pytorch torchtune issues

pytorch / torchtune

A Native-PyTorch Library for LLM Fine-tuning

BSD 3-Clause "New" or "Revised" License

3.52k stars 282 forks source link

issues

Newest

Newest Most commented Recently updated Oldest Least commented Least recently updated

update ao to latest patch

#1132 ebsmothers opened 13 hours ago
1
Pytorch main

#1131 ekinakyurek closed 1 day ago
3
Optimize attention mask creation for sample packing.

#1130 joecummings opened 1 day ago
0
Updated CausalAttention doc string

#1129 pbontrager closed 18 hours ago
1
Quantization for Llama-70b raises CUDA OOM

#1128 lulmer opened 1 day ago
2
[WIP][CLIP ENCODER] Vision Transform for Clip encoder

#1127 felipemello1 opened 2 days ago
1
add metadata format to save_file in `FullModelHFCheckpointer`

#1126 man-shar closed 1 day ago
5
Pin ao==0.3

#1125 ebsmothers closed 2 days ago
3
How to finetune a model for sequence classification tasks using the dataset like stanfordnlp/imdb from huggingface?

#1124 JonasQN opened 2 days ago
1
Improve dataset documentation

#1123 RdoubleA opened 2 days ago
2
Missing lm_head.weight key when using Gemma 7B distributed LoRA recipe with gemma-7b-it

#1122 aubreyjstrier opened 2 days ago
1
Dynamic Batch size

#1121 pbontrager opened 3 days ago
1
Support for Phi-3-mini-128k-instruct and larger context length models

#1120 dcsuka opened 3 days ago
0
NF4 quantization of linear layers without LoRA applied

#1119 winglian opened 3 days ago
2
Add in special tokens for Gemma and Mistral in tokenizer

#1118 joecummings opened 3 days ago
0
[WIP] Multimodal dataset + template transforms

#1117 RdoubleA closed 2 days ago
1
HF to Tune wrapper

#1116 ScottHoang opened 4 days ago
2
DoRA

#1115 calvinpelletier opened 4 days ago
1
Pin setuptools version

#1114 ebsmothers closed 4 days ago
5
fix typos

#1113 mdeff closed 4 days ago
3
CUDA illegal memory access exception

#1112 l3utterfly opened 5 days ago
5
ImportError: cannot import name 'packaging' from 'pkg_resources'

#1111 JadarTheObscurity closed 4 days ago
3
Missing non-LoRA key tok_embeddings.weight from base model dict

#1110 vasicvuk opened 6 days ago
2
Add 'on-the-fly' sample packing

#1109 joecummings closed 1 day ago
5
Save intermediate checkpoints during training

#1107 l3utterfly opened 1 week ago
4
[WIP] [DO NOT LAND] Everything is transform + merge instruct/chat

#1106 RdoubleA opened 1 week ago
1
[doc] Add QAT tutorial

#1105 andrewor14 opened 1 week ago
3
Was Pure bfloat16 or MixedPrecision bfloat16 Used for LLama3 pre-training?

#1104 jasonkrone closed 5 days ago
1
Distributed training on a subset of GPU does not work

#1103 lulmer closed 1 week ago
4
Llama3 finetune error: ValueError(f"Invalid {class_type} class: '{component_name}'") from None

#1102 l3utterfly closed 1 week ago
5
LORA and merged lora weights discrepancy

#1101 Optimox closed 1 week ago
6
want dora and nef-tune supports!

#1100 jeffchy opened 1 week ago
3
How to provide our own daaset.json or csv file to the llama2:7B finetune using tune run

#1099 himanshushukla12 opened 1 week ago
13
Run unit tests on GPUs in CI

#1098 ebsmothers opened 1 week ago
1
Investigate possible memory leak with sample packing

#1097 joecummings closed 1 day ago
5
Add safe-serialization to FullModelHFCheckpointer

#1096 jeffrey-fong closed 1 week ago
6
Pin ``numpy<=1.26.4``

#1095 joecummings closed 1 week ago
1
Fix the Gemma generation

#1094 solitude-alive closed 1 week ago
7
Support NF4 quantization of linear layers without LoRA applied

#1093 ebsmothers opened 2 weeks ago
6
High memory usage on Llama3-70B full finetune during checkpoint save

#1092 ebsmothers opened 2 weeks ago
1
Update API ref for torchtune/data

#1091 joecummings closed 2 weeks ago
5
Profiler v2

#1089 jeromeku closed 3 days ago
13
fix torchtune import from torchao after refactor

#1088 jerryzh168 closed 2 weeks ago
9
Fix precision + QLoRA state dict tests, DTensor init

#1087 ebsmothers closed 1 week ago
2
fix typos

#1086 mdeff closed 2 weeks ago
6
[doc] update to the models documentation

#1085 calvinpelletier closed 2 weeks ago
4
[CLIP][IMAGE TRANSFORMS] Image transforms for clip encoder

#1084 felipemello1 opened 2 weeks ago
1
[WIP] Sample packing deep dive

#1083 RdoubleA opened 2 weeks ago
2
Tokenizer redesign for better model-specific feature support

#1082 RdoubleA opened 2 weeks ago
6
Fix WandBLogger to allow resuming runs with updated config values

#1081 parthsarthi03 closed 2 weeks ago
6