Evan-Lin/dpo-llama-chat
Browse files- README.md +83 -0
- adapter_config.json +31 -0
- adapter_model.safetensors +3 -0
- special_tokens_map.json +24 -0
- tokenizer.json +0 -0
- tokenizer.model +3 -0
- tokenizer_config.json +42 -0
- training_args.bin +3 -0
README.md
ADDED
@@ -0,0 +1,83 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
library_name: peft
|
3 |
+
tags:
|
4 |
+
- trl
|
5 |
+
- dpo
|
6 |
+
- generated_from_trainer
|
7 |
+
base_model: meta-llama/Llama-2-7b-chat-hf
|
8 |
+
model-index:
|
9 |
+
- name: dpo-llama-chat-without-none
|
10 |
+
results: []
|
11 |
+
---
|
12 |
+
|
13 |
+
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
14 |
+
should probably proofread and complete it, then remove this comment. -->
|
15 |
+
|
16 |
+
# dpo-llama-chat-without-none
|
17 |
+
|
18 |
+
This model is a fine-tuned version of [meta-llama/Llama-2-7b-chat-hf](https://huggingface.co/meta-llama/Llama-2-7b-chat-hf) on the None dataset.
|
19 |
+
It achieves the following results on the evaluation set:
|
20 |
+
- Loss: 0.0177
|
21 |
+
- Rewards/chosen: -5.0689
|
22 |
+
- Rewards/rejected: -14.1911
|
23 |
+
- Rewards/accuracies: 0.9933
|
24 |
+
- Rewards/margins: 9.1223
|
25 |
+
- Logps/rejected: -230.9264
|
26 |
+
- Logps/chosen: -147.4190
|
27 |
+
- Logits/rejected: -0.6518
|
28 |
+
- Logits/chosen: -0.6273
|
29 |
+
|
30 |
+
## Model description
|
31 |
+
|
32 |
+
More information needed
|
33 |
+
|
34 |
+
## Intended uses & limitations
|
35 |
+
|
36 |
+
More information needed
|
37 |
+
|
38 |
+
## Training and evaluation data
|
39 |
+
|
40 |
+
More information needed
|
41 |
+
|
42 |
+
## Training procedure
|
43 |
+
|
44 |
+
### Training hyperparameters
|
45 |
+
|
46 |
+
The following hyperparameters were used during training:
|
47 |
+
- learning_rate: 0.0001
|
48 |
+
- train_batch_size: 2
|
49 |
+
- eval_batch_size: 2
|
50 |
+
- seed: 42
|
51 |
+
- distributed_type: multi-GPU
|
52 |
+
- num_devices: 2
|
53 |
+
- gradient_accumulation_steps: 64
|
54 |
+
- total_train_batch_size: 256
|
55 |
+
- total_eval_batch_size: 4
|
56 |
+
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
57 |
+
- lr_scheduler_type: cosine
|
58 |
+
- lr_scheduler_warmup_steps: 50
|
59 |
+
- training_steps: 1000
|
60 |
+
|
61 |
+
### Training results
|
62 |
+
|
63 |
+
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
|
64 |
+
|:-------------:|:-----:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
|
65 |
+
| 0.2627 | 1.92 | 100 | 0.2451 | -1.0090 | -3.0456 | 0.9240 | 2.0366 | -119.4711 | -106.8200 | -0.8708 | -0.8600 |
|
66 |
+
| 0.0386 | 3.84 | 200 | 0.0708 | -2.3491 | -7.1127 | 0.9813 | 4.7636 | -160.1419 | -120.2212 | -0.8381 | -0.8197 |
|
67 |
+
| 0.0147 | 5.76 | 300 | 0.0424 | -2.9584 | -8.9699 | 0.9840 | 6.0115 | -178.7142 | -126.3143 | -0.7736 | -0.7537 |
|
68 |
+
| 0.005 | 7.69 | 400 | 0.0294 | -3.7997 | -11.1138 | 0.9907 | 7.3141 | -200.1534 | -134.7274 | -0.7233 | -0.6967 |
|
69 |
+
| 0.002 | 9.61 | 500 | 0.0186 | -4.4609 | -12.4831 | 0.9947 | 8.0222 | -213.8460 | -141.3395 | -0.6848 | -0.6600 |
|
70 |
+
| 0.0008 | 11.53 | 600 | 0.0193 | -4.8300 | -13.5310 | 0.9947 | 8.7010 | -224.3250 | -145.0307 | -0.6646 | -0.6397 |
|
71 |
+
| 0.0006 | 13.45 | 700 | 0.0166 | -4.8968 | -13.8306 | 0.9947 | 8.9338 | -227.3210 | -145.6982 | -0.6552 | -0.6304 |
|
72 |
+
| 0.0005 | 15.37 | 800 | 0.0179 | -4.9970 | -14.0354 | 0.9933 | 9.0383 | -229.3687 | -146.7007 | -0.6498 | -0.6253 |
|
73 |
+
| 0.0005 | 17.29 | 900 | 0.0156 | -5.0417 | -14.1518 | 0.9947 | 9.1101 | -230.5333 | -147.1470 | -0.6492 | -0.6245 |
|
74 |
+
| 0.0005 | 19.21 | 1000 | 0.0177 | -5.0689 | -14.1911 | 0.9933 | 9.1223 | -230.9264 | -147.4190 | -0.6518 | -0.6273 |
|
75 |
+
|
76 |
+
|
77 |
+
### Framework versions
|
78 |
+
|
79 |
+
- PEFT 0.7.1
|
80 |
+
- Transformers 4.36.2
|
81 |
+
- Pytorch 2.1.2+cu121
|
82 |
+
- Datasets 2.16.1
|
83 |
+
- Tokenizers 0.15.1
|
adapter_config.json
ADDED
@@ -0,0 +1,31 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"alpha_pattern": {},
|
3 |
+
"auto_mapping": null,
|
4 |
+
"base_model_name_or_path": "meta-llama/Llama-2-7b-chat-hf",
|
5 |
+
"bias": "none",
|
6 |
+
"fan_in_fan_out": false,
|
7 |
+
"inference_mode": true,
|
8 |
+
"init_lora_weights": true,
|
9 |
+
"layers_pattern": null,
|
10 |
+
"layers_to_transform": null,
|
11 |
+
"loftq_config": {},
|
12 |
+
"lora_alpha": 16,
|
13 |
+
"lora_dropout": 0.05,
|
14 |
+
"megatron_config": null,
|
15 |
+
"megatron_core": "megatron.core",
|
16 |
+
"modules_to_save": null,
|
17 |
+
"peft_type": "LORA",
|
18 |
+
"r": 8,
|
19 |
+
"rank_pattern": {},
|
20 |
+
"revision": null,
|
21 |
+
"target_modules": [
|
22 |
+
"v_proj",
|
23 |
+
"out_proj",
|
24 |
+
"wte",
|
25 |
+
"k_proj",
|
26 |
+
"fc_out",
|
27 |
+
"q_proj",
|
28 |
+
"fc_in"
|
29 |
+
],
|
30 |
+
"task_type": "CAUSAL_LM"
|
31 |
+
}
|
adapter_model.safetensors
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:141dd4ed5d7add238576f244954362a8b9fdbb6e8f8f2bbb1bbfcb8157e84cff
|
3 |
+
size 25191360
|
special_tokens_map.json
ADDED
@@ -0,0 +1,24 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"bos_token": {
|
3 |
+
"content": "<s>",
|
4 |
+
"lstrip": false,
|
5 |
+
"normalized": false,
|
6 |
+
"rstrip": false,
|
7 |
+
"single_word": false
|
8 |
+
},
|
9 |
+
"eos_token": {
|
10 |
+
"content": "</s>",
|
11 |
+
"lstrip": false,
|
12 |
+
"normalized": false,
|
13 |
+
"rstrip": false,
|
14 |
+
"single_word": false
|
15 |
+
},
|
16 |
+
"pad_token": "</s>",
|
17 |
+
"unk_token": {
|
18 |
+
"content": "<unk>",
|
19 |
+
"lstrip": false,
|
20 |
+
"normalized": false,
|
21 |
+
"rstrip": false,
|
22 |
+
"single_word": false
|
23 |
+
}
|
24 |
+
}
|
tokenizer.json
ADDED
The diff for this file is too large to render.
See raw diff
|
|
tokenizer.model
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:9e556afd44213b6bd1be2b850ebbbd98f5481437a8021afaf58ee7fb1818d347
|
3 |
+
size 499723
|
tokenizer_config.json
ADDED
@@ -0,0 +1,42 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"add_bos_token": true,
|
3 |
+
"add_eos_token": false,
|
4 |
+
"added_tokens_decoder": {
|
5 |
+
"0": {
|
6 |
+
"content": "<unk>",
|
7 |
+
"lstrip": false,
|
8 |
+
"normalized": false,
|
9 |
+
"rstrip": false,
|
10 |
+
"single_word": false,
|
11 |
+
"special": true
|
12 |
+
},
|
13 |
+
"1": {
|
14 |
+
"content": "<s>",
|
15 |
+
"lstrip": false,
|
16 |
+
"normalized": false,
|
17 |
+
"rstrip": false,
|
18 |
+
"single_word": false,
|
19 |
+
"special": true
|
20 |
+
},
|
21 |
+
"2": {
|
22 |
+
"content": "</s>",
|
23 |
+
"lstrip": false,
|
24 |
+
"normalized": false,
|
25 |
+
"rstrip": false,
|
26 |
+
"single_word": false,
|
27 |
+
"special": true
|
28 |
+
}
|
29 |
+
},
|
30 |
+
"bos_token": "<s>",
|
31 |
+
"chat_template": "{% if messages[0]['role'] == 'system' %}{% set loop_messages = messages[1:] %}{% set system_message = messages[0]['content'] %}{% else %}{% set loop_messages = messages %}{% set system_message = false %}{% endif %}{% for message in loop_messages %}{% if (message['role'] == 'user') != (loop.index0 % 2 == 0) %}{{ raise_exception('Conversation roles must alternate user/assistant/user/assistant/...') }}{% endif %}{% if loop.index0 == 0 and system_message != false %}{% set content = '<<SYS>>\\n' + system_message + '\\n<</SYS>>\\n\\n' + message['content'] %}{% else %}{% set content = message['content'] %}{% endif %}{% if message['role'] == 'user' %}{{ bos_token + '[INST] ' + content.strip() + ' [/INST]' }}{% elif message['role'] == 'assistant' %}{{ ' ' + content.strip() + ' ' + eos_token }}{% endif %}{% endfor %}",
|
32 |
+
"clean_up_tokenization_spaces": false,
|
33 |
+
"eos_token": "</s>",
|
34 |
+
"legacy": false,
|
35 |
+
"model_max_length": 1000000000000000019884624838656,
|
36 |
+
"pad_token": "</s>",
|
37 |
+
"padding_side": "right",
|
38 |
+
"sp_model_kwargs": {},
|
39 |
+
"tokenizer_class": "LlamaTokenizer",
|
40 |
+
"unk_token": "<unk>",
|
41 |
+
"use_default_system_prompt": false
|
42 |
+
}
|
training_args.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:bf62c053ab3ff182f15262d4beb891b4d1ce3a524d83c6987386d6d7a82fd48c
|
3 |
+
size 4728
|