mgoin commited on
Commit
cca17b4
1 Parent(s): 01c34b2

Create README.md

Browse files
Files changed (1) hide show
  1. README.md +64 -0
README.md ADDED
@@ -0,0 +1,64 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ base_model: meta-llama/Llama-2-7b-hf
3
+ inference: true
4
+ model_type: llama
5
+ pipeline_tag: text-generation
6
+ datasets:
7
+ - HuggingFaceH4/ultrachat_200k
8
+ tags:
9
+ - chat
10
+ ---
11
+
12
+ # Llama-2-7b-ultrachat
13
+
14
+ This repo contains a [Llama 2 7B](https://huggingface.co/meta-llama/Llama-2-7b-hf) finetuned for chat tasks using the [UltraChat 200k](https://huggingface.co/datasets/HuggingFaceH4/ultrachat_200k) dataset.
15
+
16
+ **Authors**: Neural Magic, Cerebras
17
+
18
+ ## Usage
19
+
20
+ Below we share some code snippets on how to get quickly started with running the model.
21
+
22
+ ### Sparse Transfer
23
+
24
+ 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).
25
+
26
+ ### Running the model
27
+
28
+ 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).
29
+
30
+ ```python
31
+ # pip install transformers accelerate
32
+ from transformers import AutoTokenizer, AutoModelForCausalLM
33
+
34
+ tokenizer = AutoTokenizer.from_pretrained("neuralmagic/Llama-2-7b-ultrachat")
35
+ model = AutoModelForCausalLM.from_pretrained("neuralmagic/Llama-2-7b-ultrachat", device_map="auto")
36
+
37
+ input_text = "Write me a poem about Machine Learning."
38
+ input_ids = tokenizer.apply_chat_template(input_text, add_generation_prompt=True, return_tensors="pt").to("cuda")
39
+
40
+ outputs = model.generate(**input_ids)
41
+ print(tokenizer.decode(outputs[0]))
42
+ ```
43
+
44
+ ## Evaluation Benchmark Results
45
+
46
+ Model evaluation metrics and results.
47
+
48
+ | Benchmark | Metric | Llama-2-7b-ultrachat | Llama-2-7b-pruned50-retrained-ultrachat |
49
+ |------------------------------------------------|---------------|-------------|-------------------------------|
50
+ | [MMLU](https://arxiv.org/abs/2009.03300) | 5-shot, top-1 | xxxx | xxxx |
51
+ | [HellaSwag](https://arxiv.org/abs/1905.07830) | 0-shot | xxxx | xxxx |
52
+ | [WinoGrande](https://arxiv.org/abs/1907.10641) | partial score | xxxx | xxxx |
53
+ | [ARC-c](https://arxiv.org/abs/1911.01547) | | xxxx | xxxx |
54
+ | [TruthfulQA](https://arxiv.org/abs/2109.07958) | 5-shot | xxxx | xxxx |
55
+ | [HumanEval](https://arxiv.org/abs/2107.03374) | pass@1 | xxxx | xxxx |
56
+ | [GSM8K](https://arxiv.org/abs/2110.14168) | maj@1 | xxxx | xxxx |
57
+
58
+ ## Model Training Details
59
+
60
+ Coming soon.
61
+
62
+ ## Help
63
+
64
+ 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)