Update README.md
Browse files
README.md
CHANGED
@@ -8,9 +8,15 @@ base_model: tiiuae/falcon-7b
|
|
8 |
inference: false
|
9 |
---
|
10 |
|
11 |
-
#
|
12 |
|
13 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
14 |
|
15 |
## Model Summary
|
16 |
|
@@ -22,7 +28,7 @@ Falcon-7b-openassistant-peft is a chatbot model for dialogue generation. It was
|
|
22 |
|
23 |
## Model Details
|
24 |
|
25 |
-
The model was fine-tuned in 8-bit precision using π€ `peft` adapters, `transformers`, and `bitsandbytes`. Training relied on a method called "Low Rank Adapters" ([LoRA](https://arxiv.org/pdf/2106.09685.pdf)), specifically the [QLoRA](https://arxiv.org/abs/2305.14314) variant. The run took approximately 6.25 hours and was executed on a workstation with a single A100-SXM NVIDIA GPU with 37 GB of available memory. See attached [Colab Notebook](https://huggingface.co/dfurman/
|
26 |
|
27 |
### Model Date
|
28 |
|
@@ -45,7 +51,7 @@ To prompt the chat model, use the following format:
|
|
45 |
<bot>:"""
|
46 |
```
|
47 |
|
48 |
-
**
|
49 |
```
|
50 |
Dear friends,
|
51 |
|
@@ -65,7 +71,7 @@ Daniel
|
|
65 |
<bot>:
|
66 |
```
|
67 |
|
68 |
-
**
|
69 |
```
|
70 |
Here are some things to do in San Francisco:
|
71 |
|
@@ -120,7 +126,7 @@ from peft import PeftModel, PeftConfig
|
|
120 |
from transformers import AutoModelForCausalLM, AutoTokenizer
|
121 |
|
122 |
# load the model
|
123 |
-
peft_model_id = "dfurman/
|
124 |
config = PeftConfig.from_pretrained(peft_model_id)
|
125 |
|
126 |
model = AutoModelForCausalLM.from_pretrained(
|
@@ -173,7 +179,7 @@ print(generated_text.split("<human>: ")[1].split("<bot>: ")[-1])
|
|
173 |
|
174 |
## Reproducibility
|
175 |
|
176 |
-
See attached [Colab Notebook](https://huggingface.co/dfurman/
|
177 |
|
178 |
### CUDA Info
|
179 |
|
|
|
8 |
inference: false
|
9 |
---
|
10 |
|
11 |
+
# Falcon-7B-Chat-v0.1 π¦
|
12 |
|
13 |
+
<div align="left">
|
14 |
+
|
15 |
+
<img src="./falcon.webp" width="150px">
|
16 |
+
|
17 |
+
</div>
|
18 |
+
|
19 |
+
Falcon-7B-Chat-v0.1 is a chatbot model for dialogue generation. It was built by fine-tuning [Falcon-7B](https://huggingface.co/tiiuae/falcon-7b) on the [OpenAssistant/oasst1](https://huggingface.co/datasets/OpenAssistant/oasst1) dataset. This repo only includes the LoRA adapters from fine-tuning with π€'s [peft](https://github.com/huggingface/peft) package.
|
20 |
|
21 |
## Model Summary
|
22 |
|
|
|
28 |
|
29 |
## Model Details
|
30 |
|
31 |
+
The model was fine-tuned in 8-bit precision using π€ `peft` adapters, `transformers`, and `bitsandbytes`. Training relied on a method called "Low Rank Adapters" ([LoRA](https://arxiv.org/pdf/2106.09685.pdf)), specifically the [QLoRA](https://arxiv.org/abs/2305.14314) variant. The run took approximately 6.25 hours and was executed on a workstation with a single A100-SXM NVIDIA GPU with 37 GB of available memory. See attached [Colab Notebook](https://huggingface.co/dfurman/Falcon-7B-Chat-v0.1/blob/main/finetune_falcon7b_oasst1_with_bnb_peft.ipynb) for the code and hyperparams used to train the model.
|
32 |
|
33 |
### Model Date
|
34 |
|
|
|
51 |
<bot>:"""
|
52 |
```
|
53 |
|
54 |
+
**Falcon-7B-Chat-v0.1**:
|
55 |
```
|
56 |
Dear friends,
|
57 |
|
|
|
71 |
<bot>:
|
72 |
```
|
73 |
|
74 |
+
**Falcon-7B-Chat-v0.1**:
|
75 |
```
|
76 |
Here are some things to do in San Francisco:
|
77 |
|
|
|
126 |
from transformers import AutoModelForCausalLM, AutoTokenizer
|
127 |
|
128 |
# load the model
|
129 |
+
peft_model_id = "dfurman/Falcon-7B-Chat-v0.1"
|
130 |
config = PeftConfig.from_pretrained(peft_model_id)
|
131 |
|
132 |
model = AutoModelForCausalLM.from_pretrained(
|
|
|
179 |
|
180 |
## Reproducibility
|
181 |
|
182 |
+
See attached [Colab Notebook](https://huggingface.co/dfurman/Falcon-7B-Chat-v0.1/blob/main/finetune_falcon7b_oasst1_with_bnb_peft.ipynb) for the code (and hyperparams) used to train the model.
|
183 |
|
184 |
### CUDA Info
|
185 |
|