hlnchen commited on
Commit
3986381
·
verified ·
1 Parent(s): 7891534

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +30 -5
README.md CHANGED
@@ -10,23 +10,26 @@ tags:
10
  ---
11
  # CoDA-v0-Base
12
 
13
- ## Overview
14
- CoDA is Salesforce AI Research's open, lightweight and diffusion-based language model.
 
15
 
16
  [Technical Report](https://github.com/SalesforceAIResearch/CoDA/blob/main/technical_report.pdf)
17
 
18
  [Code](https://github.com/SalesforceAIResearch/CoDA/)
19
 
 
 
20
 
21
 
22
- ## Requirements
23
  ```
24
  torch==2.8.0
25
  transformers>=4.47.1
26
  flash-attn==2.8.3
27
  ```
28
 
29
- ## Quickstart
30
  Here is a code snippet for loading the model, tokenizer and run unmasking for a partially finished code.
31
  ```python
32
  import torch
@@ -81,4 +84,26 @@ unmasked_output = tokenizer.batch_decode(generated_ids, skip_special_tokens=True
81
 
82
  ```
83
 
84
- ## Citation
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
10
  ---
11
  # CoDA-v0-Base
12
 
13
+ ## Overview 🎯
14
+ CoDA is Salesforce AI Research's open diffusion language model.
15
+
16
 
17
  [Technical Report](https://github.com/SalesforceAIResearch/CoDA/blob/main/technical_report.pdf)
18
 
19
  [Code](https://github.com/SalesforceAIResearch/CoDA/)
20
 
21
+ The code repo contains a unified training pipeline from pre-training to post-training, evaluation harnesses, and a simple Fast-API based serving backend.
22
+
23
 
24
 
25
+ ## Requirements 📦
26
  ```
27
  torch==2.8.0
28
  transformers>=4.47.1
29
  flash-attn==2.8.3
30
  ```
31
 
32
+ ## Quickstart 🚀
33
  Here is a code snippet for loading the model, tokenizer and run unmasking for a partially finished code.
34
  ```python
35
  import torch
 
84
 
85
  ```
86
 
87
+ ## Benchmark 📊
88
+ Comparison of code-generation performance across standard and plus-enhanced benchmarks. Evalplus is computed as the mean pass@1 on enhanced variants. Bold marks results where CoDA produces the strongest diffusion-model performance.
89
+
90
+ | Model | Humaneval Instruct | Humaneval Plus | MBPP Instruct | MBPP Plus | Evalplus |
91
+ | --- | --- | --- | --- | --- | --- |
92
+ | CoDA-Base | 29.3 | 23.8 | 35.2 | 46.0 | 34.9 |
93
+ | CoDA-Instruct | 54.3 | 47.6 | 47.2 | **63.2** | **55.4** |
94
+ | Dream-Base | 56.7 | 50.0 | 68.7 | 57.4 | 53.7 |
95
+ | Dream-7B-Instruct | 57.9 | 53.7 | 68.3 | 56.1 | 54.9 |
96
+ | LLaDA-8B-Instruct | 35.4 | 31.7 | 31.5 | 28.6 | 30.2 |
97
+ | Qwen3-1.7B | 66.5 | 61.6 | 46.2 | 65.9 | 63.8 |
98
+ | Qwen2.5-Coder-1.5B | 43.9 | 36.6 | 69.2 | 58.6 | 47.6 |
99
+ | Qwen2.5-Coder-1.5B-Instruct | 70.7 | 66.5 | 69.2 | 59.4 | 62.3 |
100
+ | Gemma-3-1B-it | 39.6 | 35.4 | 39.4 | 63.5 | 49.5 |
101
+ | LLaMA-3.2-1B-Instruct | 35.4 | 31.1 | 24.4 | 53.7 | 42.4 |
102
+
103
+ ## Deployment 🛠️
104
+ Checkout our [Deployment Guide](https://github.com/SalesforceAIResearch/CoDA?tab=readme-ov-file#deployment-guide-%EF%B8%8F)!
105
+
106
+ ## Citation 📚
107
+ ```
108
+ coming soon
109
+ ```