 Kai Wu
|
41434dc825
formatted and removed duplicated or unused function get_total_flops() and byte2mb()
|
1 rok pred |
|
Kai Wu
|
f2e80bae22
created a FlopMeasure class on top of FlopCounterMode instead of keep of copy of our own tflop_counter.py
|
1 rok pred |
 Matthias Reso
|
a414ca6a57
Update chat format for llama3
|
1 rok pred |
|
Kai Wu
|
69e46887b4
handling incorrect profiling early stop caused by max_train_steps and add profiler.step() for each train step
|
1 rok pred |
|
Matthias Reso
|
113ea18bf1
Replace LlamaTokenizer with AutoTokenizer
|
1 rok pred |
 Beto
|
5979dbe996
Merging local with remote
|
1 rok pred |
|
Kai Wu
|
34e0bf4c6e
second draft of this feature, seems to be working now
|
1 rok pred |
|
Beto
|
d4cbfa1cc1
Merging upstream llama-recipes to current repo
|
1 rok pred |
|
Kai Wu
|
a35519ee90
fixed typo and handling unexpected exit
|
1 rok pred |
|
Kai Wu
|
2a5de9b448
first draft of flop counter feature
|
1 rok pred |
 Hamid Shojanazeri
|
aaa9e2c863
Adding a feature that will stop the training/eval process after reaching some max_steps (#428)
|
1 rok pred |
|
Kai Wu
|
e6f69f84ad
add max_steps_reached to reduce redundancy
|
1 rok pred |
 rahul-sarvam
|
0efb8bd31e
Update README.md
|
1 rok pred |
|
rahul-sarvam
|
687c2dc5d8
Update README.md
|
1 rok pred |
 Rahul A R
|
2fa8e69b62
add new argument: tokenizer_name
|
1 rok pred |
|
Rahul A R
|
f8183b96fe
use new tokenizer_name argument and resize embeddings if required
|
1 rok pred |
|
Rahul A R
|
1e4e3e00fc
adding new multilingual recipe
|
1 rok pred |
|
Kai Wu
|
362cda0fa6
fixing test_gradient_accumulation and test_save_to_json
|
1 rok pred |
|
Kai Wu
|
fa0a389f74
add max_step feature for training and eval
|
1 rok pred |
 Suraj Subramanian
|
201daff2d1
Add note on CUDA version + remove 'test' from pytorch whl url
|
1 rok pred |
|
Hamid Shojanazeri
|
37c8f72211
Update location and name of llm.py example notebook (#417)
|
1 rok pred |
 Thomas Robinson
|
79266217ef
Update location and name of llm.py example notebook
|
1 rok pred |
|
Hamid Shojanazeri
|
f7aa02af9f
only save training params on rank 0 (#415)
|
1 rok pred |
 jpgard
|
6954b16b3b
only save training params on rank 0
|
1 rok pred |
 varunfb
|
a404c9249c
Notebook to demonstrate using llama and llama-guard together using OctoAI
|
1 rok pred |
|
Beto
|
18d76ed36f
merging into private llama recipes repo
|
1 rok pred |
|
Hamid Shojanazeri
|
64e189914f
update due to peft new release (#407)
|
1 rok pred |
|
Hamid Shojanazeri
|
11f51db28c
adding the kbit prep in the code
|
1 rok pred |
|
Hamid Shojanazeri
|
f058ff6ccd
update due to peft new release
|
1 rok pred |
|
Hamid Shojanazeri
|
6a7478a6aa
Reorg inference throughput folder structure (#404)
|
1 rok pred |
