chaoscodes
commited on
Commit
•
37c43e7
1
Parent(s):
77b87b3
Update README.md
Browse files
README.md
CHANGED
@@ -59,25 +59,13 @@ Here we list our data distribution in each stage:
|
|
59 |
|
60 |
| Corpus | Basic pretraining | Continual pretraining with specific domain | Cooldown |
|
61 |
| ------------- | ----------------- | ------------------------------------------ | -------- |
|
62 |
-
|
|
63 |
-
| C4 | 35.0 | 35.0 | 35.0 |
|
64 |
-
| CommonCrawl | 70.1 | 70.1 | 70.1 |
|
65 |
-
| Github | 6.5 | 6.5 | 6.5 |
|
66 |
-
| StackExchange | 4.2 | 4.2 | 4.2 |
|
67 |
-
| ArXiv | 5.7 | 5.7 | 5.7 |
|
68 |
-
| Wikipedia | 4.5 | 4.5 | 4.5 |
|
69 |
|
70 |
### TinyLlama_v1.1_math_code
|
71 |
|
72 |
| Corpus | Basic pretraining | Continual pretraining with specific domain | Cooldown |
|
73 |
| ------------- | ----------------- | ------------------------------------------ | -------- |
|
74 |
-
|
|
75 |
-
| C4 | 35.0 | 21.6 | 21.6 |
|
76 |
-
| CommonCrawl | 70.1 | 43.0 | 43.0 |
|
77 |
-
| Github | 6.5 | - | - |
|
78 |
-
| StackExchange | 4.2 | 2.6 | 2.6 |
|
79 |
-
| ArXiv | 5.7 | 5.0 | 5.0 |
|
80 |
-
| Wikipedia | 4.5 | 2.8 | 2.8 |
|
81 |
| starcoder | - | 15.0 | 15.0 |
|
82 |
| proof_pile | - | 10.0 | 10.0 |
|
83 |
|
@@ -85,13 +73,7 @@ Here we list our data distribution in each stage:
|
|
85 |
|
86 |
| orpus | Basic pretraining | Continual pretraining with specific domain | Cooldown |
|
87 |
| ------------- | ----------------- | ------------------------------------------ | -------- |
|
88 |
-
|
|
89 |
-
| C4 | 35.0 | 14.6 | 14.6 |
|
90 |
-
| CommonCrawl | 70.1 | 29.3 | 29.3 |
|
91 |
-
| Github | 6.5 | - | - |
|
92 |
-
| StackExchange | 4.2 | 1.8 | 1.8 |
|
93 |
-
| ArXiv | 5.7 | 2.4 | 2.4 |
|
94 |
-
| Wikipedia | 4.5 | 1.9 | 1.9 |
|
95 |
| skypile | - | 50.0 | 50.0 |
|
96 |
|
97 |
### How to use
|
|
|
59 |
|
60 |
| Corpus | Basic pretraining | Continual pretraining with specific domain | Cooldown |
|
61 |
| ------------- | ----------------- | ------------------------------------------ | -------- |
|
62 |
+
| Slimpajama | 100.0 | 100.0 | 100.0 |
|
|
|
|
|
|
|
|
|
|
|
|
|
63 |
|
64 |
### TinyLlama_v1.1_math_code
|
65 |
|
66 |
| Corpus | Basic pretraining | Continual pretraining with specific domain | Cooldown |
|
67 |
| ------------- | ----------------- | ------------------------------------------ | -------- |
|
68 |
+
| Slimpajama | 100.0 | 75.0 | 75.0 |
|
|
|
|
|
|
|
|
|
|
|
|
|
69 |
| starcoder | - | 15.0 | 15.0 |
|
70 |
| proof_pile | - | 10.0 | 10.0 |
|
71 |
|
|
|
73 |
|
74 |
| orpus | Basic pretraining | Continual pretraining with specific domain | Cooldown |
|
75 |
| ------------- | ----------------- | ------------------------------------------ | -------- |
|
76 |
+
| Slimpajama | 100.0 | 50.0 | 50.0 |
|
|
|
|
|
|
|
|
|
|
|
|
|
77 |
| skypile | - | 50.0 | 50.0 |
|
78 |
|
79 |
### How to use
|