 artek0chumak
|
1e227240e5
select adapter by name in handler.py
|
2 éve |
|
artek0chumak
|
e452df25cc
Add skeleton for peft init
|
2 éve |
|
artek0chumak
|
da204f1285
Add loading into device directly
|
2 éve |
|
artek0chumak
|
92612aebed
style
|
2 éve |
|
artek0chumak
|
ddd770000b
Add layer specific loading
|
2 éve |
|
artek0chumak
|
c02ae4c9ba
Fix long downloading
|
2 éve |
|
artek0chumak
|
9ea77b0911
Safer way to download pefts
|
2 éve |
|
artek0chumak
|
6563bf1750
Add first functional code
|
2 éve |
|
artek0chumak
|
02e1c95415
Add possible tests
|
2 éve |
 Alexander Borzunov
|
b28f5016ea
Delete deprecated petals.cli scripts (#336)
|
2 éve |
|
Alexander Borzunov
|
fa095f6461
Use 4-bit for llama by default, use bitsandbytes 0.40.0.post3 (#340)
|
2 éve |
|
Alexander Borzunov
|
158013a671
Implement direct server-to-server communication (#331)
|
2 éve |
|
Alexander Borzunov
|
4d9c26fe5c
Allow free_disk_space_for() remove arbitrary files from Petals cache (#339)
|
2 éve |
|
Alexander Borzunov
|
de930918a0
Support loading blocks in 4-bit (QLoRA NF4 format, disabled by default) (#333)
|
2 éve |
|
Alexander Borzunov
|
66a47c763e
Require pydantic < 2.0 (2.0 is incompatible with hivemind 1.1.8) (#337)
|
2 éve |
|
Alexander Borzunov
|
10c72acdf4
Fix warmup steps and minor issues in benchmarks (#334)
|
2 éve |
|
Alexander Borzunov
|
d126ee3053
Add benchmark scripts (#319)
|
2 éve |
|
Alexander Borzunov
|
fecee8c4dc
Show license links when loading models (#332)
|
2 éve |
|
Alexander Borzunov
|
47a2b1ee65
Fix llama's lm_head.weight.requires_grad (#330)
|
2 éve |
|
Alexander Borzunov
|
7a37513f77
Add AutoDistributed{Model, ModelForCausalLM, ModelForSequenceClassification} (#329)
|
2 éve |
|
Alexander Borzunov
|
cb3f018f9f
Add LLaMA support (#323)
|
2 éve |
 Max Ryabinin
|
5c0733711a
Use number of tokens for attn_cache_size (#286)
|
2 éve |
|
Max Ryabinin
|
c839173e57
Determine block dtype in a unified manner (#325)
|
2 éve |
|
Max Ryabinin
|
3e7ae5116d
Remove unused imports and attributes (#324)
|
2 éve |
|
Alexander Borzunov
|
675bacb592
Bump version to 1.1.5 (#312)
|
2 éve |
|
Alexander Borzunov
|
e026952338
Abort speedtest if it runs too long (#316)
|
2 éve |
|
Alexander Borzunov
|
6eb306a605
Raise error for unexpected .generate() kwargs (#315)
|
2 éve |
|
Alexander Borzunov
|
d9e7bfc949
Divide compute throughput by average no. of used blocks (#314)
|
2 éve |
|
Alexander Borzunov
|
6137b1b4b0
Replace .make_sequence(..., mode="random") with mode="max_throughput" (#313)
|
2 éve |
|
Alexander Borzunov
|
0a313bf6c5
Update hivemind to 1.1.8, enable efficient bfloat16 encoding (#311)
|
2 éve |
