Commit Graph - exo - L'Électron Rare

electron-rare/exo

Fork 0

Commit Graph

Select branches

Hide Pull Requests

JakeHillion/images-formatted

JakeHillion/poulwnyystws

JakeHillion/qolrqmoxuotr

JakeHillion/qotmmkpkktsv

JakeHillion/qqwursqotloo

JakeHillion/rwopytsxwtlp

JakeHillion/syyqxzsolzox

JakeHillion/torrents

JakeHillion/vmwnonkzsnuo

JakeHillion/vnwuyypxwqqs

JakeHillion/wnpzkupouslp

JakeHillion/yvqrzonuknxu

add-glm5-support

ai-slop

aiohttp

alexcheema/add-qwen3-coder-next

alexcheema/add-step-3.5-flash

alexcheema/continuous-batching

alexcheema/dashboard-stop-button

alexcheema/debug-state-catchup-race

alexcheema/event-index-liveness

alexcheema/external-model-paths

alexcheema/fix-download-resume-buttons

alexcheema/fix-instance-preparing-delay

alexcheema/fix-local-network-permission-reboot

alexcheema/fix-stuck-download-after-cancel

alexcheema/graceful-loading-shutdown

alexcheema/meta-instance

alexcheema/mlx-distributed-transfer

alexcheema/model-picker-downloaded-filter

alexcheema/model-selection-ux

alexcheema/mtp-speculative-decoding

alexcheema/no-silent-instance-override

alexcheema/rdma-health-detection

alexcheema/reproduce-gpu-lock

alexcheema/restore-thunderbolt-bridge-launchdaemon

alexcheema/robust-hf-config-parsing

alexcheema/runner-health-check

alexcheema/speculative-decoding

alexcheema/topology-download-indicators

alexcheema/trust-remote-code-cli-flag

architecture-based-tokenizer-checks

babbler

bootstrap-peer

bump-required-mem

ciaran/download-coordinator-crash

ciaran/eval

ciaran/handle-disconnects

ciaran/image-demo

ciaran/storage-management

consistent-placement-api

dashboard-show-failure-reasons

david/batched-kernels

david/lcb_eval

david/mla-context-parallel

david/new-kernels

david/new-kernels-v2

david/speculative-mtp

download-pause-delete

e2e-tests

failed=stopped

feat/bandwidth-aware-placement

feat/dashboard-light-mode

feat/dashboard-light-mode-v2

feat/e2e-chaos-tests

feat/mac-studio-rdma-port-warning

feat/meta-instance-dashboard

feat/meta-instance-dashboard-ui

feat/prefill-eta

fix-instance-preview-type

fix/1025

fix/instance-preview-type-mismatch

fix/no-downloads-model-loading

fix/rdma-tb5-tooltip

fix/tb5-detection

graduate-image-gen

handle-message-too-large

iroh

iroh-migration

jaccl-build-fix

leo/add-cuda-typings

leo/add-glm-5

leo/add-instance-settings

leo/add-logprobs-to-chatcompletion

leo/add-more-tensor-strategies

leo/add-tool-calling

leo/add-uneven-sharding

leo/address-rdma-gpu-locks

leo/dgx-spark-integrations

leo/exo-bench-conc-fix

leo/fix-small-pipeline-rdma

leo/fix-step35flash-parser

leo/gguf-maybe

leo/handle-low-memory-situations

leo/log-kimi-k25

leo/prefill-decode-really

leo/prepare-batch-implementation

leo/prioritise-thunderbolt-ip-for-ring

leo/profile-socket-connection-metrics

leo/reduce-log-spam

leo/send-from-last-rank

leo/test-branch

leo/test-model-hanging

leo/test-stuff-1

leo/tmp-test

leo/try-custom-mlx-branch

leo/use-sm121-branch

main

meta-instance-split/jaccl-sidechannel

new-bridge-script

nid-persist

optimize-dashboard

releases/v1.0.65

releases/v1.0.66

releases/v1.0.67

remove-custom-discovery

runner-opts

runner-refactor

runner-shutdown-vs-failed

rust-explore-2

sami/dashboard-tests

sami/flash

sami/iOS-app

sami/image-gen-safety

simplify-downloads

splitting-rust-rewrite-3

streaming-delete-hang

support-mlx-device-backends

test-app

vllm-nix

gh-screenshot-assets

pr-1465-screenshots

test-screenshots-tmp

v0.0.1-alpha

v0.0.10-alpha

v0.0.11-alpha

v0.0.12-alpha

v0.0.13-alpha

v0.0.15-alpha

v0.0.2-alpha

v0.0.3-alpha

v0.0.4-alpha

v0.0.5-alpha

v0.0.7-alpha

v0.0.8-alpha

v0.0.9-alpha

v1.0

v1.0.60-alpha.1

v1.0.60-alpha0

v1.0.61

v1.0.61-alpha.0

v1.0.61-alpha.1

v1.0.61-alpha.2

v1.0.62

v1.0.63

v1.0.64

v1.0.65

v1.0.66

v1.0.67

v1.0.68

v1.0.69

vv0.0.10-alpha

7ed4639540 use structured concurrency in download coordinator (#1722) Evan QuineyandGitHub 2026-03-13 14:55:37 +00:00
29d4165fe2 Add step logo condition to FamilyLogos component (#1676) Mazin SharafandGitHub 2026-03-14 00:31:46 +11:00
8b2fb9330c Handle download coordinator exception ciaran/download-coordinator-crash ciaranbor 2026-03-12 23:44:57 +00:00
e9e23e556e Have loading progress Ryuichi Leo Takashige 2026-03-12 19:17:26 +00:00
12af7c9586 fix: shield macmon cleanup from cancellation to prevent orphaned process (#1714) ecohash-coandGitHub 2026-03-13 07:54:30 -05:00
f28b2fd037 Extract mlx revision from uv lock (#1715) ciaranborandGitHub 2026-03-13 12:34:54 +00:00
169ea2a5e8 Fix GPT OSS by not retokenizing prompts Ryuichi Leo Takashige 2026-03-12 17:56:52 +00:00
4a7901c548 Fix patches Ryuichi Leo Takashige 2026-03-12 17:37:31 +00:00
7bb5cb4fc7 Allow memory profiling to be unstable Ryuichi Leo Takashige 2026-03-12 17:34:53 +00:00
493e342f83 Skip impossible shardings Ryuichi Leo Takashige 2026-03-12 17:33:12 +00:00
283b1809c9 Move VLLM runner into VLLM engine Ryuichi Leo Takashige 2026-03-12 17:20:30 +00:00
35030119e3 ExoBench and ExoEval for CUDA Ryuichi Leo Takashige 2026-03-12 17:09:51 +00:00
1350a409ff CUDA TYPINGS leo/add-cuda-typings Ryuichi Leo Takashige 2026-03-12 16:25:48 +00:00
585dfe3549 Merge branch 'main' into leo/dgx-spark-integrations rltakashigeandGitHub 2026-03-12 15:41:29 +00:00
5d7a005a13 Destroy process group on Keyboard Interrupt Ryuichi Leo Takashige 2026-03-12 15:33:32 +00:00
87b7c5ef8b Pass CI Ryuichi Leo Takashige 2026-03-12 15:20:10 +00:00
957ebbd21f Set max token length as max context length if no max tokens set Ryuichi Leo Takashige 2026-03-12 15:14:51 +00:00
ea18a62581 fix: guard against ZeroDivisionError in mlx_lm stats (#1707) Mustafa Alp YılmazandGitHub 2026-03-12 17:48:36 +03:00
4b6dd7588f lockgit status rltakashige 2026-03-12 14:41:33 +00:00
3f4f7c9ba6 Only do for aarch64 linux Ryuichi Leo Takashige 2026-03-12 11:25:41 +00:00
1331465ba0 Add missing runner features Ryuichi Leo Takashige 2026-03-12 00:31:52 +00:00
0782d90ec5 fix: show partial download progress on initial dashboard load (#1706)
Miguel Cruz GitHub Evan
2026-03-12 06:39:34 -04:00
8f94727f14 Ignore missing modules if type stubs exist Ryuichi Leo Takashige 2026-03-11 23:52:05 +00:00
9ee23ee0d3 Make vllm inference runner closer to the normal inference runner Ryuichi Leo Takashige 2026-03-11 23:41:09 +00:00
f75d36cbe0 Fix cache patch Ryuichi Leo Takashige 2026-03-11 21:57:15 +00:00
2683ac7b61 Add Torch typings Ryuichi Leo Takashige 2026-03-11 21:40:40 +00:00
404b9769ac Download models without model.safetensors.index Ryuichi Leo Takashige 2026-03-11 21:27:21 +00:00
3e097f7243 only download a single copy of the model. Ryuichi Leo Takashige 2026-03-11 20:57:29 +00:00
0c8615f25c Only import VLLM once.. Ryuichi Leo Takashige 2026-03-11 19:19:25 +00:00
ba35a4ba13 Move VLLM into the runner and add type stubs Ryuichi Leo Takashige 2026-03-11 19:07:36 +00:00
9a83fa6cdf Patch VLLM to load multiple models dynamically Ryuichi Leo Takashige 2026-03-11 17:21:17 +00:00
659c1bc737 Progress: Run EXO-CUDA through nix! Ryuichi Leo Takashige 2026-03-11 12:52:00 +00:00
34df811b92 Vibe coding design baby Ryuichi Leo Takashige 2026-03-10 17:21:15 +00:00
ca5870a2e8 Some Linux Laptop/Desktop detection and goodbye penguin Ryuichi Leo Takashige 2026-03-10 17:01:59 +00:00
6be6ea5fd2 Fix placement preview Ryuichi Leo Takashige 2026-03-10 16:55:28 +00:00
5e9d27b753 Show Sparks and Linux in topology Ryuichi Leo Takashige 2026-03-10 13:03:11 +00:00
cfc8f09004 Fast direct USB connectivity Ryuichi Leo Takashige 2026-03-10 11:21:16 +00:00
f221a6c85c Normalise Responses API tool call format (#1704) rltakashigeandGitHub 2026-03-11 18:10:12 +00:00
acb4818057 Point mlx-lm to fix/float32-logprobs branch for float32 log_softmax fix david/lcb_eval james 2026-03-11 09:02:29 -07:00
2994b41089 fix: validate num_key_value_heads in tensor sharding placement (#1669) Mustafa Alp YılmazandGitHub 2026-03-11 16:46:33 +03:00
b26b50053d Fix repetition_penalty crash: default repetition_context_size to 64 dmcc73andClaude Opus 4.6 2026-03-10 20:41:41 +00:00
ebc7e6c100 Add top_k, min_p, repetition_penalty CLI args to exo_eval.py dmcc73andClaude Opus 4.6 2026-03-10 20:30:15 +00:00
ce67529524 vibe coding gguf leo/gguf-maybe Ryuichi Leo Takashige 2026-03-10 18:32:48 +00:00
c3f3d0eeb9 Vibe coding design baby Ryuichi Leo Takashige 2026-03-10 17:21:15 +00:00
7199e556ef Some Linux Laptop/Desktop detection and goodbye penguin Ryuichi Leo Takashige 2026-03-10 17:01:59 +00:00
f376c24895 Fix placement preview Ryuichi Leo Takashige 2026-03-10 16:55:28 +00:00
38f0c09175 fix: use StreamingDetokenizer in batch generator to fix emoji/UTF-8 corruption (#1691)
Mustafa Alp Yılmaz GitHub Ryuichi Leo Takashige
2026-03-10 19:10:22 +03:00
d63277edbd Show Sparks and Linux in topology Ryuichi Leo Takashige 2026-03-10 13:03:11 +00:00
dd121ff966 Fast direct USB connectivity Ryuichi Leo Takashige 2026-03-10 11:21:16 +00:00
f36fd56c38 Include power usage in bench responses (#1692) ciaranborandGitHub 2026-03-10 16:02:55 +00:00
82c54dd6d6 Add support for Nemotron sharding (#1693) rltakashigeandGitHub 2026-03-10 15:51:07 +00:00
54ad960ffb Sort LCB dataset by question_id to match official ordering dmcc73andClaude Opus 4.6 2026-03-10 14:17:33 +00:00
81c5d9379f Use official LCB dataset loading for consistent problem ordering dmcc73andClaude Opus 4.6 2026-03-10 14:16:42 +00:00
cd718ed726 Add LCB-compatible mode to exo_eval.py for uniform benchmarking dmcc73andClaude Opus 4.6 2026-03-10 13:42:35 +00:00
3536161f15 Fix stale state.runners (#1684) ciaranborandGitHub 2026-03-10 09:32:25 +00:00
a6aa07ed83 feat(dashboard): add mobile drawer support for sidebars (#1677) ArvidSUandGitHub 2026-03-10 05:18:11 +01:00
613d47aa1f test adding new kernels in exo davidmcc73andciaranbor 2026-03-09 19:57:57 +00:00
131ad0ff36 Implement continuous batching (#1642)
rltakashige GitHub Evan Quiney
2026-03-09 15:04:45 +00:00
d01636100a Ciaran/gpt oss tool call finish reason (#1673) ciaranborandGitHub 2026-03-06 21:09:20 +00:00
79c8dbeacd Ciaran/minor download bugs (#1674) ciaranborandGitHub 2026-03-06 17:46:18 +00:00
0096159728 up gossipsub limit (#1671) Evan QuineyandGitHub 2026-03-06 14:20:51 +00:00
7a36d3968d docs: Update documentation for v1.0.68 release (#1667)
Andrei Onel GitHubaskmanu[bot] <192355599+askmanu[bot]@users.noreply.github.com>Evan Quiney
2026-03-06 11:32:46 +00:00
eee3432738 feat(mlx): add repetition_penalty and repetition_context_size to chat completions (#1665) Mustafa Alp YılmazandGitHub 2026-03-06 13:51:25 +03:00
e8c3a873a6 feat(dashboard): improve model picker feedback and HuggingFace search (#1661)
Alex Cheema GitHub Evan
2026-03-06 18:23:08 +08:00
afab3095b0 fix: KVPrefixCache Regression (#1668) Michael HarriganandGitHub 2026-03-06 05:11:08 -05:00
ed3a974c61 fix: sign all Sparkle framework nested binaries inside-out alexcheema/fix-local-network-permission-reboot Alex CheemaandClaude Opus 4.6 2026-03-06 01:48:15 -08:00
e90289635e fix: preserve entitlements and remove --deep codesign to fix local network permission on reboot Alex CheemaandClaude Opus 4.6 2026-03-06 01:07:19 -08:00
b9d40e8e35 Ciaran/re download bug (#1658) ciaranborandGitHub 2026-03-05 14:18:17 +00:00
3a4d635d0c Fix copy code button not working in dashboard (#1659)
wysie GitHub Wysie
2026-03-05 20:41:17 +08:00
8155e14972 fix(dashboard): only suggest RDMA for TB5 hardware, not all Thunderbolt fix/tb5-detection Alex CheemaandClaude Opus 4.6 2026-03-04 13:32:09 -08:00
69a084cf97 disconnect on mdns expiry remove-custom-discovery Evan 2026-03-04 18:55:21 +00:00
9f184de322 clean iroh Evan 2026-02-18 17:53:52 +00:00
b59bb9dba6 clippy Evan 2026-03-04 17:25:44 +00:00
530c125462 simplify Evan 2026-03-04 16:07:34 +00:00
738cf1628d no more wakerdeque Evan 2026-03-04 16:03:42 +00:00
1dc8d9c9ad forgot this bit Evan 2026-03-04 13:28:33 +00:00
7b2673c3c7 maybemaybe Evan 2026-02-26 15:40:07 +00:00
8485805042 fix(worker): emit error chunks when a runner dies mid-command (#1645)
Miguel Miranda Dias GitHub Evan
2026-03-04 18:15:58 +01:00
4de8f801c7 #Add reasoning parms to chat completion and responses APIs (#1654) ciaranborandGitHub 2026-03-04 14:27:49 +00:00
5777bf3c39 fix: coerce tool-call argument types from tool schema (#1651)
Owleksiy GitHub Evan
2026-03-04 04:22:53 -08:00
886192f1e6 ignore closed resource errors when trying to cancel a task (#1652) Evan QuineyandGitHub 2026-03-03 16:48:43 +00:00
2ab66e112d or maybe this bump-required-mem Evan 2026-03-03 15:25:57 +00:00
aea2653270 bump required memory to load a model by 10% Evan 2026-03-03 15:14:15 +00:00
d914acd64e check if we have a task before we delete it (#1634) Evan QuineyandGitHub 2026-03-03 15:32:12 +00:00
3239c55e40 Move mx_all_gather_tasks into utils_mlx leo/prepare-batch-implementation Ryuichi Leo Takashige 2026-03-03 14:39:23 +00:00
37296c8249 Refactor runner for implementing batching (#1632)
rltakashige GitHub Evan Quiney
2026-03-03 14:38:55 +00:00
725264cc33 pass CI yet again Ryuichi Leo Takashige 2026-03-03 14:33:14 +00:00
401ccfbd30 Merge branch 'main' into leo/prepare-batch-implementation rltakashigeandGitHub 2026-03-03 14:32:25 +00:00
28817d3ee3 Add support for Qwen3.5 (#1644)
Daiz GitHub hw Ryuichi Leo Takashige Evan
2026-03-03 23:31:57 +09:00
06beffe0e2 Pass CI Ryuichi Leo Takashige 2026-03-03 14:18:34 +00:00
e9193581bc Batch cleanup (#1649)
Evan Quiney GitHub Ryuichi Leo Takashige
2026-03-03 14:06:13 +00:00
5087674b60 runner opts runner-opts Evan 2026-02-26 17:51:31 +00:00
16a3ab5155 sloppin it up ai-slop Evan 2026-02-28 13:30:01 +00:00
69628383c5 Match with image runner Ryuichi Leo TakashigeandEvan 2026-03-02 03:11:51 +00:00
f77a672126 Refactor runner for implementing batching Ryuichi Leo TakashigeandEvan 2026-02-27 13:22:23 +00:00
0e1b9501c3 fix: mini topology sidebar navigates home on click (#1616) Alex CheemaandGitHub 2026-03-03 02:46:52 -08:00
f0d4ccbeb3 feat: add POST /v1/cancel/{command_id} endpoint (#1579)
Mustafa Alp Yılmaz GitHub Evan
2026-03-03 13:38:52 +03:00
bee33f78a0 model_cards: add pinning by git sha JakeHillion/rwopytsxwtlp Jake Hillion 2026-02-26 23:55:37 +00:00
858dc808df fix: replace Master event_sender after EventRouter recreation (#1630) (#1637)
wysie GitHub Wysie
2026-03-02 18:05:47 +08:00
75318b4dc9 router: remove unreachable code in get_node_id_keypair JakeHillion/syyqxzsolzox Jake Hillion 2026-02-27 17:29:13 +00:00

1 2 3 4 5 ...