Alex Cheema
|
0bebf8dfde
|
fix indent
|
2025-01-30 20:21:28 +00:00 |
|
Alex Cheema
|
55c4385db5
|
cleanup tmp files on failed download
|
2025-01-30 20:11:06 +00:00 |
|
Alex Cheema
|
90690a7d10
|
Merge pull request #647 from deftdawg/patch-1
Add 4-bit to the end of DeepSeek V3/R1 model descriptions
|
2025-01-30 19:49:38 +00:00 |
|
Alex Cheema
|
130d998d36
|
Merge pull request #659 from exo-explore/robustdownload
ensure exo dir on start, retry with exp backoff on file downloads
|
2025-01-30 19:49:00 +00:00 |
|
Alex Cheema
|
788c49784c
|
retry fetch_file_list also
|
2025-01-30 19:45:12 +00:00 |
|
Alex Cheema
|
6b1c8635fc
|
ensure exo dir on start, retry with exp backoff on file downloads
|
2025-01-30 19:40:35 +00:00 |
|
Alex Cheema
|
24c410c19c
|
Merge pull request #653 from exo-explore/tinyfixes
Tiny fixes
|
2025-01-29 19:08:05 +00:00 |
|
Alex Cheema
|
f6ed830ba6
|
Merge pull request #651 from exo-explore/parallelise_model_loadin
parallelise model loading
|
2025-01-29 19:07:25 +00:00 |
|
Alex Cheema
|
e6b4f2993c
|
fix prompt output spacing in tui
|
2025-01-29 19:01:30 +00:00 |
|
DeftDawg
|
a25e02c913
|
Add 4-bit to the end of DeepSeek V3/R1 model descriptions
|
2025-01-29 14:00:13 -05:00 |
|
Alex Cheema
|
3675804f4d
|
throttle repo progress events and only send them out if something changed
|
2025-01-29 18:55:54 +00:00 |
|
Alex Cheema
|
96f1aecb05
|
only in_progress if any given file is in_progress
|
2025-01-29 18:43:43 +00:00 |
|
Alex Cheema
|
23a5030604
|
even if part of a file is downloaded it may not be in_progress
|
2025-01-29 18:39:23 +00:00 |
|
Alex Cheema
|
31b56e862f
|
make a singleton thread pool executor for tinygrad since we always want it to run on the same thread
|
2025-01-29 18:37:09 +00:00 |
|
Alex Cheema
|
9f6c688d62
|
update tinygrad
|
2025-01-29 18:06:38 +00:00 |
|
Alex Cheema
|
4887be5103
|
parallelise model loading
|
2025-01-29 02:32:59 +00:00 |
|
Alex Cheema
|
75091e206b
|
Merge pull request #650 from exo-explore/chatgpttimeout
increase chatgpt api response timeout to 900 seconds
|
2025-01-29 02:03:52 +00:00 |
|
Alex Cheema
|
141de0d011
|
increase chatgpt api response timeout to 900 seconds
|
2025-01-29 02:03:00 +00:00 |
|
Alex Cheema
|
263b18a31e
|
Merge pull request #649 from eclecticc/amd_fix
Fix AMD device capabilities fields
|
2025-01-29 02:01:06 +00:00 |
|
Nirav Patel
|
9cf6818f10
|
Fix AMD device capabilities fields
|
2025-01-28 16:58:58 -08:00 |
|
Alex Cheema
|
837ed5d980
|
Merge pull request #648 from exo-explore/modelasyncload
Fixes
|
2025-01-28 23:39:11 +00:00 |
|
Alex Cheema
|
9c1bea97e8
|
fix embed_tokens for last layer in qwen models
|
2025-01-28 23:09:45 +00:00 |
|
Alex Cheema
|
af171f06fa
|
propagate prompts to other nodes so they can display them, cleaner prompt/output output
|
2025-01-28 21:50:49 +00:00 |
|
Alex Cheema
|
edfa53a4c2
|
Merge pull request #646 from exo-explore/modelasyncload
make sure mlx stuff is on separate thread non blocking
|
2025-01-28 18:56:19 +00:00 |
|
Alex Cheema
|
4a5b80a958
|
make sure mlx stuff is on separate thread non blocking
|
2025-01-28 18:56:00 +00:00 |
|
Alex Cheema
|
92d1bc01de
|
Merge pull request #645 from exo-explore/modelasyncload
load mlx model shard on mlx thread so it doesnt block
|
2025-01-28 18:49:47 +00:00 |
|
Alex Cheema
|
6662d5668c
|
load mlx model shard on mlx thread so it doesnt block
|
2025-01-28 18:49:19 +00:00 |
|
Alex Cheema
|
a0d673fa3a
|
Merge pull request #640 from exo-explore/simpledownload
Simple download
|
2025-01-27 19:38:11 +00:00 |
|
Alex Cheema
|
7c649085a1
|
fix eta/speed for resuming an existing download, using the session downloaded bytes
|
2025-01-27 19:23:18 +00:00 |
|
Alex Cheema
|
90e0e2761f
|
ignore not_started progress updates
|
2025-01-27 06:05:59 +00:00 |
|
Alex Cheema
|
265586f7b4
|
set timeout on get too
|
2025-01-27 06:05:40 +00:00 |
|
Alex Cheema
|
4748bb7dc7
|
increase file download timeout to 30min
|
2025-01-27 05:49:17 +00:00 |
|
Alex Cheema
|
ae770db4f3
|
increase download chunks to 1MB
|
2025-01-27 05:37:50 +00:00 |
|
Alex Cheema
|
82f75d0ccf
|
increase hf download http timeout 15 mins for large downloads
|
2025-01-27 05:20:30 +00:00 |
|
Alex Cheema
|
295f41c5cc
|
increase bench job timeout to give enough time to download
|
2025-01-27 05:03:35 +00:00 |
|
Alex Cheema
|
19a27c5bfd
|
HF_HOME -> EXO_HOME
|
2025-01-27 02:59:23 +00:00 |
|
Alex Cheema
|
d7ca9b7732
|
show each node id in the tinychat topology viz
|
2025-01-27 02:20:22 +00:00 |
|
Alex Cheema
|
b349e48b0d
|
fix visual bug where frontend would show the full hf repo size, but in some cases that includes redundant files so we should use the model index in those cases too
|
2025-01-27 02:13:05 +00:00 |
|
Alex Cheema
|
21586063f6
|
use llama-3.2-1b in tinygrad test
|
2025-01-27 01:35:33 +00:00 |
|
Alex Cheema
|
277d63d860
|
special case when a model doesnt have a model index file, then use wildcard for allow_patterns
|
2025-01-27 01:26:15 +00:00 |
|
Alex Cheema
|
74379ef671
|
log download logs with DEBUG>=6 very verbose
|
2025-01-27 01:11:54 +00:00 |
|
Alex Cheema
|
3c7bd48aa3
|
get rid of some more hf bloat
|
2025-01-27 01:08:46 +00:00 |
|
Alex Cheema
|
1df023023e
|
remove a lot of hf bloat
|
2025-01-27 01:06:47 +00:00 |
|
Alex Cheema
|
b89495f444
|
rewrite ShardDownloader, simplify significantly
|
2025-01-27 00:37:57 +00:00 |
|
Alex Cheema
|
903950f64e
|
Merge pull request #638 from exo-explore/deepseekv3fix
add exception for mlx-community/DeepSeek-R1-3bit and mlx-community/DeepSeek-V3-3bit in tokenizers test
|
2025-01-26 20:33:22 +00:00 |
|
Alex Cheema
|
a3766f538a
|
add exception for mlx-community/DeepSeek-R1-3bit and mlx-community/DeepSeek-V3-3bit in tokenizers test
|
2025-01-26 20:32:48 +00:00 |
|
Alex Cheema
|
9711d632e0
|
Merge pull request #637 from exo-explore/deepseekv3fix
fix post_init deepseek v3
|
2025-01-26 20:31:53 +00:00 |
|
Alex Cheema
|
82ef086010
|
add deepseek-v3-3bit and deepseek-r1-3bit
|
2025-01-26 20:31:28 +00:00 |
|
Alex Cheema
|
55ea366932
|
fix post_init deepseek v3
|
2025-01-26 20:27:31 +00:00 |
|
Alex Cheema
|
63318983de
|
Merge pull request #631 from sigseg5/main
Some adaptivity fixes in tinychat
|
2025-01-26 19:20:58 +00:00 |
|