mention how to load v0.1 in readme
Browse files
README.md
CHANGED
@@ -40,6 +40,11 @@ To build SmolLM-Instruct, we finetuned the base models on publicly available dat
|
|
40 |
|
41 |
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/).
|
42 |
|
|
|
|
|
|
|
|
|
|
|
43 |
## Usage
|
44 |
|
45 |
### Local Applications
|
@@ -90,6 +95,8 @@ We train the models using the [alignment-handbook](https://github.com/huggingfac
|
|
90 |
- warmup ratio 0.1
|
91 |
- global batch size 262k tokens
|
92 |
|
|
|
|
|
93 |
# Citation
|
94 |
```bash
|
95 |
@misc{allal2024SmolLM,
|
|
|
40 |
|
41 |
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/).
|
42 |
|
43 |
+
You can load v0.1 checkpoint by specifying `revision="v0.1"` in the transformers code:
|
44 |
+
```python
|
45 |
+
model = AutoModelForCausalLM.from_pretrained("HuggingFaceTB/SmolLM-1.7B-Instruct", revision="v0.1")
|
46 |
+
```
|
47 |
+
|
48 |
## Usage
|
49 |
|
50 |
### Local Applications
|
|
|
95 |
- warmup ratio 0.1
|
96 |
- global batch size 262k tokens
|
97 |
|
98 |
+
You can find the training recipe here: https://github.com/huggingface/alignment-handbook/tree/smollm/recipes/smollm
|
99 |
+
|
100 |
# Citation
|
101 |
```bash
|
102 |
@misc{allal2024SmolLM,
|