loubnabnl HF staff commited on
Commit
2957b4b
1 Parent(s): 01de89e

mention how to load v0.1 in readme

Browse files
Files changed (1) hide show
  1. README.md +7 -0
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,