Update README.md
Browse files
README.md
CHANGED
@@ -1,3 +1,96 @@
|
|
1 |
---
|
2 |
license: apache-2.0
|
|
|
|
|
|
|
3 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
---
|
2 |
license: apache-2.0
|
3 |
+
tags:
|
4 |
+
- code
|
5 |
+
- mistral
|
6 |
---
|
7 |
+
|
8 |
+
# Mistral-7B-codealpaca
|
9 |
+
|
10 |
+
We are thrilled to introduce the Mistral-7B-codealpaca-test14 model. This variant is optimized and demonstrates potential in assisting developers as a coding companion. We welcome contributions from testers and enthusiasts to help evaluate its performance.
|
11 |
+
|
12 |
+
## Training Details
|
13 |
+
|
14 |
+
The model was trained using 3xRTX 3090 in a homelab setup.
|
15 |
+
[![Built with Axolotl](https://raw.githubusercontent.com/OpenAccess-AI-Collective/axolotl/main/image/axolotl-badge-web.png)](https://github.com/OpenAccess-AI-Collective/axolotl)
|
16 |
+
|
17 |
+
|
18 |
+
## Quantised Model Links:
|
19 |
+
|
20 |
+
1.
|
21 |
+
2.
|
22 |
+
3.
|
23 |
+
|
24 |
+
## Dataset:
|
25 |
+
|
26 |
+
- Dataset Name: theblackcat102/evol-codealpaca-v1
|
27 |
+
- Dataset Link: [theblackcat102/evol-codealpaca-v1](https://huggingface.co/datasets/theblackcat102/evol-codealpaca-v1)
|
28 |
+
|
29 |
+
## Prompt template: Alpaca
|
30 |
+
|
31 |
+
```
|
32 |
+
Below is an instruction that describes a task. Write a response that appropriately completes the request.
|
33 |
+
|
34 |
+
### Instruction:
|
35 |
+
{prompt}
|
36 |
+
|
37 |
+
### Response:
|
38 |
+
|
39 |
+
```
|
40 |
+
|
41 |
+
|
42 |
+
## Performance (evalplus)
|
43 |
+
|
44 |
+
### The results from evalplus for the Mistral-7B-codealpaca are still pending.
|
45 |
+
|
46 |
+
For reference, we've provided the performance of the original Mistral model alongside Mistral-7B-code-16k-qlora model.
|
47 |
+
|
48 |
+
** [Nondzu/Mistral-7B-code-16k-qlora](https://huggingface.co/Nondzu/Mistral-7B-code-16k-qlora)**:
|
49 |
+
|
50 |
+
- Base: `{'pass@1': 0.3353658536585366}`
|
51 |
+
- Base + Extra: `{'pass@1': 0.2804878048780488}`
|
52 |
+
|
53 |
+
** [mistralai/Mistral-7B-Instruct-v0.1](https://huggingface.co/mistralai/Mistral-7B-Instruct-v0.1)**:
|
54 |
+
|
55 |
+
- Base: `{'pass@1': 0.2926829268292683}`
|
56 |
+
- Base + Extra: `{'pass@1': 0.24390243902439024}`
|
57 |
+
|
58 |
+
## Model Configuration:
|
59 |
+
|
60 |
+
The following are the configurations for the Mistral-7B-codealpaca-lora:
|
61 |
+
|
62 |
+
```yaml
|
63 |
+
base_model: mistralai/Mistral-7B-Instruct-v0.1
|
64 |
+
base_model_config: mistralai/Mistral-7B-Instruct-v0.1
|
65 |
+
model_type: MistralForCausalLM
|
66 |
+
tokenizer_type: LlamaTokenizer
|
67 |
+
is_mistral_derived_model: true
|
68 |
+
load_in_8bit: true
|
69 |
+
load_in_4bit: false
|
70 |
+
strict: false
|
71 |
+
datasets:
|
72 |
+
- path: theblackcat102/evol-codealpaca-v1
|
73 |
+
type: oasst
|
74 |
+
dataset_prepared_path:
|
75 |
+
val_set_size: 0.01
|
76 |
+
output_dir: ./nondzu/Mistral-7B-codealpaca-test14
|
77 |
+
adapter: lora
|
78 |
+
sequence_len: 4096
|
79 |
+
sample_packing: true
|
80 |
+
pad_to_sequence_len: true
|
81 |
+
lora_r: 32
|
82 |
+
lora_alpha: 16
|
83 |
+
lora_dropout: 0.05
|
84 |
+
lora_target_modules:
|
85 |
+
lora_target_linear: true
|
86 |
+
```
|
87 |
+
|
88 |
+
|
89 |
+
![image/png](https://cdn-uploads.huggingface.co/production/uploads/63729f35acef705233c87909/5nPgL3ajROKf7dttf4BO0.png)
|
90 |
+
|
91 |
+
|
92 |
+
## Additional Projects:
|
93 |
+
|
94 |
+
For other related projects, you can check out:
|
95 |
+
|
96 |
+
- [LlamaTor on GitHub](https://github.com/Nondzu/LlamaTor)
|