Text Generation
Transformers
PyTorch
English
gpt2a
custom_code
crumb commited on
Commit
2b64783
1 Parent(s): b96b864

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +75 -1
README.md CHANGED
@@ -10,6 +10,8 @@ language:
10
 
11
  A modified GPT-2 model with only 25 million non-embedding params that outbenches GPT-2(124m), Pythia-70m/160m, and Cerebras-111m, it has ScaledSinusoidal position embeddings, embedding layernorm, no biases, and was trained on only 8 billion tokens of the SlimPajama dataset at home on 2xA6000. (On the graphic it's mis-labeled as cramp-41m)
12
 
 
 
13
  | model | avg | arc | hellaswag | mmlu | truthfulqa |
14
  | --- | --- | --- | --- | --- | --- |
15
  | cramp-25m | 30.57 | 21.76 | 27.35 | 25.53 | 47.66 |
@@ -17,8 +19,80 @@ A modified GPT-2 model with only 25 million non-embedding params that outbenches
17
  | pythia 70m deduped | 30.25 | 21.08 | 27.17 | 25.26 | 47.51 |
18
  | pythia 70m | 30.46 | 21.59 | 27.29 | 25.9 | 47.06 |
19
  | pythia 160m deduped | 31.16 | 24.06 | 30.34 | 24.95 | 44.34 |
20
- | pythia 160m | 30.58 | 22.78 | 30.34 | 24.95 | 44.26 |
 
 
 
 
 
 
 
 
 
 
 
 
 
21
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
22
 
23
  ![image/png](https://cdn-uploads.huggingface.co/production/uploads/6079949388160e14e4e2e499/NzTdlxtBDp4drBRZgJiXt.png)
24
 
 
10
 
11
  A modified GPT-2 model with only 25 million non-embedding params that outbenches GPT-2(124m), Pythia-70m/160m, and Cerebras-111m, it has ScaledSinusoidal position embeddings, embedding layernorm, no biases, and was trained on only 8 billion tokens of the SlimPajama dataset at home on 2xA6000. (On the graphic it's mis-labeled as cramp-41m)
12
 
13
+
14
+ **OLD BENCHMARK**
15
  | model | avg | arc | hellaswag | mmlu | truthfulqa |
16
  | --- | --- | --- | --- | --- | --- |
17
  | cramp-25m | 30.57 | 21.76 | 27.35 | 25.53 | 47.66 |
 
19
  | pythia 70m deduped | 30.25 | 21.08 | 27.17 | 25.26 | 47.51 |
20
  | pythia 70m | 30.46 | 21.59 | 27.29 | 25.9 | 47.06 |
21
  | pythia 160m deduped | 31.16 | 24.06 | 30.34 | 24.95 | 44.34 |
22
+ | pythia 160m | 30.58 | 22.78 | 30.34 | 24.95 | 44.26 |
23
+
24
+ ***NEW BENCHMARK**
25
+
26
+ | Tasks |Version|Filter|n-shot| Metric |Value | |Stderr|
27
+ |-------------|------:|------|-----:|--------|-----:|---|-----:|
28
+ |arc_challenge| 1|none | 25|acc |0.1724|± |0.0110|
29
+ | | |none | 25|acc_norm|0.2031|± |0.0118|
30
+ |truthfulqa_mc2| 2|none | 0|acc |0.4767|± |0.0156|
31
+ |hellaswag| 1|none | 10|acc |0.2687|± |0.0044|
32
+ | | |none | 10|acc_norm|0.2773|± |0.0045|
33
+ |winogrande| 1|none | 5|acc |0.5028|± |0.0141|
34
+
35
+ *MMLU*
36
 
37
+ | Tasks |Version|Filter|n-shot|Metric|Value | |Stderr|
38
+ |-----------------------------------|------:|------|-----:|------|-----:|---|-----:|
39
+ |world_religions | 0|none | 5|acc |0.1813|± |0.0295|
40
+ |virology | 0|none | 5|acc |0.1928|± |0.0307|
41
+ |us_foreign_policy | 0|none | 5|acc |0.2900|± |0.0456|
42
+ |sociology | 0|none | 5|acc |0.2438|± |0.0304|
43
+ |security_studies | 0|none | 5|acc |0.2367|± |0.0272|
44
+ |public_relations | 0|none | 5|acc |0.2455|± |0.0412|
45
+ |professional_psychology | 0|none | 5|acc |0.2271|± |0.0169|
46
+ |professional_medicine | 0|none | 5|acc |0.4375|± |0.0301|
47
+ |professional_law | 0|none | 5|acc |0.2490|± |0.0110|
48
+ |professional_accounting | 0|none | 5|acc |0.2589|± |0.0261|
49
+ |prehistory | 0|none | 5|acc |0.2963|± |0.0254|
50
+ |philosophy | 0|none | 5|acc |0.2315|± |0.0240|
51
+ |nutrition | 0|none | 5|acc |0.2222|± |0.0238|
52
+ |moral_scenarios | 0|none | 5|acc |0.2313|± |0.0141|
53
+ |moral_disputes | 0|none | 5|acc |0.2168|± |0.0222|
54
+ |miscellaneous | 0|none | 5|acc |0.2708|± |0.0159|
55
+ |medical_genetics | 0|none | 5|acc |0.3000|± |0.0461|
56
+ |marketing | 0|none | 5|acc |0.1923|± |0.0258|
57
+ |management | 0|none | 5|acc |0.1942|± |0.0392|
58
+ |machine_learning | 0|none | 5|acc |0.2054|± |0.0383|
59
+ |logical_fallacies | 0|none | 5|acc |0.2393|± |0.0335|
60
+ |jurisprudence | 0|none | 5|acc |0.2130|± |0.0396|
61
+ |international_law | 0|none | 5|acc |0.2562|± |0.0398|
62
+ |human_sexuality | 0|none | 5|acc |0.2366|± |0.0373|
63
+ |human_aging | 0|none | 5|acc |0.2063|± |0.0272|
64
+ |high_school_world_history | 0|none | 5|acc |0.2700|± |0.0289|
65
+ |high_school_us_history | 0|none | 5|acc |0.2206|± |0.0291|
66
+ |high_school_statistics | 0|none | 5|acc |0.4722|± |0.0340|
67
+ |high_school_psychology | 0|none | 5|acc |0.2257|± |0.0179|
68
+ |high_school_physics | 0|none | 5|acc |0.2384|± |0.0348|
69
+ |high_school_microeconomics | 0|none | 5|acc |0.3403|± |0.0308|
70
+ |high_school_mathematics | 0|none | 5|acc |0.2630|± |0.0268|
71
+ |high_school_macroeconomics | 0|none | 5|acc |0.2051|± |0.0205|
72
+ |high_school_government_and_politics| 0|none | 5|acc |0.2280|± |0.0303|
73
+ |high_school_geography | 0|none | 5|acc |0.3535|± |0.0341|
74
+ |high_school_european_history | 0|none | 5|acc |0.2909|± |0.0355|
75
+ |high_school_computer_science | 0|none | 5|acc |0.2400|± |0.0429|
76
+ |high_school_chemistry | 0|none | 5|acc |0.2759|± |0.0314|
77
+ |high_school_biology | 0|none | 5|acc |0.3161|± |0.0265|
78
+ |global_facts | 0|none | 5|acc |0.2000|± |0.0402|
79
+ |formal_logic | 0|none | 5|acc |0.1825|± |0.0346|
80
+ |elementary_mathematics | 0|none | 5|acc |0.2566|± |0.0225|
81
+ |electrical_engineering | 0|none | 5|acc |0.2414|± |0.0357|
82
+ |econometrics | 0|none | 5|acc |0.2544|± |0.0410|
83
+ |conceptual_physics | 0|none | 5|acc |0.2809|± |0.0294|
84
+ |computer_security | 0|none | 5|acc |0.2000|± |0.0402|
85
+ |college_physics | 0|none | 5|acc |0.3431|± |0.0472|
86
+ |college_medicine | 0|none | 5|acc |0.2197|± |0.0316|
87
+ |college_mathematics | 0|none | 5|acc |0.3100|± |0.0465|
88
+ |college_computer_science | 0|none | 5|acc |0.3100|± |0.0465|
89
+ |college_chemistry | 0|none | 5|acc |0.3400|± |0.0476|
90
+ |college_biology | 0|none | 5|acc |0.2083|± |0.0340|
91
+ |clinical_knowledge | 0|none | 5|acc |0.2189|± |0.0254|
92
+ |business_ethics | 0|none | 5|acc |0.2000|± |0.0402|
93
+ |astronomy | 0|none | 5|acc |0.2237|± |0.0339|
94
+ |anatomy | 0|none | 5|acc |0.3333|± |0.0407|
95
+ |abstract_algebra | 0|none | 5|acc |0.2200|± |0.0416|
96
 
97
  ![image/png](https://cdn-uploads.huggingface.co/production/uploads/6079949388160e14e4e2e499/NzTdlxtBDp4drBRZgJiXt.png)
98