Update README.md
Browse files
README.md
CHANGED
@@ -1,14 +1,16 @@
|
|
1 |
## Performance
|
2 |
|
|
|
|
|
3 |
### Low-level Question-Answering
|
4 |
|
5 |
-
This model has reached 75.
|
6 |
|
7 |
It also outperforms the following close-source models with much larger model capacities:
|
8 |
|
9 |
| Model | *dev* | *test* |
|
10 |
| ---- | ---- | ---- |
|
11 |
-
| **Co-Instruct-Preview (mPLUG-Owl2) (This Model)** | **75.
|
12 |
| \*GPT-4V-Turbo | 74.41\% | 74.10\% |
|
13 |
| \*Qwen-VL-**Max** | 73.63\% | 73.90\% |
|
14 |
| \*GPT-4V (Nov. 2023) | 71.78\% | 73.44\% |
|
@@ -23,8 +25,8 @@ It also outperforms the following close-source models with much larger model cap
|
|
23 |
|
24 |
| Model | live | agi | livec | test_spaq | csiq | test_kadid | test_koniq | konvid | maxwell_test |
|
25 |
|--------------------------|--------------|--------------|-------------|-------------|-------------|-------------|-------------|-------------|--------------|
|
26 |
-
|**Co-Instruct-Preview (mPLUG-Owl2) (This Model)** | **0.
|
27 |
-
| Q-Instruct (mPLUG-Owl2, Nov. 2023) | 0.749/0.747 | 0.710
|
28 |
|
29 |
|
30 |
We are also constructing multi-image benchmark sets (image pairs, triple-quadruple images), and the results on multi-image benchmarks will be released soon!
|
@@ -38,7 +40,7 @@ from transformers import AutoModelForCausalLM
|
|
38 |
model = AutoModelForCausalLM.from_pretrained("q-future/co-instruct-preview",
|
39 |
trust_remote_code=True,
|
40 |
torch_dtype=torch.float16,
|
41 |
-
attn_implementation="
|
42 |
device_map={"":"cuda:0"})
|
43 |
```
|
44 |
|
|
|
1 |
## Performance
|
2 |
|
3 |
+
*Updated Feb 1st.*
|
4 |
+
|
5 |
### Low-level Question-Answering
|
6 |
|
7 |
+
This model has reached 75.90\%(*13\% better than previous version*)/76.52\%(*10\% better than previous version*) on Q-Bench A1 *dev/test* (multi-choice questions).
|
8 |
|
9 |
It also outperforms the following close-source models with much larger model capacities:
|
10 |
|
11 |
| Model | *dev* | *test* |
|
12 |
| ---- | ---- | ---- |
|
13 |
+
| **Co-Instruct-Preview (mPLUG-Owl2) (This Model)** | **75.90\%** | **76.52\%** |
|
14 |
| \*GPT-4V-Turbo | 74.41\% | 74.10\% |
|
15 |
| \*Qwen-VL-**Max** | 73.63\% | 73.90\% |
|
16 |
| \*GPT-4V (Nov. 2023) | 71.78\% | 73.44\% |
|
|
|
25 |
|
26 |
| Model | live | agi | livec | test_spaq | csiq | test_kadid | test_koniq | konvid | maxwell_test |
|
27 |
|--------------------------|--------------|--------------|-------------|-------------|-------------|-------------|-------------|-------------|--------------|
|
28 |
+
|**Co-Instruct-Preview (mPLUG-Owl2) (This Model)** | **0.803/0.756** | **0.719**/0.732 | **0.827/0.835** | **0.946/0.937** | **0.711/0.727** | **0.782/0.766** | 0.886/**0.935** | **0.818/0.790** | **0.735/0.714** |
|
29 |
+
| Q-Instruct (mPLUG-Owl2, Nov. 2023) | 0.749/0.747 | 0.710/**0.753** | 0.781/0.791 | 0.921/0.917 | 0.693/0.723 | 0.670/0.665 | **0.904**/0.921 | 0.766/0.738 | 0.650/0.649 |
|
30 |
|
31 |
|
32 |
We are also constructing multi-image benchmark sets (image pairs, triple-quadruple images), and the results on multi-image benchmarks will be released soon!
|
|
|
40 |
model = AutoModelForCausalLM.from_pretrained("q-future/co-instruct-preview",
|
41 |
trust_remote_code=True,
|
42 |
torch_dtype=torch.float16,
|
43 |
+
attn_implementation="eager",
|
44 |
device_map={"":"cuda:0"})
|
45 |
```
|
46 |
|