 Matthias Reso
|
8b0a233c1a
Use new chat format in custom dataset
|
1 gadu atpakaļ |
|
Matthias Reso
|
83fae41195
Add test for chat completion formatting
|
1 gadu atpakaļ |
|
Matthias Reso
|
6d9d48d619
Use apply_chat_template instead of custom functions
|
1 gadu atpakaļ |
|
Matthias Reso
|
5efea160a2
Adapt test_finetuning to new model
|
1 gadu atpakaļ |
|
Matthias Reso
|
739483f262
Adjust test_grammar_datasets to stable sort
|
1 gadu atpakaļ |
|
Matthias Reso
|
b96e435cda
Adjust test_samsum_dataset to second model
|
1 gadu atpakaļ |
|
Matthias Reso
|
fac41298b0
Adapt test_custom_dataset to new model
|
1 gadu atpakaļ |
|
Matthias Reso
|
960014a3bb
Fix test_custom_dataset by introducing a stable sort algorithm
|
1 gadu atpakaļ |
|
Matthias Reso
|
b5583b31d5
Adapt test_grammar_dataset to new model
|
1 gadu atpakaļ |
|
Matthias Reso
|
17a6d16289
Test batching for both llama versions
|
1 gadu atpakaļ |
|
Matthias Reso
|
a414ca6a57
Update chat format for llama3
|
1 gadu atpakaļ |
|
Matthias Reso
|
113ea18bf1
Replace LlamaTokenizer with AutoTokenizer
|
1 gadu atpakaļ |
 Hamid Shojanazeri
|
aaa9e2c863
Adding a feature that will stop the training/eval process after reaching some max_steps (#428)
|
1 gadu atpakaļ |
 Kai Wu
|
e6f69f84ad
add max_steps_reached to reduce redundancy
|
1 gadu atpakaļ |
|
Kai Wu
|
362cda0fa6
fixing test_gradient_accumulation and test_save_to_json
|
1 gadu atpakaļ |
|
Kai Wu
|
fa0a389f74
add max_step feature for training and eval
|
1 gadu atpakaļ |
|
Hamid Shojanazeri
|
37c8f72211
Update location and name of llm.py example notebook (#417)
|
1 gadu atpakaļ |
 Thomas Robinson
|
79266217ef
Update location and name of llm.py example notebook
|
1 gadu atpakaļ |
|
Hamid Shojanazeri
|
f7aa02af9f
only save training params on rank 0 (#415)
|
1 gadu atpakaļ |
 jpgard
|
6954b16b3b
only save training params on rank 0
|
1 gadu atpakaļ |
|
Hamid Shojanazeri
|
64e189914f
update due to peft new release (#407)
|
1 gadu atpakaļ |
|
Hamid Shojanazeri
|
11f51db28c
adding the kbit prep in the code
|
1 gadu atpakaļ |
|
Hamid Shojanazeri
|
f058ff6ccd
update due to peft new release
|
1 gadu atpakaļ |
|
Hamid Shojanazeri
|
6a7478a6aa
Reorg inference throughput folder structure (#404)
|
1 gadu atpakaļ |
 Chester Hu
|
367e4869ac
Reorg inference throughput folder structure
|
1 gadu atpakaļ |
|
Hamid Shojanazeri
|
d6eb83f6c5
Add llm class so that externally-hosted models can be called (#398)
|
1 gadu atpakaļ |
|
Thomas Robinson
|
0346d0d5b8
Add documentation and examples
|
1 gadu atpakaļ |
|
Hamid Shojanazeri
|
43a1e5cdb0
Fix dead links after directory structure refactor (#397)
|
1 gadu atpakaļ |
 Suraj Subramanian
|
e2a35420c0
Remove octoai link that is 401-ing
|
1 gadu atpakaļ |
|
Suraj Subramanian
|
12602f32e2
Merge branch 'main' into subramen-patch-deadlinks
|
1 gadu atpakaļ |
