema | |
implement a health check for peers and discovery should only return healthy peers
| Commit: | acc94b5 | |
|---|---|---|
| Author: | Varshith | |
chatgpt api integration
| Commit: | 2084784 | |
|---|---|---|
| Author: | Alex Cheema | |
per-request kv cache, remove all explicit reset functionality as it wasnt used. fixes #67
| Commit: | 1475c73 | |
|---|---|---|
| Author: | Alex Cheema | |
fix inference_state serialization. related: #40 #44 #45
| Commit: | 4b592f9 | |
|---|---|---|
| Author: | Alex Cheema | |
exo topology visualisation that shows the topology of the network, device capabilities and the currently active node using opaque statuses. fixes #36. ready for #33
| Commit: | 54c9860 | |
|---|---|---|
| Author: | Alex Cheema | |
more robust grpc discovery with asyncio and proper error handling, add flops to device capabilities. fixes #23 and progress on #33
| Commit: | 8a35fd8 | |
|---|---|---|
| Author: | Alex Cheema | |
support chatgpt api endpoint fron any node #24
| Commit: | 03ba31c | |
|---|---|---|
| Author: | Alex Cheema | |
| Committer: | Alex Cheema | |
wip state
| Commit: | e17905e | |
|---|---|---|
| Author: | Alex Cheema | |
add global reset
| Commit: | d2184f5 | |
|---|---|---|
| Author: | Alex Cheema | |
keep track of already visited peers in global operations: collect_topology
| Commit: | 5bbde22 | |
|---|---|---|
| Author: | Alex Cheema | |
move everything under exo module
| Commit: | b01f69b | |
|---|---|---|
| Author: | Alex Cheema | |
add support for multiple concurrent requests with request ids
| Commit: | 36b8456 | |
|---|---|---|
| Author: | Alex Cheema | |
| Committer: | Alex Cheema | |
collect global topology with local peer visibility, ring memory weighted partitioning strategy
| Commit: | 6c8c9ee | |
|---|---|---|
| Author: | Alex Cheema | |
topology with partitioning strategy
| Commit: | 563dcb5 | |
|---|---|---|
| Author: | Alex Cheema | |
mlx sharded implementation with example of distributed inference
| Commit: | a21f59f | |
|---|---|---|
| Author: | Alex Cheema | |
scaffolding for networking, inference and orchestration