ekurtic commited on
Commit
5a9c1de
1 Parent(s): 9a87e5e

Create README.md

Browse files
Files changed (1) hide show
  1. README.md +62 -0
README.md ADDED
@@ -0,0 +1,62 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ base_model: neuralmagic/Llama-2-7b-pruned70-retrained
3
+ inference: true
4
+ model_type: llama
5
+ pipeline_tag: text-generation
6
+ datasets:
7
+ - openai/gsm8k
8
+ tags:
9
+ - sparse
10
+ ---
11
+
12
+ # Llama-2-7b-gsm8k-pruned_70
13
+
14
+ This repo contains a [70% sparse Llama 2 7B](https://huggingface.co/neuralmagic/Llama-2-7b-pruned70-retrained) finetuned for arithmetic reasoning using the [GSM8k](https://huggingface.co/datasets/openai/gsm8k) dataset.
15
+
16
+ Official model weights from [Enabling High-Sparsity Foundational Llama Models with Efficient Pretraining and Deployment](https://arxiv.org/abs/2405.03594).
17
+
18
+ **Authors**: Neural Magic, Cerebras
19
+
20
+ ## Usage
21
+
22
+ Below we share some code snippets on how to get quickly started with running the model.
23
+
24
+ ### Sparse Transfer
25
+
26
+ By leveraging a pre-sparsified model's structure, you can efficiently fine-tune on new data, leading to reduced hyperparameter tuning, training times, and computational costs. Learn about this process [here](https://neuralmagic.github.io/docs-v2/get-started/transfer).
27
+
28
+ ### Running the model
29
+
30
+ This model may be run with the transformers library. For accelerated inference with sparsity, deploy with [nm-vllm](https://github.com/neuralmagic/nm-vllm) or [deepsparse](https://github.com/neuralmagic/deepsparse).
31
+
32
+ ```python
33
+ # pip install transformers accelerate
34
+ from transformers import AutoTokenizer, AutoModelForCausalLM
35
+
36
+ tokenizer = AutoTokenizer.from_pretrained("neuralmagic/Llama-2-7b-gsm8k-pruned_70")
37
+ model = AutoModelForCausalLM.from_pretrained("neuralmagic/Llama-2-7b-gsm8k-pruned_70", device_map="auto")
38
+
39
+ input_text = "Natalia sold clips to 48 of her friends in April, and then she sold half as many clips in May. How many clips did Natalia sell altogether in April and May?"
40
+ input_ids = tokenizer.apply_chat_template(input_text, add_generation_prompt=True, return_tensors="pt").to("cuda")
41
+
42
+ outputs = model.generate(**input_ids)
43
+ print(tokenizer.decode(outputs[0]))
44
+ ```
45
+
46
+ ## Evaluation Benchmark Results
47
+
48
+ Model evaluation metrics and results.
49
+
50
+ | Benchmark | Metric | Llama-2-7b-gsm8k | Llama-2-7b-gsm8k-pruned_70 |
51
+ |:----:|:----:|:----:|:----:|
52
+ | [GSM8K](https://arxiv.org/abs/2110.14168) | 0-shot | 35.5% | 34.3% |
53
+
54
+
55
+ ## Model Training Details
56
+
57
+ This model was obtained by sparse-tranfer of the sparse foundational model [Llama-2-7b-pruned70-retrained](https://huggingface.co/neuralmagic/Llama-2-7b-pruned70-retrained) on the [GSM8k](https://huggingface.co/datasets/openai/gsm8k) dataset.
58
+ Sparse-transfer was performed with [SquareHead](https://arxiv.org/abs/2310.06927) knowledge distillation with [Llama-2-7b-gsm8k](https://huggingface.co/neuralmagic/Llama-2-7b-gsm8k) as teacher.
59
+
60
+ ## Help
61
+
62
+ For further support, and discussions on these models and AI in general, join [Neural Magic's Slack Community](https://join.slack.com/t/discuss-neuralmagic/shared_invite/zt-q1a1cnvo-YBoICSIw3L1dmQpjBeDurQ)