 Alex Cheema
|
31b56e862f
make a singleton thread pool executor for tinygrad since we always want it to run on the same thread
|
il y a 7 mois |
|
Alex Cheema
|
9f6c688d62
update tinygrad
|
il y a 7 mois |
|
Alex Cheema
|
4887be5103
parallelise model loading
|
il y a 7 mois |
|
Alex Cheema
|
141de0d011
increase chatgpt api response timeout to 900 seconds
|
il y a 7 mois |
 Alex Cheema
|
837ed5d980
Merge pull request #648 from exo-explore/modelasyncload
|
il y a 7 mois |
|
Alex Cheema
|
9c1bea97e8
fix embed_tokens for last layer in qwen models
|
il y a 7 mois |
|
Alex Cheema
|
af171f06fa
propagate prompts to other nodes so they can display them, cleaner prompt/output output
|
il y a 7 mois |
|
Alex Cheema
|
edfa53a4c2
Merge pull request #646 from exo-explore/modelasyncload
|
il y a 7 mois |
|
Alex Cheema
|
4a5b80a958
make sure mlx stuff is on separate thread non blocking
|
il y a 7 mois |
|
Alex Cheema
|
92d1bc01de
Merge pull request #645 from exo-explore/modelasyncload
|
il y a 7 mois |
|
Alex Cheema
|
6662d5668c
load mlx model shard on mlx thread so it doesnt block
|
il y a 7 mois |
|
Alex Cheema
|
a0d673fa3a
Merge pull request #640 from exo-explore/simpledownload
|
il y a 7 mois |
|
Alex Cheema
|
7c649085a1
fix eta/speed for resuming an existing download, using the session downloaded bytes
|
il y a 7 mois |
|
Alex Cheema
|
90e0e2761f
ignore not_started progress updates
|
il y a 7 mois |
|
Alex Cheema
|
265586f7b4
set timeout on get too
|
il y a 7 mois |
|
Alex Cheema
|
4748bb7dc7
increase file download timeout to 30min
|
il y a 7 mois |
|
Alex Cheema
|
ae770db4f3
increase download chunks to 1MB
|
il y a 7 mois |
|
Alex Cheema
|
82f75d0ccf
increase hf download http timeout 15 mins for large downloads
|
il y a 7 mois |
|
Alex Cheema
|
295f41c5cc
increase bench job timeout to give enough time to download
|
il y a 7 mois |
|
Alex Cheema
|
19a27c5bfd
HF_HOME -> EXO_HOME
|
il y a 7 mois |
|
Alex Cheema
|
d7ca9b7732
show each node id in the tinychat topology viz
|
il y a 7 mois |
|
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
|
il y a 7 mois |
|
Alex Cheema
|
21586063f6
use llama-3.2-1b in tinygrad test
|
il y a 7 mois |
|
Alex Cheema
|
277d63d860
special case when a model doesnt have a model index file, then use wildcard for allow_patterns
|
il y a 7 mois |
|
Alex Cheema
|
74379ef671
log download logs with DEBUG>=6 very verbose
|
il y a 7 mois |
|
Alex Cheema
|
3c7bd48aa3
get rid of some more hf bloat
|
il y a 7 mois |
|
Alex Cheema
|
1df023023e
remove a lot of hf bloat
|
il y a 7 mois |
|
Alex Cheema
|
b89495f444
rewrite ShardDownloader, simplify significantly
|
il y a 7 mois |
|
Alex Cheema
|
903950f64e
Merge pull request #638 from exo-explore/deepseekv3fix
|
il y a 7 mois |
|
Alex Cheema
|
a3766f538a
add exception for mlx-community/DeepSeek-R1-3bit and mlx-community/DeepSeek-V3-3bit in tokenizers test
|
il y a 7 mois |
