潘其威(William)
|
1e353a8dc5
|
Merge pull request #24 from PanQiWei/speedup_quantization
Offloading and Multiple devices quantization/inference
|
2023-04-28 18:50:12 +08:00 |
|
PanQiWei
|
bdb713b5a3
|
add batch_size to model.quant() api
|
2023-04-28 18:26:07 +08:00 |
|
PanQiWei
|
3dfc87bec3
|
return module in .to function
|
2023-04-28 17:20:46 +08:00 |
|
PanQiWei
|
a69a73a22c
|
fix device mismatch when directly using model to inference after quantization
|
2023-04-28 16:41:46 +08:00 |
|
qwopqwop200
|
3f90a22632
|
fix bug
|
2023-04-28 08:26:58 +09:00 |
|
PanQiWei
|
d0cd5af5d3
|
make code more robust
|
2023-04-28 01:29:12 +08:00 |
|
PanQiWei
|
51d2e53130
|
add support to cpu offloading and multi gpus inference on quantized model
|
2023-04-28 00:53:57 +08:00 |
|
PanQiWei
|
b14dca9207
|
disk offload assertion
|
2023-04-27 21:31:53 +08:00 |
|
PanQiWei
|
7a3397e7ba
|
add cpu offload when doing quantization
|
2023-04-27 21:25:24 +08:00 |
|
PanQiWei
|
498de923f2
|
support multi gpus quantization
|
2023-04-27 18:48:43 +08:00 |
|
qwopqwop200
|
8b6ee04aee
|
add option
|
2023-04-27 17:29:36 +09:00 |
|
PanQiWei
|
a2abff983e
|
support dispatch layers to different devices when loading pretrained model before quantization
|
2023-04-27 02:24:08 +08:00 |
|
PanQiWei
|
950f203260
|
add 'n_positions' to sequence length search list
|
2023-04-27 01:09:10 +08:00 |
|
PanQiWei
|
893c3264cb
|
make layer ignorance more robust
|
2023-04-26 19:35:19 +08:00 |
|
PanQiWei
|
f2359f56cb
|
add support to use push_to_hub to upload and share quantized model
|
2023-04-26 16:55:01 +08:00 |
|
PanQiWei
|
975f100d0f
|
init Quantizer() at GPTQ() init stage
|
2023-04-25 23:13:09 +08:00 |
|
PanQiWei
|
062b34f31a
|
add inference_mode and autocast context manager to generate function
|
2023-04-25 20:47:33 +08:00 |
|
PanQiWei
|
31d683f85b
|
add option to choose whether autotune warmup or not after quantization
|
2023-04-25 20:29:05 +08:00 |
|
PanQiWei
|
9c405b1628
|
add triton support
|
2023-04-25 20:05:22 +08:00 |
|
PanQiWei
|
832dc4a7a1
|
refactor file structure
|
2023-04-25 18:58:20 +08:00 |
|
PanQiWei
|
419160b733
|
always trust remote code
|
2023-04-25 12:52:49 +08:00 |
|
PanQiWei
|
f748dad2e1
|
always trust remote code
|
2023-04-25 12:13:46 +08:00 |
|
PanQiWei
|
7d3a625cee
|
fix mismatch GPTNeoxForCausalLM's lm_head
|
2023-04-24 20:51:56 +08:00 |
|
PanQiWei
|
1a8c460262
|
fix problem that some models required more positional arguments in transformer layer's forward function
|
2023-04-24 14:46:21 +08:00 |
|
PanQiWei
|
37c0a80092
|
fix problem that some models can't get seqlen from model.config.max_position_embeddings
|
2023-04-24 14:24:00 +08:00 |
|
PanQiWei
|
4763c0b9a1
|
fix bugs
|
2023-04-23 19:27:16 +08:00 |
|
PanQiWei
|
a830a62bc3
|
fix bugs for attention_mask and position_ids
|
2023-04-20 18:32:21 +08:00 |
|
PanQiWei
|
bcc7e0a051
|
make BaseGPTQForCausalLM as nn.Module, add more shortcut apis and fix some bugs
|
2023-04-17 01:15:30 +08:00 |
|
PanQiWei
|
969ec250ad
|
add shortcut to model.to method
|
2023-04-17 00:34:14 +08:00 |
|
PanQiWei
|
12ae4d024c
|
fix gptj forward and add torch.no_grad context manager
|
2023-04-17 00:15:41 +08:00 |
|
PanQiWei
|
87402e01c9
|
fix some errors at quantized model loading
|
2023-04-16 22:32:13 +08:00 |
|
PanQiWei
|
036a151f10
|
fix save_dir not exist
|
2023-04-16 22:20:54 +08:00 |
|
PanQiWei
|
42b28f032f
|
shortcut model.config
|
2023-04-16 22:12:42 +08:00 |
|
PanQiWei
|
0c4b039925
|
add setup.py
|
2023-04-16 21:06:15 +08:00 |
|
PanQiWei
|
229b61e20e
|
first init
|
2023-04-14 01:09:40 +08:00 |
|