lee101 commited on
Commit
3622bf0
·
verified ·
1 Parent(s): 7b7dd1b

Publish Gemma Roleplay v2 adapter files

Browse files
Files changed (1) hide show
  1. adapter/README.md +73 -0
adapter/README.md ADDED
@@ -0,0 +1,73 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ library_name: peft
3
+ model_name: gemma-roleplay-v2-lora
4
+ tags:
5
+ - base_model:adapter:google/gemma-4-E4B-it
6
+ - lora
7
+ - sft
8
+ - transformers
9
+ - trl
10
+ - roleplay
11
+ license: gemma
12
+ base_model: google/gemma-4-E4B-it
13
+ pipeline_tag: text-generation
14
+ ---
15
+
16
+ # Gemma Roleplay v2 LoRA adapter
17
+
18
+ This is the PEFT adapter for [Gemma Roleplay v2](https://huggingface.co/text-generator/llmtrain),
19
+ trained from [google/gemma-4-E4B-it](https://huggingface.co/google/gemma-4-E4B-it)
20
+ with QLoRA SFT. It is intended for fictional consenting-adult roleplay and
21
+ creative chat. See the parent model card for usage, limitations, and the live
22
+ hosted inference endpoint.
23
+
24
+ ## Quick start
25
+
26
+ ```python
27
+ from transformers import AutoModelForCausalLM, AutoTokenizer
28
+ from peft import PeftModel
29
+
30
+ base = "google/gemma-4-E4B-it"
31
+ tokenizer = AutoTokenizer.from_pretrained(base)
32
+ model = AutoModelForCausalLM.from_pretrained(base, torch_dtype="auto", device_map="auto")
33
+ model = PeftModel.from_pretrained(model, "text-generator/llmtrain", subfolder="adapter")
34
+ inputs = tokenizer.apply_chat_template(
35
+ [{"role": "user", "content": "Write a short scene in a haunted hotel."}],
36
+ add_generation_prompt=True, return_tensors="pt",
37
+ ).to(model.device)
38
+ output = model.generate(inputs, max_new_tokens=128, do_sample=True, temperature=0.85)
39
+ print(tokenizer.decode(output[0, inputs.shape[-1]:], skip_special_tokens=True))
40
+ ```
41
+
42
+ ## Training procedure
43
+
44
+
45
+
46
+
47
+
48
+ This model was trained with SFT.
49
+
50
+ ### Framework versions
51
+
52
+ - PEFT 0.18.0
53
+ - TRL: 1.8.0
54
+ - Transformers: 5.5.0
55
+ - Pytorch: 2.9.1
56
+ - Datasets: 4.0.0
57
+ - Tokenizers: 0.22.2
58
+
59
+ ## Citations
60
+
61
+
62
+
63
+ Cite TRL as:
64
+
65
+ ```bibtex
66
+ @software{vonwerra2020trl,
67
+ title = {{TRL: Transformers Reinforcement Learning}},
68
+ author = {von Werra, Leandro and Belkada, Younes and Tunstall, Lewis and Beeching, Edward and Thrush, Tristan and Lambert, Nathan and Huang, Shengyi and Rasul, Kashif and Gallouédec, Quentin},
69
+ license = {Apache-2.0},
70
+ url = {https://github.com/huggingface/trl},
71
+ year = {2020}
72
+ }
73
+ ```