Remove delete inputs_embeds in _update_model_kwargs_for_generation
238dc29
duzx16
commited on
Commit History
Add warning
29868f4
duzx16
commited on
Merge branch 'main' into pr/45
34a7f82
duzx16
commited on
Fix input embeds
35ca523
duzx16
commited on
Update modeling_chatglm.py for inputs_embeds
3118bcb
Change mask positions to batch
4de8efe
duzx16
commited on
Add empty_init option
eb55ff0
duzx16
commited on
Fix attention score on mps
cde457b
duzx16
commited on
Use gmask in first place
9324de7
duzx16
commited on
Update code for slim
63ce1ba
duzx16
commited on
Fix position ids expand
f82b180
duzx16
commited on
Fix generate
fb23542
duzx16
commited on
Fix attention mask for prefix prompt
08bc851
duzx16
commited on
No padding for chat function
4b7ffbf
duzx16
commited on
Implement batch generation
cc96a22
duzx16
commited on
Fix position id for training
11c270c
duzx16
commited on
Add support for loading quantized model
2e1be30
duzx16
commited on
Use dynamic dtype for prompts
c949d03
duzx16
commited on
Fix backward for quantization
0cfae21
duzx16
commited on
Implement gradient checkpointing
aea6cef
duzx16
commited on
Fix bugs
0564795
duzx16
commited on
Add pad_token_id in config.json
2200e2b
duzx16
commited on
Set ignore_index for CrossEntropyLoss
5c64357
duzx16
commited on
Support batch training
8127ab6
duzx16
commited on
Merge branch 'main' into dev_pt
fbda120
duzx16
commited on
Add p-tuning v2
812f43f
duzx16
commited on
Fix context length in get_position_ids
096f3de
duzx16
commited on
Close CPU fusion on Mac
4a9b711
duzx16
commited on
Fix Chinese punctuation
d2bbc82
duzx16
commited on
Remove hardcode bos_token_id
2460dc2
duzx16
commited on
Add support for streaming output
42095d4
duzx16
commited on
Fix overflow in FP16
220f772
duzx16
commited on
Set is_parallelizable to False
f9f74fd
duzx16
commited on
Add logit processor for NaN or Inf scores
c3dece3
duzx16
commited on
Fix default history argument
9d1509a
duzx16
commited on
Add support for float32
d4832e8
duzx16
commited on
Fix past_key_values
cd8041e
duzx16
commited on
