Update README.md
Browse files
README.md
CHANGED
@@ -72,7 +72,16 @@ The following two different evaluations are performed.
|
|
72 |
Perplexity (PPL) is a metric used to evaluate the performance of language models. It measures how well a probability distribution or a language model predicts a sample. A **lower perplexity** score indicates better performance (i.e., the model is more confident in its predictions).
|
73 |
|
74 |
|
|
|
75 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
76 |
```python
|
77 |
from evaluate import load
|
78 |
import datasets
|
@@ -96,13 +105,6 @@ print(round(results["mean_perplexity"], 2))
|
|
96 |
|
97 |
|
98 |
|
99 |
-
#### Main Results
|
100 |
-
|
101 |
-
| Model | Perplexity Score |
|
102 |
-
|---------------------------------------------|----------|
|
103 |
-
| **Llama-3.1-8B-Instruct** | 842611366.59 |
|
104 |
-
| **Llama-3.1-10B-Instruct** | 2890.31 |
|
105 |
-
|
106 |
|
107 |
### Harness Evaluation
|
108 |
|
@@ -117,7 +119,7 @@ The library used is [lm-evaluation-harness repository](https://github.com/Eleuth
|
|
117 |
| **Llama-3.1-8B-Instruct** | **73** | **71.1** | **87.9** |
|
118 |
|
119 |
|
120 |
-
|
121 |
|
122 |
```python
|
123 |
# install from https://github.com/EleutherAI/lm-evaluation-harness
|
@@ -130,8 +132,6 @@ tasks_list = ["arc_challenge", "gpqa", "ifeval", "mmlu_pro", "hellaswag"] # Ben
|
|
130 |
model_path='rwmasood/llama-3.1-10b-instruct'
|
131 |
model_name_or_path = "./output/checkpoint-2800"
|
132 |
|
133 |
-
```
|
134 |
-
|
135 |
# Run evaluation
|
136 |
results = evaluator.simple_evaluate(
|
137 |
model="hf", # Hugging Face model
|
|
|
72 |
Perplexity (PPL) is a metric used to evaluate the performance of language models. It measures how well a probability distribution or a language model predicts a sample. A **lower perplexity** score indicates better performance (i.e., the model is more confident in its predictions).
|
73 |
|
74 |
|
75 |
+
#### Main Results
|
76 |
|
77 |
+
| Model | Perplexity Score |
|
78 |
+
|---------------------------------------------|----------|
|
79 |
+
| **Llama-3.1-8B-Instruct** | 842611366.59 |
|
80 |
+
| **Llama-3.1-10B-Instruct** | 2890.31 |
|
81 |
+
|
82 |
+
|
83 |
+
|
84 |
+
#### Scripts to generate evalution results
|
85 |
```python
|
86 |
from evaluate import load
|
87 |
import datasets
|
|
|
105 |
|
106 |
|
107 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
108 |
|
109 |
### Harness Evaluation
|
110 |
|
|
|
119 |
| **Llama-3.1-8B-Instruct** | **73** | **71.1** | **87.9** |
|
120 |
|
121 |
|
122 |
+
#### Scripts to generate evalution results
|
123 |
|
124 |
```python
|
125 |
# install from https://github.com/EleutherAI/lm-evaluation-harness
|
|
|
132 |
model_path='rwmasood/llama-3.1-10b-instruct'
|
133 |
model_name_or_path = "./output/checkpoint-2800"
|
134 |
|
|
|
|
|
135 |
# Run evaluation
|
136 |
results = evaluator.simple_evaluate(
|
137 |
model="hf", # Hugging Face model
|