Hugging Face's logo

Efficient-Large-Model
/
NVILA-AWQ

Model card Files Community
2
Louym commited on
Commit
1b41d63
·
1 Parent(s): 6fae2d2

[Minor] Add NVILA-Lite-15B

Browse files
Files changed (6) hide show
  1. NVILA-15B-smooth-scale.pt → NVILA-15B-VT-smooth-scale.pt +0 -0
  2. NVILA-8B-smooth-scale.pt → NVILA-8B-VT-smooth-scale.pt +0 -0
  3. NVILA-Lite-8B-stage2-VT-smooth-scale.pt → NVILA-Lite-15B-VT-smooth-scale.pt +2 -2
  4. NVILA-Lite-8B-stage2-w4-g128-awq-v2.pt → NVILA-Lite-15B-w4-g128-awq-v2.pt +2 -2
  5. NVILA-Lite-8B-smooth-scale.pt → NVILA-Lite-8B-VT-smooth-scale.pt +0 -0
  6. README.md +3 -2
NVILA-15B-smooth-scale.pt → NVILA-15B-VT-smooth-scale.pt RENAMED
File without changes
NVILA-8B-smooth-scale.pt → NVILA-8B-VT-smooth-scale.pt RENAMED
File without changes
NVILA-Lite-8B-stage2-VT-smooth-scale.pt → NVILA-Lite-15B-VT-smooth-scale.pt RENAMED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:32037536b7ccd2f2f9dfa52f8b3fbfe0ce634fc5087f848930af9e82031eb9ab
3
- size 1145404
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fddaee3f268310c28655e3b091010267fa782d918cb75233a56a9ae7e0e64ef0
3
+ size 1144408
NVILA-Lite-8B-stage2-w4-g128-awq-v2.pt → NVILA-Lite-15B-w4-g128-awq-v2.pt RENAMED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:68d035f2f577e9c425dfa5ea53080d92c0553e7a2c65806f391d3bc171974b50
3
- size 5663776696
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:854faff6cceb95da75e650c7e3da7f416f5206150bb1ce4b6fc02b097d113edd
3
+ size 10131195405
NVILA-Lite-8B-smooth-scale.pt → NVILA-Lite-8B-VT-smooth-scale.pt RENAMED
File without changes
README.md CHANGED
@@ -9,8 +9,9 @@ One-command demo to chat with quantized NVILA models via [llm-awq](https://githu
9
  '''
10
  cd llm-awq/tinychat
11
  python nvila_demo.py --model-path PATH/TO/NVILA \
12
- --quant_path NVILA-AWQ/NVILA-8B-w4-g128-awq-v2.pt \
13
  --media PATH/TO/ANY/IMAGES/VIDEOS \
14
- --act_scale_path NVILA-AWQ/NVILA-8B-smooth-scale.pt \
15
  --all --chunk --model_type nvila
16
  '''
 
 
9
  '''
10
  cd llm-awq/tinychat
11
  python nvila_demo.py --model-path PATH/TO/NVILA \
12
+ --quant_path NVILA-8B-w4-g128-awq-v2.pt \
13
  --media PATH/TO/ANY/IMAGES/VIDEOS \
14
+ --act_scale_path NVILA-8B-VT-smooth-scale.pt \
15
  --all --chunk --model_type nvila
16
  '''
17
+ This command will download the quantized NVILA model and run a chat demo. If you’ve already downloaded the files, simply set the path to your local copies.