Hugging Face's logo

MediaTek-Research
/
Breeze-7B-Instruct-v0_1

Text Generation
Transformers
PyTorch
Safetensors
Chinese
English
mistral
conversational
text-generation-inference
Model card Files Community
5
YC-Chen commited on
Commit
2359491
·
1 Parent(s): 6979fc8

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +43 -33
README.md CHANGED
@@ -26,28 +26,50 @@ and is comparable with Mistral-7B-Instruct-v0.1 on MMLU and MT-Bench in English.
26
  - **Model type:** Causal decoder-only transformer language model
27
  - **Language:** English and Traditional Chinese (zh-tw)
28
 
29
- ## Performance
30
-
31
- | **[Traditional Chinese Benchmarks]** | TMMLU+ 5-shot (ACC) | DRCD 3-shot (EM) | MT-Bench-tw (Score) |
32
- |-----------------------------------------------------|---------------------|------------------|---------------------|
33
- | MediaTek-Research/Breeze-7B-Base-v0.1 | | | - |
34
- | MediaTek-Research/Breeze-7B-Instruct-v0.1 | | | |
35
- | mistralai/Mistral-7B-v0.1 | | | - |
36
- | mistralai/Mistral-7B-Instruct-v0.1 | | | |
37
- | yentinglin/Taiwan-LLM-7B-v2.1-base | | | - |
38
- | yentinglin/Taiwan-LLM-7B-v2.1-chat | | | |
39
- | yentinglin/Taiwan-LLM-13B-v2.0-base | | | - |
40
- | yentinglin/Taiwan-LLM-13B-v2.0-chat | | | |
41
- | 01-ai/Yi-6B | | | - |
42
- | 01-ai/Yi-6B-Chat | | | |
43
- | 01-ai/Yi-34B | | | - |
44
- | 01-ai/Yi-34B-Chat | | | |
45
- | Qwen/Qwen-7B | | | - |
46
- | Qwen/Qwen-7B-Chat | | | |
47
- | Qwen/Qwen-14B | | | - |
48
- | Qwen/Qwen-14B-Chat | | | |
49
- | gpt-3.5-turbo-0613 | | 76.30 | |
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
50
 
 
51
 
52
  | **[English Benchmarks]** | MMLU (ACC) <br/> 5-shot, log-likelihood, w/o chat template | MT-Bench (Score) <br/>|
53
  |-----------------------------------------------------|---------------------|------------------|
@@ -70,18 +92,6 @@ and is comparable with Mistral-7B-Instruct-v0.1 on MMLU and MT-Bench in English.
70
  | gpt-3.5-turbo-0613 | | |
71
 
72
 
73
- | **[Inference Metrics on Traditional Chinese]** | Speed (char/sec) | Compression Ratio | Max Character Size |
74
- |--------------------------------------------------------------------|-------------------|-------------------|--------------------|
75
- | MediaTek-Research/Breeze-7B-Base-v0.1 | | | | |
76
- | mistralai/Mistral-7B-v0.1 | | | |
77
- | yentinglin/Taiwan-LLM-7B-v2.1-base | | | |
78
- | yentinglin/Taiwan-LLM-13B-v2.0-base | | | |
79
- | 01-ai/Yi-6B | | | |
80
- | 01-ai/Yi-34B | | | |
81
- | Qwen/Qwen-7B | | | |
82
- | Qwen/Qwen-14B | | | |
83
-
84
-
85
  ## Use in Transformers
86
 
87
  First install direct dependencies:
 
26
  - **Model type:** Causal decoder-only transformer language model
27
  - **Language:** English and Traditional Chinese (zh-tw)
28
 
29
+ ## Base Model Performance
30
+
31
+ | Models | TMMLU+ (ACC) | DRCD (EM) | MMLU (ACC) |
32
+ |-----------------------------------------------------|--------------|-----------|------------|
33
+ | | 5 shot | 3 shot | 5 shot |
34
+ | MediaTek-Research/Breeze-7B-Base-v0.1 | | | |
35
+ | mistralai/Mistral-7B-v0.1 | | | |
36
+ | yentinglin/Taiwan-LLM-7B-v2.1-base | | | |
37
+ | yentinglin/Taiwan-LLM-13B-v2.0-base | | | |
38
+ | 01-ai/Yi-6B | | | |
39
+ | 01-ai/Yi-34B | | | |
40
+ | Qwen/Qwen-7B | | | |
41
+ | Qwen/Qwen-14B | | | |
42
+
43
+ ## Inference Performance
44
+
45
+ | Models | Speed (char/sec) | Compression Ratio | Max Character Size |
46
+ |--------------------------------------------------------------------|-------------------|-------------------|--------------------|
47
+ | MediaTek-Research/Breeze-7B-Base-v0.1 | | | | |
48
+ | mistralai/Mistral-7B-v0.1 | | | |
49
+ | yentinglin/Taiwan-LLM-7B-v2.1-base | | | |
50
+ | yentinglin/Taiwan-LLM-13B-v2.0-base | | | |
51
+ | 01-ai/Yi-6B | | | |
52
+ | 01-ai/Yi-34B | | | |
53
+ | Qwen/Qwen-7B | | | |
54
+ | Qwen/Qwen-14B | | | |
55
+
56
+ ## Chat Model Performance
57
+
58
+ | Models | TMMLU+ (ACC) | DRCD (EM) | MT-Bench-tw (Score) | MMLU (ACC) | MT-Bench (Score) |
59
+ |-----------------------------------------------------|--------------|-----------|---------------------|------------|------------------|
60
+ | | 5 shot | 3 shot | 0 shot | 5 shot | 0 shot |
61
+ | MediaTek-Research/Breeze-7B-Instruct-v0.1 | | | | | |
62
+ | mistralai/Mistral-7B-Instruct-v0.1 | | | | | |
63
+ | yentinglin/Taiwan-LLM-7B-v2.1-chat | | | | | |
64
+ | yentinglin/Taiwan-LLM-13B-v2.0-chat | | | | | |
65
+ | 01-ai/Yi-6B-Chat | | | | | |
66
+ | 01-ai/Yi-34B-Chat | | | | | |
67
+ | Qwen/Qwen-7B-Chat | | | | | |
68
+ | Qwen/Qwen-14B-Chat | | | | | |
69
+ | gpt-3.5-turbo-0613 | | 76.30 | | | |
70
+
71
 
72
+ ## dupeca.
73
 
74
  | **[English Benchmarks]** | MMLU (ACC) <br/> 5-shot, log-likelihood, w/o chat template | MT-Bench (Score) <br/>|
75
  |-----------------------------------------------------|---------------------|------------------|
 
92
  | gpt-3.5-turbo-0613 | | |
93
 
94
 
 
 
 
 
 
 
 
 
 
 
 
 
95
  ## Use in Transformers
96
 
97
  First install direct dependencies: