Commit History

add gradient checkpointing for the final_layernorm module.
3d854f8

zhaoqf123 commited on

Update model license
1d240ba

zxdu20 commited on

Update change log
a10da4c

zxdu20 commited on

Upload 8 files
942945d

zxdu20 commited on

Update slack link
55cced3

zxdu20 commited on

Update decode method in tokenizer
4d0fc39

zxdu20 commited on

Add version
658202d

zxdu20 commited on

Fix position ids in 1d position encoding
a8ede82

zxdu20 commited on

Add test for modeling_chatglm
f831824

zxdu20 commited on

Fix input embeds
35ca523

zxdu20 commited on

Update slack link
0829959

zxdu20 commited on

Change mask positions to batch
4de8efe

zxdu20 commited on

Always add gmask in token ids
3a99d79

zxdu20 commited on

Fix bug
53f0197

zxdu20 commited on

Add empty_init option
eb55ff0

zxdu20 commited on

Update README
9692905

zxdu20 commited on

Fix eos token in tokenizer
aa51e62

zxdu20 commited on

Fix attention score on mps
cde457b

zxdu20 commited on

Update dependency
acd41f7

zxdu20 commited on

Merge branch 'main' of https://huggingface.co/THUDM/chatglm-6b
6650ae3

zxdu20 commited on

Fix tokenizer config saving
7e69b85

zxdu20 commited on

Fix LogitsProcessor using slim checkpoint (#29)
61eee50

zxdu20 bcol commited on

Use gmask in first place
9324de7

zxdu20 commited on

Update slim checkpoint (#28)
d467eff

zxdu20 commited on

Merge branch 'slim' of https://huggingface.co/THUDM/chatglm-6b into slim
06a22a3

zxdu20 commited on

Add gmask token id
36b7f2d

zxdu20 commited on

Update slim checkpoint
6461061

zxdu20 commited on

Update code for slim
63ce1ba

zxdu20 commited on

Drop icetk dependency
72985e8

zxdu20 commited on

Fix decode method for torch tensor
23ad39b

zxdu20 commited on

Support single integer or empty list as input to decode (#7)
fdb7a60

zxdu20 peakji commited on

Fix position ids expand
f82b180

zxdu20 commited on

Fix generate
fb23542

zxdu20 commited on

Fix attention mask for prefix prompt
08bc851

zxdu20 commited on

No padding for chat function
4b7ffbf

zxdu20 commited on

Fix attention_mask and position_ids
373fd6b

zxdu20 commited on

Fix encode method
e22cddf

zxdu20 commited on

Fix batch input
e1494f2

zxdu20 commited on

Implement batch generation
cc96a22

zxdu20 commited on

Fix position id for training
11c270c

zxdu20 commited on

Add support for loading quantized model
2e1be30

zxdu20 commited on

Use dynamic dtype for prompts
c949d03

zxdu20 commited on

Fix backward for quantization
0cfae21

zxdu20 commited on

Implement gradient checkpointing
aea6cef

zxdu20 commited on

Fix bugs
0564795

zxdu20 commited on

Add pad_token_id in config.json
2200e2b

zxdu20 commited on

Change padding side
db22499

zxdu20 commited on

Set ignore_index for CrossEntropyLoss
5c64357

zxdu20 commited on