issues
search
flowersteam
/
lamorel
Lamorel is a Python library designed for RL practitioners eager to use Large Language Models (LLMs).
MIT License
174
stars
14
forks
source link
issues
Newest
Newest
Most commented
Recently updated
Oldest
Least commented
Least recently updated
#40: Stability issues with gradient accumulation
#41
ClementRomac
closed
3 months ago
0
Stability issues in PPO examples
#40
ClementRomac
closed
3 months ago
0
A few questions & suggestions on `HF_LLM.py`
#39
sadra-barikbin
opened
4 months ago
1
Adding a log_softmax in LogScoringModuleFn to properly handle logits
#38
ClementRomac
closed
4 months ago
0
Missing log_softmax in score
#37
ClementRomac
closed
4 months ago
0
Better handling of empty candidates in forward
#36
ClementRomac
opened
4 months ago
2
#34 causal logprobs
#35
ClementRomac
closed
4 months ago
0
Logprobs don't match on causal models between generate and forward
#34
ClementRomac
closed
4 months ago
0
A question on a `break` statement in `HF_LLM::forward()`
#33
sadra-barikbin
opened
4 months ago
2
Finetuned Weights Loading Error
#32
AiBo123456
closed
5 months ago
3
Updated licenses in examples
#31
ClementRomac
closed
6 months ago
0
Using API in lamorel
#30
nuomizai
closed
6 months ago
2
Adapt to multi-agent training
#29
ewanlee
closed
6 months ago
2
Expand to multi-agent scenarios.
#28
ewanlee
opened
7 months ago
2
Connection closed by peer [127.0.1.1]: 14734
#27
ewanlee
closed
7 months ago
2
V0.2
#26
ClementRomac
closed
7 months ago
0
Ensuring gpu devices' id are native python integers
#25
ClementRomac
closed
7 months ago
0
Device 0 is not recognized
#24
giobin
closed
7 months ago
6
Connection error
#23
yone456
opened
8 months ago
13
AssertionError: torch distributed must be used!
#22
Jugg1er
closed
8 months ago
3
Fix device ordinal in single machine setup
#21
ClementRomac
closed
8 months ago
0
Remove the need of custom Accelerate version for single machine with single GPU
#20
ClementRomac
closed
8 months ago
0
A syntax error in __call_model
#19
Clayfigure
opened
9 months ago
3
why should we have two configs?
#18
HCHCXY
closed
8 months ago
2
How to load trained Flan-T5 model and then fine tune
#17
yanxue7
closed
8 months ago
6
Update README.md
#16
ClementRomac
closed
10 months ago
0
#14 Readme update for single-node with GPU
#15
ClementRomac
closed
10 months ago
0
Can't start PPO_finetuning example with 1 machine and 1 GPU
#14
tokarev-i-v
closed
10 months ago
1
Fix pre encode inputs
#13
ClementRomac
closed
12 months ago
0
Fix a tiny typo in `README.md`
#12
sadra-barikbin
closed
1 year ago
0
Fixing pre-encoding inputs
#11
ClementRomac
closed
1 year ago
0
Pre-encoding inputs crashes
#10
ClementRomac
closed
1 year ago
0
Put encoder's output to main device when pre-encoding candidates
#9
ClementRomac
closed
1 year ago
0
Using an encoder-decoder LLM with `pre_encode_inputs: true` doesn't work when `model_parallelism_size` > 1
#8
ClementRomac
closed
1 year ago
0
Could I directly run lamorel/examples/PPO_finuetuning/main.py?
#7
yanxue7
closed
1 year ago
4
Ppo upgrade
#6
ClementRomac
closed
1 year ago
0
Inefficient PPO example
#5
ClementRomac
closed
1 year ago
1
Forward optimizations + custom initializers
#4
ClementRomac
closed
1 year ago
0
#1 Updater are now given as instantiated objects
#3
ClementRomac
closed
1 year ago
0
What's the next step after running launch command?
#2
Shengqiang-Zhang
closed
1 year ago
2
Giving methods to updater is not possible
#1
ClementRomac
closed
1 year ago
0