danielhanchen
commited on
Commit
•
edd5803
1
Parent(s):
47f404c
Update README.md
Browse files
README.md
CHANGED
@@ -1,199 +1,131 @@
|
|
1 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2 |
library_name: transformers
|
3 |
-
|
|
|
4 |
---
|
5 |
|
6 |
-
#
|
7 |
|
8 |
-
|
9 |
|
|
|
|
|
10 |
|
|
|
11 |
|
12 |
-
|
13 |
|
14 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
15 |
|
16 |
-
|
|
|
|
|
17 |
|
18 |
-
|
19 |
|
20 |
-
|
21 |
-
|
22 |
-
|
23 |
-
- **Model type:** [More Information Needed]
|
24 |
-
- **Language(s) (NLP):** [More Information Needed]
|
25 |
-
- **License:** [More Information Needed]
|
26 |
-
- **Finetuned from model [optional]:** [More Information Needed]
|
27 |
|
28 |
-
### Model Sources [optional]
|
29 |
|
30 |
-
|
31 |
|
32 |
-
|
33 |
-
- **Paper [optional]:** [More Information Needed]
|
34 |
-
- **Demo [optional]:** [More Information Needed]
|
35 |
|
36 |
-
|
37 |
|
38 |
-
|
39 |
|
40 |
-
|
41 |
|
42 |
-
<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
|
43 |
|
44 |
-
|
|
|
|
|
|
|
45 |
|
46 |
-
|
47 |
|
48 |
-
|
|
|
|
|
|
|
49 |
|
50 |
-
|
51 |
|
52 |
-
###
|
|
|
53 |
|
54 |
-
|
55 |
|
56 |
-
|
|
|
|
|
|
|
57 |
|
58 |
-
|
|
|
|
|
|
|
59 |
|
60 |
-
|
|
|
|
|
|
|
61 |
|
62 |
-
[
|
|
|
|
|
|
|
|
|
|
|
|
|
63 |
|
64 |
-
###
|
|
|
|
|
|
|
|
|
|
|
65 |
|
66 |
-
|
67 |
|
68 |
-
|
69 |
|
70 |
-
##
|
|
|
71 |
|
72 |
-
|
|
|
|
|
|
|
|
|
73 |
|
74 |
-
|
75 |
|
76 |
-
|
77 |
-
|
78 |
-
|
79 |
-
|
80 |
-
|
81 |
-
|
82 |
-
|
83 |
-
|
84 |
-
### Training Procedure
|
85 |
-
|
86 |
-
<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
|
87 |
-
|
88 |
-
#### Preprocessing [optional]
|
89 |
-
|
90 |
-
[More Information Needed]
|
91 |
-
|
92 |
-
|
93 |
-
#### Training Hyperparameters
|
94 |
-
|
95 |
-
- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
|
96 |
-
|
97 |
-
#### Speeds, Sizes, Times [optional]
|
98 |
-
|
99 |
-
<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
|
100 |
-
|
101 |
-
[More Information Needed]
|
102 |
-
|
103 |
-
## Evaluation
|
104 |
-
|
105 |
-
<!-- This section describes the evaluation protocols and provides the results. -->
|
106 |
-
|
107 |
-
### Testing Data, Factors & Metrics
|
108 |
-
|
109 |
-
#### Testing Data
|
110 |
-
|
111 |
-
<!-- This should link to a Dataset Card if possible. -->
|
112 |
-
|
113 |
-
[More Information Needed]
|
114 |
-
|
115 |
-
#### Factors
|
116 |
-
|
117 |
-
<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
|
118 |
-
|
119 |
-
[More Information Needed]
|
120 |
-
|
121 |
-
#### Metrics
|
122 |
-
|
123 |
-
<!-- These are the evaluation metrics being used, ideally with a description of why. -->
|
124 |
-
|
125 |
-
[More Information Needed]
|
126 |
-
|
127 |
-
### Results
|
128 |
-
|
129 |
-
[More Information Needed]
|
130 |
-
|
131 |
-
#### Summary
|
132 |
-
|
133 |
-
|
134 |
-
|
135 |
-
## Model Examination [optional]
|
136 |
-
|
137 |
-
<!-- Relevant interpretability work for the model goes here -->
|
138 |
-
|
139 |
-
[More Information Needed]
|
140 |
-
|
141 |
-
## Environmental Impact
|
142 |
-
|
143 |
-
<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
|
144 |
-
|
145 |
-
Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
|
146 |
-
|
147 |
-
- **Hardware Type:** [More Information Needed]
|
148 |
-
- **Hours used:** [More Information Needed]
|
149 |
-
- **Cloud Provider:** [More Information Needed]
|
150 |
-
- **Compute Region:** [More Information Needed]
|
151 |
-
- **Carbon Emitted:** [More Information Needed]
|
152 |
-
|
153 |
-
## Technical Specifications [optional]
|
154 |
-
|
155 |
-
### Model Architecture and Objective
|
156 |
-
|
157 |
-
[More Information Needed]
|
158 |
-
|
159 |
-
### Compute Infrastructure
|
160 |
-
|
161 |
-
[More Information Needed]
|
162 |
-
|
163 |
-
#### Hardware
|
164 |
-
|
165 |
-
[More Information Needed]
|
166 |
-
|
167 |
-
#### Software
|
168 |
-
|
169 |
-
[More Information Needed]
|
170 |
-
|
171 |
-
## Citation [optional]
|
172 |
-
|
173 |
-
<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
|
174 |
-
|
175 |
-
**BibTeX:**
|
176 |
-
|
177 |
-
[More Information Needed]
|
178 |
-
|
179 |
-
**APA:**
|
180 |
-
|
181 |
-
[More Information Needed]
|
182 |
-
|
183 |
-
## Glossary [optional]
|
184 |
-
|
185 |
-
<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
|
186 |
-
|
187 |
-
[More Information Needed]
|
188 |
-
|
189 |
-
## More Information [optional]
|
190 |
-
|
191 |
-
[More Information Needed]
|
192 |
-
|
193 |
-
## Model Card Authors [optional]
|
194 |
-
|
195 |
-
[More Information Needed]
|
196 |
-
|
197 |
-
## Model Card Contact
|
198 |
-
|
199 |
-
[More Information Needed]
|
|
|
1 |
---
|
2 |
+
license: apache-2.0
|
3 |
+
base_model: HuggingFaceTB/SmolLM-360M
|
4 |
+
tags:
|
5 |
+
- alignment-handbook
|
6 |
+
- trl
|
7 |
+
- unsloth
|
8 |
+
datasets:
|
9 |
+
- Magpie-Align/Magpie-Pro-300K-Filtered
|
10 |
+
- bigcode/self-oss-instruct-sc2-exec-filter-50k
|
11 |
+
- teknium/OpenHermes-2.5
|
12 |
+
- HuggingFaceTB/everyday-conversations-llama3.1-2k
|
13 |
library_name: transformers
|
14 |
+
language:
|
15 |
+
- en
|
16 |
---
|
17 |
|
18 |
+
# Finetune Llama 3.1, Gemma 2, Mistral 2-5x faster with 70% less memory via Unsloth!
|
19 |
|
20 |
+
We have a free Google Colab Tesla T4 notebook for Llama 3.1 (8B) here - also works for SmolLM!: https://colab.research.google.com/drive/1Ys44kVvmeZtnICzWz0xgpRnrIOjZAuxp?usp=sharing
|
21 |
|
22 |
+
[<img src="https://raw.githubusercontent.com/unslothai/unsloth/main/images/Discord%20button.png" width="200"/>](https://discord.gg/unsloth)
|
23 |
+
[<img src="https://raw.githubusercontent.com/unslothai/unsloth/main/images/unsloth%20made%20with%20love.png" width="200"/>](https://github.com/unslothai/unsloth)
|
24 |
|
25 |
+
## ✨ Finetune for Free
|
26 |
|
27 |
+
All notebooks are **beginner friendly**! Add your dataset, click "Run All", and you'll get a 2x faster finetuned model which can be exported to GGUF, vLLM or uploaded to Hugging Face.
|
28 |
|
29 |
+
| Unsloth supports | Free Notebooks | Performance | Memory use |
|
30 |
+
|-----------------|--------------------------------------------------------------------------------------------------------------------------|-------------|----------|
|
31 |
+
| **Llama-3.1 8b** | [▶️ Start on Colab](https://colab.research.google.com/drive/1Ys44kVvmeZtnICzWz0xgpRnrIOjZAuxp?usp=sharing) | 2.4x faster | 58% less |
|
32 |
+
| **Phi-3.5 (mini)** | [▶️ Start on Colab](https://colab.research.google.com/drive/1lN6hPQveB_mHSnTOYifygFcrO8C1bxq4?usp=sharing) | 2x faster | 50% less |
|
33 |
+
| **Gemma-2 9b** | [▶️ Start on Colab](https://colab.research.google.com/drive/1vIrqH5uYDQwsJ4-OO3DErvuv4pBgVwk4?usp=sharing) | 2.4x faster | 58% less |
|
34 |
+
| **Mistral 7b** | [▶️ Start on Colab](https://colab.research.google.com/drive/1Dyauq4kTZoLewQ1cApceUQVNcnnNTzg_?usp=sharing) | 2.2x faster | 62% less |
|
35 |
+
| **TinyLlama** | [▶️ Start on Colab](https://colab.research.google.com/drive/1AZghoNBQaMDgWJpi4RbffGM1h6raLUj9?usp=sharing) | 3.9x faster | 74% less |
|
36 |
+
| **DPO - Zephyr** | [▶️ Start on Colab](https://colab.research.google.com/drive/15vttTpzzVXv_tJwEk-hIcQ0S9FcEWvwP?usp=sharing) | 1.9x faster | 19% less |
|
37 |
|
38 |
+
- This [conversational notebook](https://colab.research.google.com/drive/1Aau3lgPzeZKQ-98h69CCu1UJcvIBLmy2?usp=sharing) is useful for ShareGPT ChatML / Vicuna templates.
|
39 |
+
- This [text completion notebook](https://colab.research.google.com/drive/1ef-tab5bhkvWmBOObepl1WgJvfvSzn5Q?usp=sharing) is for raw text. This [DPO notebook](https://colab.research.google.com/drive/15vttTpzzVXv_tJwEk-hIcQ0S9FcEWvwP?usp=sharing) replicates Zephyr.
|
40 |
+
- \* Kaggle has 2x T4s, but we use 1. Due to overhead, 1x T4 is 5x faster.
|
41 |
|
42 |
+
# SmolLM-1.7B-Instruct
|
43 |
|
44 |
+
<center>
|
45 |
+
<img src="https://huggingface.co/datasets/HuggingFaceTB/images/resolve/main/banner_smol.png" alt="SmolLM" width="1100" height="600">
|
46 |
+
</center>
|
|
|
|
|
|
|
|
|
47 |
|
|
|
48 |
|
49 |
+
## Model Summary
|
50 |
|
51 |
+
SmolLM is a series of small language models available in three sizes: 135M, 360M, and 1.7B parameters.
|
|
|
|
|
52 |
|
53 |
+
These models are pre-trained on [SmolLM-Corpus](https://huggingface.co/datasets/HuggingFaceTB/smollm-corpus), a curated collection of high-quality educational and synthetic data designed for training LLMs. For further details, we refer to our [blogpost](https://huggingface.co/blog/smollm).
|
54 |
|
55 |
+
To build SmolLM-Instruct, we finetuned the base models on publicly available datasets.
|
56 |
|
57 |
+
## Changelog
|
58 |
|
|
|
59 |
|
60 |
+
|Release|Description|
|
61 |
+
|-|-|
|
62 |
+
|v0.1| Initial release of SmolLM-Instruct. We finetune on the permissive subset of the [WebInstructSub](https://huggingface.co/datasets/TIGER-Lab/WebInstructSub) dataset, combined with [StarCoder2-Self-OSS-Instruct](https://huggingface.co/datasets/bigcode/self-oss-instruct-sc2-exec-filter-50k). Then, we perform DPO (Direct Preference Optimization) for one epoch on [HelpSteer](https://huggingface.co/datasets/nvidia/HelpSteer) for the 135M and 1.7B models, and [argilla/dpo-mix-7k](https://huggingface.co/datasets/argilla/dpo-mix-7k) for the 360M model.|
|
63 |
+
|v0.2| We changed the finetuning mix to datasets more suitable for smol models. We train on a new dataset of 2k simple everyday conversations we generated by llama3.1-70B [everyday-conversations-llama3.1-2k](https://huggingface.co/datasets/HuggingFaceTB/everyday-conversations-llama3.1-2k/), [Magpie-Pro-300K-Filtered](https://huggingface.co/datasets/Magpie-Align/Magpie-Pro-300K-Filtered), [StarCoder2-Self-OSS-Instruct](https://huggingface.co/datasets/bigcode/self-oss-instruct-sc2-exec-filter-50k), and a small subset of [OpenHermes-2.5](https://huggingface.co/datasets/teknium/OpenHermes-2.5)|
|
64 |
|
65 |
+
v0.2 models are better at staying on topic and responding appropriately to standard prompts, such as greetings and questions about their role as AI assistants. SmolLM-360M-Instruct (v0.2) has a 63.3% win rate over SmolLM-360M-Instruct (v0.1) on AlpacaEval. You can find the details [here](https://huggingface.co/datasets/HuggingFaceTB/alpaca_eval_details/).
|
66 |
|
67 |
+
You can load v0.1 checkpoint by specifying `revision="v0.1"` in the transformers code:
|
68 |
+
```python
|
69 |
+
model = AutoModelForCausalLM.from_pretrained("HuggingFaceTB/SmolLM-1.7B-Instruct", revision="v0.1")
|
70 |
+
```
|
71 |
|
72 |
+
## Usage
|
73 |
|
74 |
+
### Local Applications
|
75 |
+
⚡ For local applications, you can find optimized implementations of the model in MLC, GGUF and Transformers.js formats, in addition to fast in-browser demos in this collection: https://huggingface.co/collections/HuggingFaceTB/local-smollms-66c0f3b2a15b4eed7fb198d0
|
76 |
|
77 |
+
We noticed that 4bit quantization degrades the quality of the 135M and 360M, so we use `q016` for MLC and ONNX/Transformers.js checkpoints for the WebGPU demos. We also suggest using temperature 0.2 and top-p 0.9.
|
78 |
|
79 |
+
### Transformers
|
80 |
+
```bash
|
81 |
+
pip install transformers
|
82 |
+
```
|
83 |
|
84 |
+
```python
|
85 |
+
# pip install transformers
|
86 |
+
from transformers import AutoModelForCausalLM, AutoTokenizer
|
87 |
+
checkpoint = "HuggingFaceTB/SmolLM-1.7B-Instruct"
|
88 |
|
89 |
+
device = "cuda" # for GPU usage or "cpu" for CPU usage
|
90 |
+
tokenizer = AutoTokenizer.from_pretrained(checkpoint)
|
91 |
+
# for multiple GPUs install accelerate and do `model = AutoModelForCausalLM.from_pretrained(checkpoint, device_map="auto")`
|
92 |
+
model = AutoModelForCausalLM.from_pretrained(checkpoint).to(device)
|
93 |
|
94 |
+
messages = [{"role": "user", "content": "What is the capital of France."}]
|
95 |
+
input_text=tokenizer.apply_chat_template(messages, tokenize=False)
|
96 |
+
print(input_text)
|
97 |
+
inputs = tokenizer.encode(input_text, return_tensors="pt").to(device)
|
98 |
+
outputs = model.generate(inputs, max_new_tokens=50, temperature=0.2, top_p=0.9, do_sample=True)
|
99 |
+
print(tokenizer.decode(outputs[0]))
|
100 |
+
```
|
101 |
|
102 |
+
### Chat in TRL
|
103 |
+
You can also use the TRL CLI to chat with the model from the terminal:
|
104 |
+
```bash
|
105 |
+
pip install trl
|
106 |
+
trl chat --model_name_or_path HuggingFaceTB/SmolLM-1.7B-Instruct --device cpu
|
107 |
+
```
|
108 |
|
109 |
+
## Limitations
|
110 |
|
111 |
+
Additionally, the generated content may not always be factually accurate, logically consistent, or free from biases present in the training data, we invite users to leverage them as assistive tools rather than definitive sources of information. We find that they can handle general knowledge questions, creative writing and basic Python programming. But they are English only and may have difficulty with arithmetics, editing tasks and complex reasoning. For more details about the models' capabilities, please refer to our [blog post](https://huggingface.co/blog/smollm).
|
112 |
|
113 |
+
## Training parameters
|
114 |
+
We train the models using the [alignment-handbook](https://github.com/huggingface/alignment-handbook) with the datasets mentioned in the changelog, using these parameters v0.2 (most of them are from Zephyr Gemma recipe):
|
115 |
|
116 |
+
- 1 epoch
|
117 |
+
- lr 1e-3
|
118 |
+
- cosine schedule
|
119 |
+
- warmup ratio 0.1
|
120 |
+
- global batch size 262k tokens
|
121 |
|
122 |
+
You can find the training recipe here: https://github.com/huggingface/alignment-handbook/tree/smollm/recipes/smollm
|
123 |
|
124 |
+
# Citation
|
125 |
+
```bash
|
126 |
+
@misc{allal2024SmolLM,
|
127 |
+
title={SmolLM - blazingly fast and remarkably powerful},
|
128 |
+
author={Loubna Ben Allal and Anton Lozhkov and Elie Bakouch and Leandro von Werra and Thomas Wolf},
|
129 |
+
year={2024},
|
130 |
+
}
|
131 |
+
```
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|