Upload folder using huggingface_hub
Browse files- README.md +58 -0
- adapter_config.json +31 -0
- adapter_model.safetensors +3 -0
- all_results.json +9 -0
- checkpoint-62/README.md +202 -0
- checkpoint-62/adapter_config.json +31 -0
- checkpoint-62/adapter_model.safetensors +3 -0
- checkpoint-62/optimizer.pt +3 -0
- checkpoint-62/qwen.tiktoken +0 -0
- checkpoint-62/rng_state.pth +3 -0
- checkpoint-62/scheduler.pt +3 -0
- checkpoint-62/special_tokens_map.json +10 -0
- checkpoint-62/tokenization_qwen.py +276 -0
- checkpoint-62/tokenizer_config.json +17 -0
- checkpoint-62/trainer_state.json +129 -0
- checkpoint-62/training_args.bin +3 -0
- llamaboard_config.yaml +66 -0
- qwen.tiktoken +0 -0
- running_log.txt +2 -0
- special_tokens_map.json +10 -0
- tokenization_qwen.py +276 -0
- tokenizer_config.json +17 -0
- train_results.json +9 -0
- trainer_log.jsonl +13 -0
- trainer_state.json +139 -0
- training_args.bin +3 -0
- training_args.yaml +32 -0
- training_loss.png +0 -0
README.md
ADDED
@@ -0,0 +1,58 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
base_model: qwen/Qwen-1_8B-Chat
|
3 |
+
library_name: peft
|
4 |
+
license: other
|
5 |
+
tags:
|
6 |
+
- llama-factory
|
7 |
+
- lora
|
8 |
+
- generated_from_trainer
|
9 |
+
model-index:
|
10 |
+
- name: train_2024-08-31-17-40-34
|
11 |
+
results: []
|
12 |
+
---
|
13 |
+
|
14 |
+
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
15 |
+
should probably proofread and complete it, then remove this comment. -->
|
16 |
+
|
17 |
+
# train_2024-08-31-17-40-34
|
18 |
+
|
19 |
+
This model is a fine-tuned version of [qwen/Qwen-1_8B-Chat](https://huggingface.co/qwen/Qwen-1_8B-Chat) on the glaive_toolcall_en dataset.
|
20 |
+
|
21 |
+
## Model description
|
22 |
+
|
23 |
+
More information needed
|
24 |
+
|
25 |
+
## Intended uses & limitations
|
26 |
+
|
27 |
+
More information needed
|
28 |
+
|
29 |
+
## Training and evaluation data
|
30 |
+
|
31 |
+
More information needed
|
32 |
+
|
33 |
+
## Training procedure
|
34 |
+
|
35 |
+
### Training hyperparameters
|
36 |
+
|
37 |
+
The following hyperparameters were used during training:
|
38 |
+
- learning_rate: 5e-05
|
39 |
+
- train_batch_size: 2
|
40 |
+
- eval_batch_size: 8
|
41 |
+
- seed: 42
|
42 |
+
- gradient_accumulation_steps: 8
|
43 |
+
- total_train_batch_size: 16
|
44 |
+
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
45 |
+
- lr_scheduler_type: cosine
|
46 |
+
- num_epochs: 1.0
|
47 |
+
|
48 |
+
### Training results
|
49 |
+
|
50 |
+
|
51 |
+
|
52 |
+
### Framework versions
|
53 |
+
|
54 |
+
- PEFT 0.12.0
|
55 |
+
- Transformers 4.44.2
|
56 |
+
- Pytorch 2.4.0
|
57 |
+
- Datasets 2.21.0
|
58 |
+
- Tokenizers 0.19.1
|
adapter_config.json
ADDED
@@ -0,0 +1,31 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"alpha_pattern": {},
|
3 |
+
"auto_mapping": null,
|
4 |
+
"base_model_name_or_path": "qwen/Qwen-1_8B-Chat",
|
5 |
+
"bias": "none",
|
6 |
+
"fan_in_fan_out": false,
|
7 |
+
"inference_mode": true,
|
8 |
+
"init_lora_weights": true,
|
9 |
+
"layer_replication": null,
|
10 |
+
"layers_pattern": null,
|
11 |
+
"layers_to_transform": null,
|
12 |
+
"loftq_config": {},
|
13 |
+
"lora_alpha": 16,
|
14 |
+
"lora_dropout": 0.0,
|
15 |
+
"megatron_config": null,
|
16 |
+
"megatron_core": "megatron.core",
|
17 |
+
"modules_to_save": null,
|
18 |
+
"peft_type": "LORA",
|
19 |
+
"r": 4,
|
20 |
+
"rank_pattern": {},
|
21 |
+
"revision": null,
|
22 |
+
"target_modules": [
|
23 |
+
"c_proj",
|
24 |
+
"c_attn",
|
25 |
+
"w1",
|
26 |
+
"w2"
|
27 |
+
],
|
28 |
+
"task_type": "CAUSAL_LM",
|
29 |
+
"use_dora": false,
|
30 |
+
"use_rslora": false
|
31 |
+
}
|
adapter_model.safetensors
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:bfd807cad2fb36649906b71719516e895d388c2396c7d72a4e31387147025c7f
|
3 |
+
size 13448712
|
all_results.json
ADDED
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"epoch": 0.992,
|
3 |
+
"num_input_tokens_seen": 610016,
|
4 |
+
"total_flos": 5596354012643328.0,
|
5 |
+
"train_loss": 0.6199409730972782,
|
6 |
+
"train_runtime": 3421.6329,
|
7 |
+
"train_samples_per_second": 0.292,
|
8 |
+
"train_steps_per_second": 0.018
|
9 |
+
}
|
checkpoint-62/README.md
ADDED
@@ -0,0 +1,202 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
base_model: qwen/Qwen-1_8B-Chat
|
3 |
+
library_name: peft
|
4 |
+
---
|
5 |
+
|
6 |
+
# Model Card for Model ID
|
7 |
+
|
8 |
+
<!-- Provide a quick summary of what the model is/does. -->
|
9 |
+
|
10 |
+
|
11 |
+
|
12 |
+
## Model Details
|
13 |
+
|
14 |
+
### Model Description
|
15 |
+
|
16 |
+
<!-- Provide a longer summary of what this model is. -->
|
17 |
+
|
18 |
+
|
19 |
+
|
20 |
+
- **Developed by:** [More Information Needed]
|
21 |
+
- **Funded by [optional]:** [More Information Needed]
|
22 |
+
- **Shared by [optional]:** [More Information Needed]
|
23 |
+
- **Model type:** [More Information Needed]
|
24 |
+
- **Language(s) (NLP):** [More Information Needed]
|
25 |
+
- **License:** [More Information Needed]
|
26 |
+
- **Finetuned from model [optional]:** [More Information Needed]
|
27 |
+
|
28 |
+
### Model Sources [optional]
|
29 |
+
|
30 |
+
<!-- Provide the basic links for the model. -->
|
31 |
+
|
32 |
+
- **Repository:** [More Information Needed]
|
33 |
+
- **Paper [optional]:** [More Information Needed]
|
34 |
+
- **Demo [optional]:** [More Information Needed]
|
35 |
+
|
36 |
+
## Uses
|
37 |
+
|
38 |
+
<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
|
39 |
+
|
40 |
+
### Direct Use
|
41 |
+
|
42 |
+
<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
|
43 |
+
|
44 |
+
[More Information Needed]
|
45 |
+
|
46 |
+
### Downstream Use [optional]
|
47 |
+
|
48 |
+
<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
|
49 |
+
|
50 |
+
[More Information Needed]
|
51 |
+
|
52 |
+
### Out-of-Scope Use
|
53 |
+
|
54 |
+
<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
|
55 |
+
|
56 |
+
[More Information Needed]
|
57 |
+
|
58 |
+
## Bias, Risks, and Limitations
|
59 |
+
|
60 |
+
<!-- This section is meant to convey both technical and sociotechnical limitations. -->
|
61 |
+
|
62 |
+
[More Information Needed]
|
63 |
+
|
64 |
+
### Recommendations
|
65 |
+
|
66 |
+
<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
|
67 |
+
|
68 |
+
Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
|
69 |
+
|
70 |
+
## How to Get Started with the Model
|
71 |
+
|
72 |
+
Use the code below to get started with the model.
|
73 |
+
|
74 |
+
[More Information Needed]
|
75 |
+
|
76 |
+
## Training Details
|
77 |
+
|
78 |
+
### Training Data
|
79 |
+
|
80 |
+
<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
|
81 |
+
|
82 |
+
[More Information Needed]
|
83 |
+
|
84 |
+
### Training Procedure
|
85 |
+
|
86 |
+
<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
|
87 |
+
|
88 |
+
#### Preprocessing [optional]
|
89 |
+
|
90 |
+
[More Information Needed]
|
91 |
+
|
92 |
+
|
93 |
+
#### Training Hyperparameters
|
94 |
+
|
95 |
+
- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
|
96 |
+
|
97 |
+
#### Speeds, Sizes, Times [optional]
|
98 |
+
|
99 |
+
<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
|
100 |
+
|
101 |
+
[More Information Needed]
|
102 |
+
|
103 |
+
## Evaluation
|
104 |
+
|
105 |
+
<!-- This section describes the evaluation protocols and provides the results. -->
|
106 |
+
|
107 |
+
### Testing Data, Factors & Metrics
|
108 |
+
|
109 |
+
#### Testing Data
|
110 |
+
|
111 |
+
<!-- This should link to a Dataset Card if possible. -->
|
112 |
+
|
113 |
+
[More Information Needed]
|
114 |
+
|
115 |
+
#### Factors
|
116 |
+
|
117 |
+
<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
|
118 |
+
|
119 |
+
[More Information Needed]
|
120 |
+
|
121 |
+
#### Metrics
|
122 |
+
|
123 |
+
<!-- These are the evaluation metrics being used, ideally with a description of why. -->
|
124 |
+
|
125 |
+
[More Information Needed]
|
126 |
+
|
127 |
+
### Results
|
128 |
+
|
129 |
+
[More Information Needed]
|
130 |
+
|
131 |
+
#### Summary
|
132 |
+
|
133 |
+
|
134 |
+
|
135 |
+
## Model Examination [optional]
|
136 |
+
|
137 |
+
<!-- Relevant interpretability work for the model goes here -->
|
138 |
+
|
139 |
+
[More Information Needed]
|
140 |
+
|
141 |
+
## Environmental Impact
|
142 |
+
|
143 |
+
<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
|
144 |
+
|
145 |
+
Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
|
146 |
+
|
147 |
+
- **Hardware Type:** [More Information Needed]
|
148 |
+
- **Hours used:** [More Information Needed]
|
149 |
+
- **Cloud Provider:** [More Information Needed]
|
150 |
+
- **Compute Region:** [More Information Needed]
|
151 |
+
- **Carbon Emitted:** [More Information Needed]
|
152 |
+
|
153 |
+
## Technical Specifications [optional]
|
154 |
+
|
155 |
+
### Model Architecture and Objective
|
156 |
+
|
157 |
+
[More Information Needed]
|
158 |
+
|
159 |
+
### Compute Infrastructure
|
160 |
+
|
161 |
+
[More Information Needed]
|
162 |
+
|
163 |
+
#### Hardware
|
164 |
+
|
165 |
+
[More Information Needed]
|
166 |
+
|
167 |
+
#### Software
|
168 |
+
|
169 |
+
[More Information Needed]
|
170 |
+
|
171 |
+
## Citation [optional]
|
172 |
+
|
173 |
+
<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
|
174 |
+
|
175 |
+
**BibTeX:**
|
176 |
+
|
177 |
+
[More Information Needed]
|
178 |
+
|
179 |
+
**APA:**
|
180 |
+
|
181 |
+
[More Information Needed]
|
182 |
+
|
183 |
+
## Glossary [optional]
|
184 |
+
|
185 |
+
<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
|
186 |
+
|
187 |
+
[More Information Needed]
|
188 |
+
|
189 |
+
## More Information [optional]
|
190 |
+
|
191 |
+
[More Information Needed]
|
192 |
+
|
193 |
+
## Model Card Authors [optional]
|
194 |
+
|
195 |
+
[More Information Needed]
|
196 |
+
|
197 |
+
## Model Card Contact
|
198 |
+
|
199 |
+
[More Information Needed]
|
200 |
+
### Framework versions
|
201 |
+
|
202 |
+
- PEFT 0.12.0
|
checkpoint-62/adapter_config.json
ADDED
@@ -0,0 +1,31 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"alpha_pattern": {},
|
3 |
+
"auto_mapping": null,
|
4 |
+
"base_model_name_or_path": "qwen/Qwen-1_8B-Chat",
|
5 |
+
"bias": "none",
|
6 |
+
"fan_in_fan_out": false,
|
7 |
+
"inference_mode": true,
|
8 |
+
"init_lora_weights": true,
|
9 |
+
"layer_replication": null,
|
10 |
+
"layers_pattern": null,
|
11 |
+
"layers_to_transform": null,
|
12 |
+
"loftq_config": {},
|
13 |
+
"lora_alpha": 16,
|
14 |
+
"lora_dropout": 0.0,
|
15 |
+
"megatron_config": null,
|
16 |
+
"megatron_core": "megatron.core",
|
17 |
+
"modules_to_save": null,
|
18 |
+
"peft_type": "LORA",
|
19 |
+
"r": 4,
|
20 |
+
"rank_pattern": {},
|
21 |
+
"revision": null,
|
22 |
+
"target_modules": [
|
23 |
+
"c_proj",
|
24 |
+
"c_attn",
|
25 |
+
"w1",
|
26 |
+
"w2"
|
27 |
+
],
|
28 |
+
"task_type": "CAUSAL_LM",
|
29 |
+
"use_dora": false,
|
30 |
+
"use_rslora": false
|
31 |
+
}
|
checkpoint-62/adapter_model.safetensors
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:bfd807cad2fb36649906b71719516e895d388c2396c7d72a4e31387147025c7f
|
3 |
+
size 13448712
|
checkpoint-62/optimizer.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:e5c61e60e5fe23b7a52c5dfb4349c5c80facbe1c4b0565fa2418530e797c49a6
|
3 |
+
size 27031674
|
checkpoint-62/qwen.tiktoken
ADDED
The diff for this file is too large to render.
See raw diff
|
|
checkpoint-62/rng_state.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:a74cb2999fd09c30c2676c95a55b375947ad04fa23df46ea458fa59f07eaee5c
|
3 |
+
size 13990
|
checkpoint-62/scheduler.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:d65ff43cc7d5ac74f560f6c36d19c34fda721bbf1bacbbdd9237f934985640c0
|
3 |
+
size 1064
|
checkpoint-62/special_tokens_map.json
ADDED
@@ -0,0 +1,10 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"eos_token": {
|
3 |
+
"content": "<|im_end|>",
|
4 |
+
"lstrip": false,
|
5 |
+
"normalized": false,
|
6 |
+
"rstrip": false,
|
7 |
+
"single_word": false
|
8 |
+
},
|
9 |
+
"pad_token": "<|im_end|>"
|
10 |
+
}
|
checkpoint-62/tokenization_qwen.py
ADDED
@@ -0,0 +1,276 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# Copyright (c) Alibaba Cloud.
|
2 |
+
#
|
3 |
+
# This source code is licensed under the license found in the
|
4 |
+
# LICENSE file in the root directory of this source tree.
|
5 |
+
|
6 |
+
"""Tokenization classes for QWen."""
|
7 |
+
|
8 |
+
import base64
|
9 |
+
import logging
|
10 |
+
import os
|
11 |
+
import unicodedata
|
12 |
+
from typing import Collection, Dict, List, Set, Tuple, Union
|
13 |
+
|
14 |
+
import tiktoken
|
15 |
+
from transformers import PreTrainedTokenizer, AddedToken
|
16 |
+
|
17 |
+
logger = logging.getLogger(__name__)
|
18 |
+
|
19 |
+
|
20 |
+
VOCAB_FILES_NAMES = {"vocab_file": "qwen.tiktoken"}
|
21 |
+
|
22 |
+
PAT_STR = r"""(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\r\n\p{L}\p{N}]?\p{L}+|\p{N}| ?[^\s\p{L}\p{N}]+[\r\n]*|\s*[\r\n]+|\s+(?!\S)|\s+"""
|
23 |
+
ENDOFTEXT = "<|endoftext|>"
|
24 |
+
IMSTART = "<|im_start|>"
|
25 |
+
IMEND = "<|im_end|>"
|
26 |
+
# as the default behavior is changed to allow special tokens in
|
27 |
+
# regular texts, the surface forms of special tokens need to be
|
28 |
+
# as different as possible to minimize the impact
|
29 |
+
EXTRAS = tuple((f"<|extra_{i}|>" for i in range(205)))
|
30 |
+
# changed to use actual index to avoid misconfiguration with vocabulary expansion
|
31 |
+
SPECIAL_START_ID = 151643
|
32 |
+
SPECIAL_TOKENS = tuple(
|
33 |
+
enumerate(
|
34 |
+
(
|
35 |
+
(
|
36 |
+
ENDOFTEXT,
|
37 |
+
IMSTART,
|
38 |
+
IMEND,
|
39 |
+
)
|
40 |
+
+ EXTRAS
|
41 |
+
),
|
42 |
+
start=SPECIAL_START_ID,
|
43 |
+
)
|
44 |
+
)
|
45 |
+
SPECIAL_TOKENS_SET = set(t for i, t in SPECIAL_TOKENS)
|
46 |
+
|
47 |
+
|
48 |
+
def _load_tiktoken_bpe(tiktoken_bpe_file: str) -> Dict[bytes, int]:
|
49 |
+
with open(tiktoken_bpe_file, "rb") as f:
|
50 |
+
contents = f.read()
|
51 |
+
return {
|
52 |
+
base64.b64decode(token): int(rank)
|
53 |
+
for token, rank in (line.split() for line in contents.splitlines() if line)
|
54 |
+
}
|
55 |
+
|
56 |
+
|
57 |
+
class QWenTokenizer(PreTrainedTokenizer):
|
58 |
+
"""QWen tokenizer."""
|
59 |
+
|
60 |
+
vocab_files_names = VOCAB_FILES_NAMES
|
61 |
+
|
62 |
+
def __init__(
|
63 |
+
self,
|
64 |
+
vocab_file,
|
65 |
+
errors="replace",
|
66 |
+
extra_vocab_file=None,
|
67 |
+
**kwargs,
|
68 |
+
):
|
69 |
+
super().__init__(**kwargs)
|
70 |
+
|
71 |
+
# how to handle errors in decoding UTF-8 byte sequences
|
72 |
+
# use ignore if you are in streaming inference
|
73 |
+
self.errors = errors
|
74 |
+
|
75 |
+
self.mergeable_ranks = _load_tiktoken_bpe(vocab_file) # type: Dict[bytes, int]
|
76 |
+
self.special_tokens = {
|
77 |
+
token: index
|
78 |
+
for index, token in SPECIAL_TOKENS
|
79 |
+
}
|
80 |
+
|
81 |
+
# try load extra vocab from file
|
82 |
+
if extra_vocab_file is not None:
|
83 |
+
used_ids = set(self.mergeable_ranks.values()) | set(self.special_tokens.values())
|
84 |
+
extra_mergeable_ranks = _load_tiktoken_bpe(extra_vocab_file)
|
85 |
+
for token, index in extra_mergeable_ranks.items():
|
86 |
+
if token in self.mergeable_ranks:
|
87 |
+
logger.info(f"extra token {token} exists, skipping")
|
88 |
+
continue
|
89 |
+
if index in used_ids:
|
90 |
+
logger.info(f'the index {index} for extra token {token} exists, skipping')
|
91 |
+
continue
|
92 |
+
self.mergeable_ranks[token] = index
|
93 |
+
# the index may be sparse after this, but don't worry tiktoken.Encoding will handle this
|
94 |
+
|
95 |
+
enc = tiktoken.Encoding(
|
96 |
+
"Qwen",
|
97 |
+
pat_str=PAT_STR,
|
98 |
+
mergeable_ranks=self.mergeable_ranks,
|
99 |
+
special_tokens=self.special_tokens,
|
100 |
+
)
|
101 |
+
assert (
|
102 |
+
len(self.mergeable_ranks) + len(self.special_tokens) == enc.n_vocab
|
103 |
+
), f"{len(self.mergeable_ranks) + len(self.special_tokens)} != {enc.n_vocab} in encoding"
|
104 |
+
|
105 |
+
self.decoder = {
|
106 |
+
v: k for k, v in self.mergeable_ranks.items()
|
107 |
+
} # type: dict[int, bytes|str]
|
108 |
+
self.decoder.update({v: k for k, v in self.special_tokens.items()})
|
109 |
+
|
110 |
+
self.tokenizer = enc # type: tiktoken.Encoding
|
111 |
+
|
112 |
+
self.eod_id = self.tokenizer.eot_token
|
113 |
+
self.im_start_id = self.special_tokens[IMSTART]
|
114 |
+
self.im_end_id = self.special_tokens[IMEND]
|
115 |
+
|
116 |
+
def __getstate__(self):
|
117 |
+
# for pickle lovers
|
118 |
+
state = self.__dict__.copy()
|
119 |
+
del state["tokenizer"]
|
120 |
+
return state
|
121 |
+
|
122 |
+
def __setstate__(self, state):
|
123 |
+
# tokenizer is not python native; don't pass it; rebuild it
|
124 |
+
self.__dict__.update(state)
|
125 |
+
enc = tiktoken.Encoding(
|
126 |
+
"Qwen",
|
127 |
+
pat_str=PAT_STR,
|
128 |
+
mergeable_ranks=self.mergeable_ranks,
|
129 |
+
special_tokens=self.special_tokens,
|
130 |
+
)
|
131 |
+
self.tokenizer = enc
|
132 |
+
|
133 |
+
def __len__(self) -> int:
|
134 |
+
return self.tokenizer.n_vocab
|
135 |
+
|
136 |
+
def get_vocab(self) -> Dict[bytes, int]:
|
137 |
+
return self.mergeable_ranks
|
138 |
+
|
139 |
+
def convert_tokens_to_ids(
|
140 |
+
self, tokens: Union[bytes, str, List[Union[bytes, str]]]
|
141 |
+
) -> List[int]:
|
142 |
+
ids = []
|
143 |
+
if isinstance(tokens, (str, bytes)):
|
144 |
+
if tokens in self.special_tokens:
|
145 |
+
return self.special_tokens[tokens]
|
146 |
+
else:
|
147 |
+
return self.mergeable_ranks.get(tokens)
|
148 |
+
for token in tokens:
|
149 |
+
if token in self.special_tokens:
|
150 |
+
ids.append(self.special_tokens[token])
|
151 |
+
else:
|
152 |
+
ids.append(self.mergeable_ranks.get(token))
|
153 |
+
return ids
|
154 |
+
|
155 |
+
def _add_tokens(
|
156 |
+
self,
|
157 |
+
new_tokens: Union[List[str], List[AddedToken]],
|
158 |
+
special_tokens: bool = False,
|
159 |
+
) -> int:
|
160 |
+
if not special_tokens and new_tokens:
|
161 |
+
raise ValueError("Adding regular tokens is not supported")
|
162 |
+
for token in new_tokens:
|
163 |
+
surface_form = token.content if isinstance(token, AddedToken) else token
|
164 |
+
if surface_form not in SPECIAL_TOKENS_SET:
|
165 |
+
raise ValueError("Adding unknown special tokens is not supported")
|
166 |
+
return 0
|
167 |
+
|
168 |
+
def save_vocabulary(self, save_directory: str, **kwargs) -> Tuple[str]:
|
169 |
+
"""
|
170 |
+
Save only the vocabulary of the tokenizer (vocabulary).
|
171 |
+
|
172 |
+
Returns:
|
173 |
+
`Tuple(str)`: Paths to the files saved.
|
174 |
+
"""
|
175 |
+
file_path = os.path.join(save_directory, "qwen.tiktoken")
|
176 |
+
with open(file_path, "w", encoding="utf8") as w:
|
177 |
+
for k, v in self.mergeable_ranks.items():
|
178 |
+
line = base64.b64encode(k).decode("utf8") + " " + str(v) + "\n"
|
179 |
+
w.write(line)
|
180 |
+
return (file_path,)
|
181 |
+
|
182 |
+
def tokenize(
|
183 |
+
self,
|
184 |
+
text: str,
|
185 |
+
allowed_special: Union[Set, str] = "all",
|
186 |
+
disallowed_special: Union[Collection, str] = (),
|
187 |
+
**kwargs,
|
188 |
+
) -> List[Union[bytes, str]]:
|
189 |
+
"""
|
190 |
+
Converts a string in a sequence of tokens.
|
191 |
+
|
192 |
+
Args:
|
193 |
+
text (`str`):
|
194 |
+
The sequence to be encoded.
|
195 |
+
allowed_special (`Literal["all"]` or `set`):
|
196 |
+
The surface forms of the tokens to be encoded as special tokens in regular texts.
|
197 |
+
Default to "all".
|
198 |
+
disallowed_special (`Literal["all"]` or `Collection`):
|
199 |
+
The surface forms of the tokens that should not be in regular texts and trigger errors.
|
200 |
+
Default to an empty tuple.
|
201 |
+
|
202 |
+
kwargs (additional keyword arguments, *optional*):
|
203 |
+
Will be passed to the underlying model specific encode method.
|
204 |
+
|
205 |
+
Returns:
|
206 |
+
`List[bytes|str]`: The list of tokens.
|
207 |
+
"""
|
208 |
+
tokens = []
|
209 |
+
text = unicodedata.normalize("NFC", text)
|
210 |
+
|
211 |
+
# this implementation takes a detour: text -> token id -> token surface forms
|
212 |
+
for t in self.tokenizer.encode(
|
213 |
+
text, allowed_special=allowed_special, disallowed_special=disallowed_special
|
214 |
+
):
|
215 |
+
tokens.append(self.decoder[t])
|
216 |
+
return tokens
|
217 |
+
|
218 |
+
def convert_tokens_to_string(self, tokens: List[Union[bytes, str]]) -> str:
|
219 |
+
"""
|
220 |
+
Converts a sequence of tokens in a single string.
|
221 |
+
"""
|
222 |
+
text = ""
|
223 |
+
temp = b""
|
224 |
+
for t in tokens:
|
225 |
+
if isinstance(t, str):
|
226 |
+
if temp:
|
227 |
+
text += temp.decode("utf-8", errors=self.errors)
|
228 |
+
temp = b""
|
229 |
+
text += t
|
230 |
+
elif isinstance(t, bytes):
|
231 |
+
temp += t
|
232 |
+
else:
|
233 |
+
raise TypeError("token should only be of type types or str")
|
234 |
+
if temp:
|
235 |
+
text += temp.decode("utf-8", errors=self.errors)
|
236 |
+
return text
|
237 |
+
|
238 |
+
@property
|
239 |
+
def vocab_size(self):
|
240 |
+
return self.tokenizer.n_vocab
|
241 |
+
|
242 |
+
def _convert_id_to_token(self, index: int) -> Union[bytes, str]:
|
243 |
+
"""Converts an id to a token, special tokens included"""
|
244 |
+
if index in self.decoder:
|
245 |
+
return self.decoder[index]
|
246 |
+
raise ValueError("unknown ids")
|
247 |
+
|
248 |
+
def _convert_token_to_id(self, token: Union[bytes, str]) -> int:
|
249 |
+
"""Converts a token to an id using the vocab, special tokens included"""
|
250 |
+
if token in self.special_tokens:
|
251 |
+
return self.special_tokens[token]
|
252 |
+
if token in self.mergeable_ranks:
|
253 |
+
return self.mergeable_ranks[token]
|
254 |
+
raise ValueError("unknown token")
|
255 |
+
|
256 |
+
def _tokenize(self, text: str, **kwargs):
|
257 |
+
"""
|
258 |
+
Converts a string in a sequence of tokens (string), using the tokenizer. Split in words for word-based
|
259 |
+
vocabulary or sub-words for sub-word-based vocabularies (BPE/SentencePieces/WordPieces).
|
260 |
+
|
261 |
+
Do NOT take care of added tokens.
|
262 |
+
"""
|
263 |
+
raise NotImplementedError
|
264 |
+
|
265 |
+
def _decode(
|
266 |
+
self,
|
267 |
+
token_ids: Union[int, List[int]],
|
268 |
+
skip_special_tokens: bool = False,
|
269 |
+
errors: str = None,
|
270 |
+
**kwargs,
|
271 |
+
) -> str:
|
272 |
+
if isinstance(token_ids, int):
|
273 |
+
token_ids = [token_ids]
|
274 |
+
if skip_special_tokens:
|
275 |
+
token_ids = [i for i in token_ids if i < self.eod_id]
|
276 |
+
return self.tokenizer.decode(token_ids, errors=errors or self.errors)
|
checkpoint-62/tokenizer_config.json
ADDED
@@ -0,0 +1,17 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"added_tokens_decoder": {},
|
3 |
+
"auto_map": {
|
4 |
+
"AutoTokenizer": [
|
5 |
+
"tokenization_qwen.QWenTokenizer",
|
6 |
+
null
|
7 |
+
]
|
8 |
+
},
|
9 |
+
"chat_template": "{% set system_message = 'You are a helpful assistant.' %}{% if messages[0]['role'] == 'system' %}{% set loop_messages = messages[1:] %}{% set system_message = messages[0]['content'] %}{% else %}{% set loop_messages = messages %}{% endif %}{% if system_message is defined %}{{ '<|im_start|>system\n' + system_message + '<|im_end|>\n' }}{% endif %}{% for message in loop_messages %}{% set content = message['content'] %}{% if message['role'] == 'user' %}{{ '<|im_start|>user\n' + content + '<|im_end|>\n<|im_start|>assistant\n' }}{% elif message['role'] == 'assistant' %}{{ content + '<|im_end|>' + '\n' }}{% endif %}{% endfor %}",
|
10 |
+
"clean_up_tokenization_spaces": true,
|
11 |
+
"eos_token": "<|im_end|>",
|
12 |
+
"model_max_length": 8192,
|
13 |
+
"pad_token": "<|im_end|>",
|
14 |
+
"padding_side": "right",
|
15 |
+
"split_special_tokens": false,
|
16 |
+
"tokenizer_class": "QWenTokenizer"
|
17 |
+
}
|
checkpoint-62/trainer_state.json
ADDED
@@ -0,0 +1,129 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"best_metric": null,
|
3 |
+
"best_model_checkpoint": null,
|
4 |
+
"epoch": 0.992,
|
5 |
+
"eval_steps": 500,
|
6 |
+
"global_step": 62,
|
7 |
+
"is_hyper_param_search": false,
|
8 |
+
"is_local_process_zero": true,
|
9 |
+
"is_world_process_zero": true,
|
10 |
+
"log_history": [
|
11 |
+
{
|
12 |
+
"epoch": 0.08,
|
13 |
+
"grad_norm": 0.7084760069847107,
|
14 |
+
"learning_rate": 4.920192797165511e-05,
|
15 |
+
"loss": 0.8189,
|
16 |
+
"num_input_tokens_seen": 54176,
|
17 |
+
"step": 5
|
18 |
+
},
|
19 |
+
{
|
20 |
+
"epoch": 0.16,
|
21 |
+
"grad_norm": 0.7297347784042358,
|
22 |
+
"learning_rate": 4.685866540361456e-05,
|
23 |
+
"loss": 0.6742,
|
24 |
+
"num_input_tokens_seen": 102528,
|
25 |
+
"step": 10
|
26 |
+
},
|
27 |
+
{
|
28 |
+
"epoch": 0.24,
|
29 |
+
"grad_norm": 684049.0625,
|
30 |
+
"learning_rate": 4.3119819680728e-05,
|
31 |
+
"loss": 0.6816,
|
32 |
+
"num_input_tokens_seen": 152912,
|
33 |
+
"step": 15
|
34 |
+
},
|
35 |
+
{
|
36 |
+
"epoch": 0.32,
|
37 |
+
"grad_norm": 0.5369439125061035,
|
38 |
+
"learning_rate": 3.822410025817406e-05,
|
39 |
+
"loss": 0.6588,
|
40 |
+
"num_input_tokens_seen": 204432,
|
41 |
+
"step": 20
|
42 |
+
},
|
43 |
+
{
|
44 |
+
"epoch": 0.4,
|
45 |
+
"grad_norm": 0.5415367484092712,
|
46 |
+
"learning_rate": 3.2484078074333954e-05,
|
47 |
+
"loss": 0.5882,
|
48 |
+
"num_input_tokens_seen": 254512,
|
49 |
+
"step": 25
|
50 |
+
},
|
51 |
+
{
|
52 |
+
"epoch": 0.48,
|
53 |
+
"grad_norm": 0.7139325737953186,
|
54 |
+
"learning_rate": 2.6266229220967818e-05,
|
55 |
+
"loss": 0.6473,
|
56 |
+
"num_input_tokens_seen": 298608,
|
57 |
+
"step": 30
|
58 |
+
},
|
59 |
+
{
|
60 |
+
"epoch": 0.56,
|
61 |
+
"grad_norm": 2127319.0,
|
62 |
+
"learning_rate": 1.9967536997783494e-05,
|
63 |
+
"loss": 0.6343,
|
64 |
+
"num_input_tokens_seen": 348928,
|
65 |
+
"step": 35
|
66 |
+
},
|
67 |
+
{
|
68 |
+
"epoch": 0.64,
|
69 |
+
"grad_norm": 5908939.5,
|
70 |
+
"learning_rate": 1.399014621105914e-05,
|
71 |
+
"loss": 0.5067,
|
72 |
+
"num_input_tokens_seen": 397216,
|
73 |
+
"step": 40
|
74 |
+
},
|
75 |
+
{
|
76 |
+
"epoch": 0.72,
|
77 |
+
"grad_norm": 1202378.875,
|
78 |
+
"learning_rate": 8.715687931944449e-06,
|
79 |
+
"loss": 0.5386,
|
80 |
+
"num_input_tokens_seen": 444832,
|
81 |
+
"step": 45
|
82 |
+
},
|
83 |
+
{
|
84 |
+
"epoch": 0.8,
|
85 |
+
"grad_norm": 0.4307733178138733,
|
86 |
+
"learning_rate": 4.480913969818098e-06,
|
87 |
+
"loss": 0.563,
|
88 |
+
"num_input_tokens_seen": 490640,
|
89 |
+
"step": 50
|
90 |
+
},
|
91 |
+
{
|
92 |
+
"epoch": 0.88,
|
93 |
+
"grad_norm": 11058075.0,
|
94 |
+
"learning_rate": 1.5561966963229924e-06,
|
95 |
+
"loss": 0.5484,
|
96 |
+
"num_input_tokens_seen": 538400,
|
97 |
+
"step": 55
|
98 |
+
},
|
99 |
+
{
|
100 |
+
"epoch": 0.96,
|
101 |
+
"grad_norm": 2679230.5,
|
102 |
+
"learning_rate": 1.2826691520262114e-07,
|
103 |
+
"loss": 0.5979,
|
104 |
+
"num_input_tokens_seen": 589856,
|
105 |
+
"step": 60
|
106 |
+
}
|
107 |
+
],
|
108 |
+
"logging_steps": 5,
|
109 |
+
"max_steps": 62,
|
110 |
+
"num_input_tokens_seen": 610016,
|
111 |
+
"num_train_epochs": 1,
|
112 |
+
"save_steps": 100,
|
113 |
+
"stateful_callbacks": {
|
114 |
+
"TrainerControl": {
|
115 |
+
"args": {
|
116 |
+
"should_epoch_stop": false,
|
117 |
+
"should_evaluate": false,
|
118 |
+
"should_log": false,
|
119 |
+
"should_save": true,
|
120 |
+
"should_training_stop": true
|
121 |
+
},
|
122 |
+
"attributes": {}
|
123 |
+
}
|
124 |
+
},
|
125 |
+
"total_flos": 5596354012643328.0,
|
126 |
+
"train_batch_size": 2,
|
127 |
+
"trial_name": null,
|
128 |
+
"trial_params": null
|
129 |
+
}
|
checkpoint-62/training_args.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:311875b8ca25e2de38752c7cd348177c9ca8d80ec40f9c44f87bd2bc51a3e94b
|
3 |
+
size 5368
|
llamaboard_config.yaml
ADDED
@@ -0,0 +1,66 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
top.booster: auto
|
2 |
+
top.checkpoint_path: []
|
3 |
+
top.finetuning_type: lora
|
4 |
+
top.model_name: Qwen-1.8B-Chat
|
5 |
+
top.quantization_bit: none
|
6 |
+
top.quantization_method: bitsandbytes
|
7 |
+
top.rope_scaling: none
|
8 |
+
top.template: qwen
|
9 |
+
train.additional_target: ''
|
10 |
+
train.badam_mode: layer
|
11 |
+
train.badam_switch_interval: 50
|
12 |
+
train.badam_switch_mode: ascending
|
13 |
+
train.badam_update_ratio: 0.05
|
14 |
+
train.batch_size: 2
|
15 |
+
train.compute_type: bf16
|
16 |
+
train.create_new_adapter: false
|
17 |
+
train.cutoff_len: 1024
|
18 |
+
train.dataset:
|
19 |
+
- glaive_toolcall_en
|
20 |
+
train.dataset_dir: data
|
21 |
+
train.ds_offload: false
|
22 |
+
train.ds_stage: none
|
23 |
+
train.freeze_extra_modules: ''
|
24 |
+
train.freeze_trainable_layers: 2
|
25 |
+
train.freeze_trainable_modules: all
|
26 |
+
train.galore_rank: 16
|
27 |
+
train.galore_scale: 0.25
|
28 |
+
train.galore_target: all
|
29 |
+
train.galore_update_interval: 200
|
30 |
+
train.gradient_accumulation_steps: 8
|
31 |
+
train.learning_rate: 5e-5
|
32 |
+
train.logging_steps: 5
|
33 |
+
train.lora_alpha: 16
|
34 |
+
train.lora_dropout: 0
|
35 |
+
train.lora_rank: 8
|
36 |
+
train.lora_target: ''
|
37 |
+
train.loraplus_lr_ratio: 0
|
38 |
+
train.lr_scheduler_type: cosine
|
39 |
+
train.mask_history: false
|
40 |
+
train.max_grad_norm: '1.0'
|
41 |
+
train.max_samples: '100000'
|
42 |
+
train.neat_packing: false
|
43 |
+
train.neftune_alpha: 0
|
44 |
+
train.num_train_epochs: '3.0'
|
45 |
+
train.optim: adamw_torch
|
46 |
+
train.packing: false
|
47 |
+
train.ppo_score_norm: false
|
48 |
+
train.ppo_whiten_rewards: false
|
49 |
+
train.pref_beta: 0.1
|
50 |
+
train.pref_ftx: 0
|
51 |
+
train.pref_loss: sigmoid
|
52 |
+
train.report_to: false
|
53 |
+
train.resize_vocab: false
|
54 |
+
train.reward_model: null
|
55 |
+
train.save_steps: 100
|
56 |
+
train.shift_attn: false
|
57 |
+
train.train_on_prompt: false
|
58 |
+
train.training_stage: Supervised Fine-Tuning
|
59 |
+
train.use_badam: false
|
60 |
+
train.use_dora: false
|
61 |
+
train.use_galore: false
|
62 |
+
train.use_llama_pro: false
|
63 |
+
train.use_pissa: false
|
64 |
+
train.use_rslora: false
|
65 |
+
train.val_size: 0
|
66 |
+
train.warmup_steps: 0
|
qwen.tiktoken
ADDED
The diff for this file is too large to render.
See raw diff
|
|
running_log.txt
ADDED
@@ -0,0 +1,2 @@
|
|
|
|
|
|
|
1 |
+
[INFO|parser.py:352] 2024-08-31 17:53:54,714 >> Process rank: 0, device: mps, n_gpu: 1, distributed training: False, compute dtype: torch.bfloat16
|
2 |
+
|
special_tokens_map.json
ADDED
@@ -0,0 +1,10 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"eos_token": {
|
3 |
+
"content": "<|im_end|>",
|
4 |
+
"lstrip": false,
|
5 |
+
"normalized": false,
|
6 |
+
"rstrip": false,
|
7 |
+
"single_word": false
|
8 |
+
},
|
9 |
+
"pad_token": "<|im_end|>"
|
10 |
+
}
|
tokenization_qwen.py
ADDED
@@ -0,0 +1,276 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# Copyright (c) Alibaba Cloud.
|
2 |
+
#
|
3 |
+
# This source code is licensed under the license found in the
|
4 |
+
# LICENSE file in the root directory of this source tree.
|
5 |
+
|
6 |
+
"""Tokenization classes for QWen."""
|
7 |
+
|
8 |
+
import base64
|
9 |
+
import logging
|
10 |
+
import os
|
11 |
+
import unicodedata
|
12 |
+
from typing import Collection, Dict, List, Set, Tuple, Union
|
13 |
+
|
14 |
+
import tiktoken
|
15 |
+
from transformers import PreTrainedTokenizer, AddedToken
|
16 |
+
|
17 |
+
logger = logging.getLogger(__name__)
|
18 |
+
|
19 |
+
|
20 |
+
VOCAB_FILES_NAMES = {"vocab_file": "qwen.tiktoken"}
|
21 |
+
|
22 |
+
PAT_STR = r"""(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\r\n\p{L}\p{N}]?\p{L}+|\p{N}| ?[^\s\p{L}\p{N}]+[\r\n]*|\s*[\r\n]+|\s+(?!\S)|\s+"""
|
23 |
+
ENDOFTEXT = "<|endoftext|>"
|
24 |
+
IMSTART = "<|im_start|>"
|
25 |
+
IMEND = "<|im_end|>"
|
26 |
+
# as the default behavior is changed to allow special tokens in
|
27 |
+
# regular texts, the surface forms of special tokens need to be
|
28 |
+
# as different as possible to minimize the impact
|
29 |
+
EXTRAS = tuple((f"<|extra_{i}|>" for i in range(205)))
|
30 |
+
# changed to use actual index to avoid misconfiguration with vocabulary expansion
|
31 |
+
SPECIAL_START_ID = 151643
|
32 |
+
SPECIAL_TOKENS = tuple(
|
33 |
+
enumerate(
|
34 |
+
(
|
35 |
+
(
|
36 |
+
ENDOFTEXT,
|
37 |
+
IMSTART,
|
38 |
+
IMEND,
|
39 |
+
)
|
40 |
+
+ EXTRAS
|
41 |
+
),
|
42 |
+
start=SPECIAL_START_ID,
|
43 |
+
)
|
44 |
+
)
|
45 |
+
SPECIAL_TOKENS_SET = set(t for i, t in SPECIAL_TOKENS)
|
46 |
+
|
47 |
+
|
48 |
+
def _load_tiktoken_bpe(tiktoken_bpe_file: str) -> Dict[bytes, int]:
|
49 |
+
with open(tiktoken_bpe_file, "rb") as f:
|
50 |
+
contents = f.read()
|
51 |
+
return {
|
52 |
+
base64.b64decode(token): int(rank)
|
53 |
+
for token, rank in (line.split() for line in contents.splitlines() if line)
|
54 |
+
}
|
55 |
+
|
56 |
+
|
57 |
+
class QWenTokenizer(PreTrainedTokenizer):
|
58 |
+
"""QWen tokenizer."""
|
59 |
+
|
60 |
+
vocab_files_names = VOCAB_FILES_NAMES
|
61 |
+
|
62 |
+
def __init__(
|
63 |
+
self,
|
64 |
+
vocab_file,
|
65 |
+
errors="replace",
|
66 |
+
extra_vocab_file=None,
|
67 |
+
**kwargs,
|
68 |
+
):
|
69 |
+
super().__init__(**kwargs)
|
70 |
+
|
71 |
+
# how to handle errors in decoding UTF-8 byte sequences
|
72 |
+
# use ignore if you are in streaming inference
|
73 |
+
self.errors = errors
|
74 |
+
|
75 |
+
self.mergeable_ranks = _load_tiktoken_bpe(vocab_file) # type: Dict[bytes, int]
|
76 |
+
self.special_tokens = {
|
77 |
+
token: index
|
78 |
+
for index, token in SPECIAL_TOKENS
|
79 |
+
}
|
80 |
+
|
81 |
+
# try load extra vocab from file
|
82 |
+
if extra_vocab_file is not None:
|
83 |
+
used_ids = set(self.mergeable_ranks.values()) | set(self.special_tokens.values())
|
84 |
+
extra_mergeable_ranks = _load_tiktoken_bpe(extra_vocab_file)
|
85 |
+
for token, index in extra_mergeable_ranks.items():
|
86 |
+
if token in self.mergeable_ranks:
|
87 |
+
logger.info(f"extra token {token} exists, skipping")
|
88 |
+
continue
|
89 |
+
if index in used_ids:
|
90 |
+
logger.info(f'the index {index} for extra token {token} exists, skipping')
|
91 |
+
continue
|
92 |
+
self.mergeable_ranks[token] = index
|
93 |
+
# the index may be sparse after this, but don't worry tiktoken.Encoding will handle this
|
94 |
+
|
95 |
+
enc = tiktoken.Encoding(
|
96 |
+
"Qwen",
|
97 |
+
pat_str=PAT_STR,
|
98 |
+
mergeable_ranks=self.mergeable_ranks,
|
99 |
+
special_tokens=self.special_tokens,
|
100 |
+
)
|
101 |
+
assert (
|
102 |
+
len(self.mergeable_ranks) + len(self.special_tokens) == enc.n_vocab
|
103 |
+
), f"{len(self.mergeable_ranks) + len(self.special_tokens)} != {enc.n_vocab} in encoding"
|
104 |
+
|
105 |
+
self.decoder = {
|
106 |
+
v: k for k, v in self.mergeable_ranks.items()
|
107 |
+
} # type: dict[int, bytes|str]
|
108 |
+
self.decoder.update({v: k for k, v in self.special_tokens.items()})
|
109 |
+
|
110 |
+
self.tokenizer = enc # type: tiktoken.Encoding
|
111 |
+
|
112 |
+
self.eod_id = self.tokenizer.eot_token
|
113 |
+
self.im_start_id = self.special_tokens[IMSTART]
|
114 |
+
self.im_end_id = self.special_tokens[IMEND]
|
115 |
+
|
116 |
+
def __getstate__(self):
|
117 |
+
# for pickle lovers
|
118 |
+
state = self.__dict__.copy()
|
119 |
+
del state["tokenizer"]
|
120 |
+
return state
|
121 |
+
|
122 |
+
def __setstate__(self, state):
|
123 |
+
# tokenizer is not python native; don't pass it; rebuild it
|
124 |
+
self.__dict__.update(state)
|
125 |
+
enc = tiktoken.Encoding(
|
126 |
+
"Qwen",
|
127 |
+
pat_str=PAT_STR,
|
128 |
+
mergeable_ranks=self.mergeable_ranks,
|
129 |
+
special_tokens=self.special_tokens,
|
130 |
+
)
|
131 |
+
self.tokenizer = enc
|
132 |
+
|
133 |
+
def __len__(self) -> int:
|
134 |
+
return self.tokenizer.n_vocab
|
135 |
+
|
136 |
+
def get_vocab(self) -> Dict[bytes, int]:
|
137 |
+
return self.mergeable_ranks
|
138 |
+
|
139 |
+
def convert_tokens_to_ids(
|
140 |
+
self, tokens: Union[bytes, str, List[Union[bytes, str]]]
|
141 |
+
) -> List[int]:
|
142 |
+
ids = []
|
143 |
+
if isinstance(tokens, (str, bytes)):
|
144 |
+
if tokens in self.special_tokens:
|
145 |
+
return self.special_tokens[tokens]
|
146 |
+
else:
|
147 |
+
return self.mergeable_ranks.get(tokens)
|
148 |
+
for token in tokens:
|
149 |
+
if token in self.special_tokens:
|
150 |
+
ids.append(self.special_tokens[token])
|
151 |
+
else:
|
152 |
+
ids.append(self.mergeable_ranks.get(token))
|
153 |
+
return ids
|
154 |
+
|
155 |
+
def _add_tokens(
|
156 |
+
self,
|
157 |
+
new_tokens: Union[List[str], List[AddedToken]],
|
158 |
+
special_tokens: bool = False,
|
159 |
+
) -> int:
|
160 |
+
if not special_tokens and new_tokens:
|
161 |
+
raise ValueError("Adding regular tokens is not supported")
|
162 |
+
for token in new_tokens:
|
163 |
+
surface_form = token.content if isinstance(token, AddedToken) else token
|
164 |
+
if surface_form not in SPECIAL_TOKENS_SET:
|
165 |
+
raise ValueError("Adding unknown special tokens is not supported")
|
166 |
+
return 0
|
167 |
+
|
168 |
+
def save_vocabulary(self, save_directory: str, **kwargs) -> Tuple[str]:
|
169 |
+
"""
|
170 |
+
Save only the vocabulary of the tokenizer (vocabulary).
|
171 |
+
|
172 |
+
Returns:
|
173 |
+
`Tuple(str)`: Paths to the files saved.
|
174 |
+
"""
|
175 |
+
file_path = os.path.join(save_directory, "qwen.tiktoken")
|
176 |
+
with open(file_path, "w", encoding="utf8") as w:
|
177 |
+
for k, v in self.mergeable_ranks.items():
|
178 |
+
line = base64.b64encode(k).decode("utf8") + " " + str(v) + "\n"
|
179 |
+
w.write(line)
|
180 |
+
return (file_path,)
|
181 |
+
|
182 |
+
def tokenize(
|
183 |
+
self,
|
184 |
+
text: str,
|
185 |
+
allowed_special: Union[Set, str] = "all",
|
186 |
+
disallowed_special: Union[Collection, str] = (),
|
187 |
+
**kwargs,
|
188 |
+
) -> List[Union[bytes, str]]:
|
189 |
+
"""
|
190 |
+
Converts a string in a sequence of tokens.
|
191 |
+
|
192 |
+
Args:
|
193 |
+
text (`str`):
|
194 |
+
The sequence to be encoded.
|
195 |
+
allowed_special (`Literal["all"]` or `set`):
|
196 |
+
The surface forms of the tokens to be encoded as special tokens in regular texts.
|
197 |
+
Default to "all".
|
198 |
+
disallowed_special (`Literal["all"]` or `Collection`):
|
199 |
+
The surface forms of the tokens that should not be in regular texts and trigger errors.
|
200 |
+
Default to an empty tuple.
|
201 |
+
|
202 |
+
kwargs (additional keyword arguments, *optional*):
|
203 |
+
Will be passed to the underlying model specific encode method.
|
204 |
+
|
205 |
+
Returns:
|
206 |
+
`List[bytes|str]`: The list of tokens.
|
207 |
+
"""
|
208 |
+
tokens = []
|
209 |
+
text = unicodedata.normalize("NFC", text)
|
210 |
+
|
211 |
+
# this implementation takes a detour: text -> token id -> token surface forms
|
212 |
+
for t in self.tokenizer.encode(
|
213 |
+
text, allowed_special=allowed_special, disallowed_special=disallowed_special
|
214 |
+
):
|
215 |
+
tokens.append(self.decoder[t])
|
216 |
+
return tokens
|
217 |
+
|
218 |
+
def convert_tokens_to_string(self, tokens: List[Union[bytes, str]]) -> str:
|
219 |
+
"""
|
220 |
+
Converts a sequence of tokens in a single string.
|
221 |
+
"""
|
222 |
+
text = ""
|
223 |
+
temp = b""
|
224 |
+
for t in tokens:
|
225 |
+
if isinstance(t, str):
|
226 |
+
if temp:
|
227 |
+
text += temp.decode("utf-8", errors=self.errors)
|
228 |
+
temp = b""
|
229 |
+
text += t
|
230 |
+
elif isinstance(t, bytes):
|
231 |
+
temp += t
|
232 |
+
else:
|
233 |
+
raise TypeError("token should only be of type types or str")
|
234 |
+
if temp:
|
235 |
+
text += temp.decode("utf-8", errors=self.errors)
|
236 |
+
return text
|
237 |
+
|
238 |
+
@property
|
239 |
+
def vocab_size(self):
|
240 |
+
return self.tokenizer.n_vocab
|
241 |
+
|
242 |
+
def _convert_id_to_token(self, index: int) -> Union[bytes, str]:
|
243 |
+
"""Converts an id to a token, special tokens included"""
|
244 |
+
if index in self.decoder:
|
245 |
+
return self.decoder[index]
|
246 |
+
raise ValueError("unknown ids")
|
247 |
+
|
248 |
+
def _convert_token_to_id(self, token: Union[bytes, str]) -> int:
|
249 |
+
"""Converts a token to an id using the vocab, special tokens included"""
|
250 |
+
if token in self.special_tokens:
|
251 |
+
return self.special_tokens[token]
|
252 |
+
if token in self.mergeable_ranks:
|
253 |
+
return self.mergeable_ranks[token]
|
254 |
+
raise ValueError("unknown token")
|
255 |
+
|
256 |
+
def _tokenize(self, text: str, **kwargs):
|
257 |
+
"""
|
258 |
+
Converts a string in a sequence of tokens (string), using the tokenizer. Split in words for word-based
|
259 |
+
vocabulary or sub-words for sub-word-based vocabularies (BPE/SentencePieces/WordPieces).
|
260 |
+
|
261 |
+
Do NOT take care of added tokens.
|
262 |
+
"""
|
263 |
+
raise NotImplementedError
|
264 |
+
|
265 |
+
def _decode(
|
266 |
+
self,
|
267 |
+
token_ids: Union[int, List[int]],
|
268 |
+
skip_special_tokens: bool = False,
|
269 |
+
errors: str = None,
|
270 |
+
**kwargs,
|
271 |
+
) -> str:
|
272 |
+
if isinstance(token_ids, int):
|
273 |
+
token_ids = [token_ids]
|
274 |
+
if skip_special_tokens:
|
275 |
+
token_ids = [i for i in token_ids if i < self.eod_id]
|
276 |
+
return self.tokenizer.decode(token_ids, errors=errors or self.errors)
|
tokenizer_config.json
ADDED
@@ -0,0 +1,17 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"added_tokens_decoder": {},
|
3 |
+
"auto_map": {
|
4 |
+
"AutoTokenizer": [
|
5 |
+
"tokenization_qwen.QWenTokenizer",
|
6 |
+
null
|
7 |
+
]
|
8 |
+
},
|
9 |
+
"chat_template": "{% set system_message = 'You are a helpful assistant.' %}{% if messages[0]['role'] == 'system' %}{% set loop_messages = messages[1:] %}{% set system_message = messages[0]['content'] %}{% else %}{% set loop_messages = messages %}{% endif %}{% if system_message is defined %}{{ '<|im_start|>system\n' + system_message + '<|im_end|>\n' }}{% endif %}{% for message in loop_messages %}{% set content = message['content'] %}{% if message['role'] == 'user' %}{{ '<|im_start|>user\n' + content + '<|im_end|>\n<|im_start|>assistant\n' }}{% elif message['role'] == 'assistant' %}{{ content + '<|im_end|>' + '\n' }}{% endif %}{% endfor %}",
|
10 |
+
"clean_up_tokenization_spaces": true,
|
11 |
+
"eos_token": "<|im_end|>",
|
12 |
+
"model_max_length": 8192,
|
13 |
+
"pad_token": "<|im_end|>",
|
14 |
+
"padding_side": "right",
|
15 |
+
"split_special_tokens": false,
|
16 |
+
"tokenizer_class": "QWenTokenizer"
|
17 |
+
}
|
train_results.json
ADDED
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"epoch": 0.992,
|
3 |
+
"num_input_tokens_seen": 610016,
|
4 |
+
"total_flos": 5596354012643328.0,
|
5 |
+
"train_loss": 0.6199409730972782,
|
6 |
+
"train_runtime": 3421.6329,
|
7 |
+
"train_samples_per_second": 0.292,
|
8 |
+
"train_steps_per_second": 0.018
|
9 |
+
}
|
trainer_log.jsonl
ADDED
@@ -0,0 +1,13 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{"current_steps": 5, "total_steps": 62, "loss": 0.8189, "learning_rate": 4.920192797165511e-05, "epoch": 0.08, "percentage": 8.06, "elapsed_time": "0:04:59", "remaining_time": "0:56:59", "throughput": 180.63, "total_tokens": 54176}
|
2 |
+
{"current_steps": 10, "total_steps": 62, "loss": 0.6742, "learning_rate": 4.685866540361456e-05, "epoch": 0.16, "percentage": 16.13, "elapsed_time": "0:09:21", "remaining_time": "0:48:40", "throughput": 182.53, "total_tokens": 102528}
|
3 |
+
{"current_steps": 15, "total_steps": 62, "loss": 0.6816, "learning_rate": 4.3119819680728e-05, "epoch": 0.24, "percentage": 24.19, "elapsed_time": "0:14:05", "remaining_time": "0:44:09", "throughput": 180.81, "total_tokens": 152912}
|
4 |
+
{"current_steps": 20, "total_steps": 62, "loss": 0.6588, "learning_rate": 3.822410025817406e-05, "epoch": 0.32, "percentage": 32.26, "elapsed_time": "0:18:45", "remaining_time": "0:39:23", "throughput": 181.6, "total_tokens": 204432}
|
5 |
+
{"current_steps": 25, "total_steps": 62, "loss": 0.5882, "learning_rate": 3.2484078074333954e-05, "epoch": 0.4, "percentage": 40.32, "elapsed_time": "0:23:35", "remaining_time": "0:34:55", "throughput": 179.77, "total_tokens": 254512}
|
6 |
+
{"current_steps": 30, "total_steps": 62, "loss": 0.6473, "learning_rate": 2.6266229220967818e-05, "epoch": 0.48, "percentage": 48.39, "elapsed_time": "0:27:52", "remaining_time": "0:29:44", "throughput": 178.51, "total_tokens": 298608}
|
7 |
+
{"current_steps": 35, "total_steps": 62, "loss": 0.6343, "learning_rate": 1.9967536997783494e-05, "epoch": 0.56, "percentage": 56.45, "elapsed_time": "0:32:46", "remaining_time": "0:25:16", "throughput": 177.47, "total_tokens": 348928}
|
8 |
+
{"current_steps": 40, "total_steps": 62, "loss": 0.5067, "learning_rate": 1.399014621105914e-05, "epoch": 0.64, "percentage": 64.52, "elapsed_time": "0:37:00", "remaining_time": "0:20:21", "throughput": 178.92, "total_tokens": 397216}
|
9 |
+
{"current_steps": 45, "total_steps": 62, "loss": 0.5386, "learning_rate": 8.715687931944449e-06, "epoch": 0.72, "percentage": 72.58, "elapsed_time": "0:41:30", "remaining_time": "0:15:40", "throughput": 178.61, "total_tokens": 444832}
|
10 |
+
{"current_steps": 50, "total_steps": 62, "loss": 0.563, "learning_rate": 4.480913969818098e-06, "epoch": 0.8, "percentage": 80.65, "elapsed_time": "0:46:06", "remaining_time": "0:11:03", "throughput": 177.36, "total_tokens": 490640}
|
11 |
+
{"current_steps": 55, "total_steps": 62, "loss": 0.5484, "learning_rate": 1.5561966963229924e-06, "epoch": 0.88, "percentage": 88.71, "elapsed_time": "0:50:03", "remaining_time": "0:06:22", "throughput": 179.23, "total_tokens": 538400}
|
12 |
+
{"current_steps": 60, "total_steps": 62, "loss": 0.5979, "learning_rate": 1.2826691520262114e-07, "epoch": 0.96, "percentage": 96.77, "elapsed_time": "0:54:46", "remaining_time": "0:01:49", "throughput": 179.48, "total_tokens": 589856}
|
13 |
+
{"current_steps": 62, "total_steps": 62, "epoch": 0.992, "percentage": 100.0, "elapsed_time": "0:57:01", "remaining_time": "0:00:00", "throughput": 178.28, "total_tokens": 610016}
|
trainer_state.json
ADDED
@@ -0,0 +1,139 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"best_metric": null,
|
3 |
+
"best_model_checkpoint": null,
|
4 |
+
"epoch": 0.992,
|
5 |
+
"eval_steps": 500,
|
6 |
+
"global_step": 62,
|
7 |
+
"is_hyper_param_search": false,
|
8 |
+
"is_local_process_zero": true,
|
9 |
+
"is_world_process_zero": true,
|
10 |
+
"log_history": [
|
11 |
+
{
|
12 |
+
"epoch": 0.08,
|
13 |
+
"grad_norm": 0.7084760069847107,
|
14 |
+
"learning_rate": 4.920192797165511e-05,
|
15 |
+
"loss": 0.8189,
|
16 |
+
"num_input_tokens_seen": 54176,
|
17 |
+
"step": 5
|
18 |
+
},
|
19 |
+
{
|
20 |
+
"epoch": 0.16,
|
21 |
+
"grad_norm": 0.7297347784042358,
|
22 |
+
"learning_rate": 4.685866540361456e-05,
|
23 |
+
"loss": 0.6742,
|
24 |
+
"num_input_tokens_seen": 102528,
|
25 |
+
"step": 10
|
26 |
+
},
|
27 |
+
{
|
28 |
+
"epoch": 0.24,
|
29 |
+
"grad_norm": 684049.0625,
|
30 |
+
"learning_rate": 4.3119819680728e-05,
|
31 |
+
"loss": 0.6816,
|
32 |
+
"num_input_tokens_seen": 152912,
|
33 |
+
"step": 15
|
34 |
+
},
|
35 |
+
{
|
36 |
+
"epoch": 0.32,
|
37 |
+
"grad_norm": 0.5369439125061035,
|
38 |
+
"learning_rate": 3.822410025817406e-05,
|
39 |
+
"loss": 0.6588,
|
40 |
+
"num_input_tokens_seen": 204432,
|
41 |
+
"step": 20
|
42 |
+
},
|
43 |
+
{
|
44 |
+
"epoch": 0.4,
|
45 |
+
"grad_norm": 0.5415367484092712,
|
46 |
+
"learning_rate": 3.2484078074333954e-05,
|
47 |
+
"loss": 0.5882,
|
48 |
+
"num_input_tokens_seen": 254512,
|
49 |
+
"step": 25
|
50 |
+
},
|
51 |
+
{
|
52 |
+
"epoch": 0.48,
|
53 |
+
"grad_norm": 0.7139325737953186,
|
54 |
+
"learning_rate": 2.6266229220967818e-05,
|
55 |
+
"loss": 0.6473,
|
56 |
+
"num_input_tokens_seen": 298608,
|
57 |
+
"step": 30
|
58 |
+
},
|
59 |
+
{
|
60 |
+
"epoch": 0.56,
|
61 |
+
"grad_norm": 2127319.0,
|
62 |
+
"learning_rate": 1.9967536997783494e-05,
|
63 |
+
"loss": 0.6343,
|
64 |
+
"num_input_tokens_seen": 348928,
|
65 |
+
"step": 35
|
66 |
+
},
|
67 |
+
{
|
68 |
+
"epoch": 0.64,
|
69 |
+
"grad_norm": 5908939.5,
|
70 |
+
"learning_rate": 1.399014621105914e-05,
|
71 |
+
"loss": 0.5067,
|
72 |
+
"num_input_tokens_seen": 397216,
|
73 |
+
"step": 40
|
74 |
+
},
|
75 |
+
{
|
76 |
+
"epoch": 0.72,
|
77 |
+
"grad_norm": 1202378.875,
|
78 |
+
"learning_rate": 8.715687931944449e-06,
|
79 |
+
"loss": 0.5386,
|
80 |
+
"num_input_tokens_seen": 444832,
|
81 |
+
"step": 45
|
82 |
+
},
|
83 |
+
{
|
84 |
+
"epoch": 0.8,
|
85 |
+
"grad_norm": 0.4307733178138733,
|
86 |
+
"learning_rate": 4.480913969818098e-06,
|
87 |
+
"loss": 0.563,
|
88 |
+
"num_input_tokens_seen": 490640,
|
89 |
+
"step": 50
|
90 |
+
},
|
91 |
+
{
|
92 |
+
"epoch": 0.88,
|
93 |
+
"grad_norm": 11058075.0,
|
94 |
+
"learning_rate": 1.5561966963229924e-06,
|
95 |
+
"loss": 0.5484,
|
96 |
+
"num_input_tokens_seen": 538400,
|
97 |
+
"step": 55
|
98 |
+
},
|
99 |
+
{
|
100 |
+
"epoch": 0.96,
|
101 |
+
"grad_norm": 2679230.5,
|
102 |
+
"learning_rate": 1.2826691520262114e-07,
|
103 |
+
"loss": 0.5979,
|
104 |
+
"num_input_tokens_seen": 589856,
|
105 |
+
"step": 60
|
106 |
+
},
|
107 |
+
{
|
108 |
+
"epoch": 0.992,
|
109 |
+
"num_input_tokens_seen": 610016,
|
110 |
+
"step": 62,
|
111 |
+
"total_flos": 5596354012643328.0,
|
112 |
+
"train_loss": 0.6199409730972782,
|
113 |
+
"train_runtime": 3421.6329,
|
114 |
+
"train_samples_per_second": 0.292,
|
115 |
+
"train_steps_per_second": 0.018
|
116 |
+
}
|
117 |
+
],
|
118 |
+
"logging_steps": 5,
|
119 |
+
"max_steps": 62,
|
120 |
+
"num_input_tokens_seen": 610016,
|
121 |
+
"num_train_epochs": 1,
|
122 |
+
"save_steps": 100,
|
123 |
+
"stateful_callbacks": {
|
124 |
+
"TrainerControl": {
|
125 |
+
"args": {
|
126 |
+
"should_epoch_stop": false,
|
127 |
+
"should_evaluate": false,
|
128 |
+
"should_log": false,
|
129 |
+
"should_save": true,
|
130 |
+
"should_training_stop": true
|
131 |
+
},
|
132 |
+
"attributes": {}
|
133 |
+
}
|
134 |
+
},
|
135 |
+
"total_flos": 5596354012643328.0,
|
136 |
+
"train_batch_size": 2,
|
137 |
+
"trial_name": null,
|
138 |
+
"trial_params": null
|
139 |
+
}
|
training_args.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:311875b8ca25e2de38752c7cd348177c9ca8d80ec40f9c44f87bd2bc51a3e94b
|
3 |
+
size 5368
|
training_args.yaml
ADDED
@@ -0,0 +1,32 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
bf16: true
|
2 |
+
cutoff_len: 1024
|
3 |
+
dataset: glaive_toolcall_en
|
4 |
+
dataset_dir: data
|
5 |
+
ddp_timeout: 180000000
|
6 |
+
do_train: true
|
7 |
+
finetuning_type: lora
|
8 |
+
flash_attn: auto
|
9 |
+
gradient_accumulation_steps: 8
|
10 |
+
include_num_input_tokens_seen: true
|
11 |
+
learning_rate: 5.0e-05
|
12 |
+
logging_steps: 5
|
13 |
+
lora_alpha: 16
|
14 |
+
lora_dropout: 0
|
15 |
+
lora_rank: 8
|
16 |
+
lora_target: all
|
17 |
+
lr_scheduler_type: cosine
|
18 |
+
max_grad_norm: 1.0
|
19 |
+
max_samples: 100000
|
20 |
+
model_name_or_path: qwen/Qwen-1_8B-Chat
|
21 |
+
num_train_epochs: 3.0
|
22 |
+
optim: adamw_torch
|
23 |
+
output_dir: saves/Qwen-1.8B-Chat/lora/train_2024-08-31-17-40-34
|
24 |
+
packing: false
|
25 |
+
per_device_train_batch_size: 2
|
26 |
+
plot_loss: true
|
27 |
+
preprocessing_num_workers: 16
|
28 |
+
report_to: none
|
29 |
+
save_steps: 100
|
30 |
+
stage: sft
|
31 |
+
template: qwen
|
32 |
+
warmup_steps: 0
|
training_loss.png
ADDED