Hugging Face's logo

BeaverAI
/
Qwen2.5-QwQ-RP-Draft-v0.2-1.5B

Text Generation
Transformers
Safetensors
qwen2
axolotl
Generated from Trainer
conversational
text-generation-inference
Model card Files Community
1
xzuyn commited on
Commit
c1cd87d
·
verified ·
1 Parent(s): 426d3e5

Create README.md

Browse files
Files changed (1) hide show
  1. README.md +325 -0
README.md ADDED
@@ -0,0 +1,325 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ library_name: transformers
3
+ license: apache-2.0
4
+ base_model: Qwen/Qwen2.5-1.5B-Instruct
5
+ tags:
6
+ - axolotl
7
+ - generated_from_trainer
8
+ datasets:
9
+ - PJMixers-Dev/allura-org_gryphe-sonnet-3.5-charcards-names-added-qwq-all-aphrodite-Shuffled
10
+ - PJMixers-Dev/anthracite-org_c2_logs_32k_llama3_qwen2_v1.3-qwq-all-aphrodite-Shuffled
11
+ - PJMixers-Dev/grimulkan_aicg-logs-augmented-system-qwq-all-aphrodite-Shuffled
12
+ - PJMixers-Dev/grimulkan_jannie-log-augmented-system-qwq-all-aphrodite-Shuffled
13
+ - PJMixers-Dev/grimulkan_PIPPA-augmented-dedup-system-qwq-all-aphrodite-Shuffled
14
+ - PJMixers-Dev/lemonilia_LimaRP-Only-NonSus-Simple-CustomShareGPT-qwq-all-aphrodite-Shuffled
15
+ - PJMixers-Dev/MinervaAI_Aesir-Preview-Anon-qwq-all-aphrodite-Shuffled
16
+ - PJMixers-Dev/NyxKrage_chub-logs-sharegpt-longest-CustomShareGPT-qwq-all-aphrodite-Shuffled
17
+ - PJMixers-Dev/PocketDoc_Dans-Prosemaxx-Cowriter-XL-8192-shrunk-l3-qwq-all-aphrodite-Shuffled
18
+ - PJMixers-Dev/PocketDoc_Dans-Personamaxx-Rainy-qwq-all-aphrodite-Shuffled
19
+ model-index:
20
+ - name: Qwen2.5-QwQ-RP-Draft-v0.2-1.5B-LoRA-WS
21
+ results: []
22
+ ---
23
+
24
+
25
+
26
+ <!-- This model card has been generated automatically according to the information the Trainer had access to. You
27
+ should probably proofread and complete it, then remove this comment. -->
28
+
29
+ [<img src="https://raw.githubusercontent.com/axolotl-ai-cloud/axolotl/main/image/axolotl-badge-web.png" alt="Built with Axolotl" width="200" height="32"/>](https://github.com/axolotl-ai-cloud/axolotl)
30
+ <details><summary>See axolotl config</summary>
31
+
32
+ axolotl version: `0.8.0.dev0`
33
+ ```yaml
34
+ mlflow_tracking_uri: http://127.0.0.1:7860
35
+ mlflow_experiment_name: Qwen2.5-QwQ-RP-Draft-v0.2-1.5B-LoRA
36
+
37
+ # Hugging Face saving config
38
+ hub_model_id: BeaverAI/Qwen2.5-QwQ-RP-Draft-v0.2-1.5B-LoRA-WS
39
+ hub_strategy: every_save
40
+
41
+ # Model checkpointing config
42
+ output_dir: ./Outputs/Qwen2.5-QwQ-RP-Draft-v0.2-1.5B-LoRA
43
+ resume_from_checkpoint:
44
+ save_steps: 10
45
+ save_safetensors: true
46
+ save_total_limit: 3
47
+ save_only_model: false
48
+
49
+ # Model architecture config
50
+ base_model: Qwen/Qwen2.5-1.5B-Instruct
51
+ model_type: AutoModelForCausalLM
52
+ tokenizer_type: AutoTokenizer
53
+
54
+ # Mixed precision training config
55
+ bf16: true
56
+ fp16: false
57
+ tf32: false
58
+
59
+ # Model loading config
60
+ load_in_8bit: false
61
+ load_in_4bit: false
62
+ strict: false
63
+
64
+ # Sequence config
65
+ sequence_len: 8192
66
+ min_sample_len: 256
67
+ sample_packing: true
68
+ eval_sample_packing: false
69
+ pad_to_sequence_len: true
70
+ train_on_inputs: false
71
+ group_by_length: false
72
+
73
+ # LoRA adapter config
74
+ adapter: lora
75
+ lora_model_dir:
76
+ lora_r: 128
77
+ lora_alpha: 128
78
+ lora_dropout: 0.125
79
+ peft_layers_to_transform:
80
+ peft_use_dora:
81
+ peft_use_rslora:
82
+ peft_layer_replication:
83
+ lora_target_modules:
84
+ - gate_proj
85
+ - down_proj
86
+ - up_proj
87
+ - q_proj
88
+ - v_proj
89
+ - k_proj
90
+ - o_proj
91
+ lora_modules_to_save:
92
+
93
+ # Fix uninitialized tokens (such as <|start_header_id|> on the base L3 models)
94
+ fix_untrained_tokens:
95
+
96
+ # Dataset config
97
+ # https://github.com/xzuyn/axolotl/blob/prompt_formats/src/axolotl/prompt_strategies/customchatml-regex-last-only.py
98
+ datasets:
99
+ - path: PJMixers-Dev/allura-org_gryphe-sonnet-3.5-charcards-names-added-qwq-all-aphrodite-Shuffled
100
+ split: train
101
+ type: customchatml-regex-last-only
102
+ - path: PJMixers-Dev/anthracite-org_c2_logs_32k_llama3_qwen2_v1.3-qwq-all-aphrodite-Shuffled
103
+ split: train
104
+ type: customchatml-regex-last-only
105
+ - path: PJMixers-Dev/grimulkan_aicg-logs-augmented-system-qwq-all-aphrodite-Shuffled
106
+ split: train
107
+ type: customchatml-regex-last-only
108
+ - path: PJMixers-Dev/grimulkan_jannie-log-augmented-system-qwq-all-aphrodite-Shuffled
109
+ split: train
110
+ type: customchatml-regex-last-only
111
+ - path: PJMixers-Dev/grimulkan_PIPPA-augmented-dedup-system-qwq-all-aphrodite-Shuffled
112
+ split: train
113
+ type: customchatml-regex-last-only
114
+ - path: PJMixers-Dev/lemonilia_LimaRP-Only-NonSus-Simple-CustomShareGPT-qwq-all-aphrodite-Shuffled
115
+ split: train
116
+ type: customchatml-regex-last-only
117
+ - path: PJMixers-Dev/MinervaAI_Aesir-Preview-Anon-qwq-all-aphrodite-Shuffled
118
+ split: train
119
+ type: customchatml-regex-last-only
120
+ - path: PJMixers-Dev/NyxKrage_chub-logs-sharegpt-longest-CustomShareGPT-qwq-all-aphrodite-Shuffled
121
+ split: train
122
+ type: customchatml-regex-last-only
123
+ - path: PJMixers-Dev/PocketDoc_Dans-Prosemaxx-Cowriter-XL-8192-shrunk-l3-qwq-all-aphrodite-Shuffled
124
+ split: train
125
+ type: customchatml-regex-last-only
126
+ - path: PJMixers-Dev/PocketDoc_Dans-Personamaxx-Rainy-qwq-all-aphrodite-Shuffled
127
+ split: train
128
+ type: customchatml-regex-last-only
129
+ test_datasets:
130
+ - path: PJMixers-Dev/allura-org_gryphe-sonnet-3.5-charcards-names-added-qwq-all-aphrodite-Shuffled
131
+ split: test
132
+ type: customchatml-regex-last-only
133
+ - path: PJMixers-Dev/anthracite-org_c2_logs_32k_llama3_qwen2_v1.3-qwq-all-aphrodite-Shuffled
134
+ split: test
135
+ type: customchatml-regex-last-only
136
+ - path: PJMixers-Dev/grimulkan_aicg-logs-augmented-system-qwq-all-aphrodite-Shuffled
137
+ split: test
138
+ type: customchatml-regex-last-only
139
+ - path: PJMixers-Dev/grimulkan_jannie-log-augmented-system-qwq-all-aphrodite-Shuffled
140
+ split: test
141
+ type: customchatml-regex-last-only
142
+ - path: PJMixers-Dev/grimulkan_PIPPA-augmented-dedup-system-qwq-all-aphrodite-Shuffled
143
+ split: test
144
+ type: customchatml-regex-last-only
145
+ - path: PJMixers-Dev/lemonilia_LimaRP-Only-NonSus-Simple-CustomShareGPT-qwq-all-aphrodite-Shuffled
146
+ split: test
147
+ type: customchatml-regex-last-only
148
+ - path: PJMixers-Dev/MinervaAI_Aesir-Preview-Anon-qwq-all-aphrodite-Shuffled
149
+ split: test
150
+ type: customchatml-regex-last-only
151
+ - path: PJMixers-Dev/NyxKrage_chub-logs-sharegpt-longest-CustomShareGPT-qwq-all-aphrodite-Shuffled
152
+ split: test
153
+ type: customchatml-regex-last-only
154
+ - path: PJMixers-Dev/PocketDoc_Dans-Prosemaxx-Cowriter-XL-8192-shrunk-l3-qwq-all-aphrodite-Shuffled
155
+ split: test
156
+ type: customchatml-regex-last-only
157
+ - path: PJMixers-Dev/PocketDoc_Dans-Personamaxx-Rainy-qwq-all-aphrodite-Shuffled
158
+ split: test
159
+ type: customchatml-regex-last-only
160
+ val_set_size: 0
161
+ eval_strategy: steps
162
+ eval_steps: 10
163
+ dataset_prepared_path: ./00-Tokenized-Datasets/Qwen2.5-QwQ-Draft-v0.2-1.5B-customchatml-regex-last-only
164
+ shuffle_merged_datasets: true
165
+ dataset_processes:
166
+
167
+ # Training hyperparameters
168
+ num_epochs: 2
169
+ gradient_accumulation_steps: 1
170
+ micro_batch_size: 16 # x4 GPUs = 64
171
+ eval_batch_size: 16 # x4 GPUs = 64
172
+ warmup_steps: 0
173
+ optimizer: came_pytorch
174
+ optim_args:
175
+ optim_target_modules:
176
+ lr_scheduler: rex
177
+ learning_rate: 2e-5
178
+ cosine_min_lr_ratio:
179
+ loraplus_lr_ratio:
180
+ loraplus_lr_embedding:
181
+ weight_decay: 0.1
182
+ max_grad_norm: 1
183
+ logging_steps: 1
184
+
185
+ # Model optimization
186
+ gradient_checkpointing: unsloth
187
+ flash_attention: true
188
+ plugins:
189
+ - axolotl.integrations.liger.LigerPlugin
190
+ - axolotl.integrations.cut_cross_entropy.CutCrossEntropyPlugin
191
+ cut_cross_entropy: true
192
+ liger_rope: true
193
+ liger_rms_norm: true
194
+ liger_layer_norm: true
195
+ liger_glu_activation: true
196
+ liger_cross_entropy: false
197
+ liger_fused_linear_cross_entropy: false
198
+ lora_mlp_kernel: false
199
+ lora_qkv_kernel: false
200
+ lora_o_kernel: false
201
+
202
+ # DeepSpeed
203
+ deepspeed: deepspeed_configs/zero3_bf16.json
204
+
205
+ # Garbage Collection
206
+ gc_steps: 1
207
+
208
+ # Debug config
209
+ debug: true
210
+ seed: 42
211
+
212
+ # Token config
213
+ special_tokens:
214
+ eos_token: "<|endoftext|>"
215
+ pad_token: "<|endoftext|>"
216
+ tokens:
217
+
218
+ ```
219
+
220
+ </details><br>
221
+
222
+ # Qwen2.5-QwQ-RP-Draft-v0.2-1.5B-LoRA-WS
223
+
224
+ This model is a fine-tuned version of [Qwen/Qwen2.5-1.5B-Instruct](https://huggingface.co/Qwen/Qwen2.5-1.5B-Instruct) on the PJMixers-Dev/allura-org_gryphe-sonnet-3.5-charcards-names-added-qwq-all-aphrodite-Shuffled, the PJMixers-Dev/anthracite-org_c2_logs_32k_llama3_qwen2_v1.3-qwq-all-aphrodite-Shuffled, the PJMixers-Dev/grimulkan_aicg-logs-augmented-system-qwq-all-aphrodite-Shuffled, the PJMixers-Dev/grimulkan_jannie-log-augmented-system-qwq-all-aphrodite-Shuffled, the PJMixers-Dev/grimulkan_PIPPA-augmented-dedup-system-qwq-all-aphrodite-Shuffled, the PJMixers-Dev/lemonilia_LimaRP-Only-NonSus-Simple-CustomShareGPT-qwq-all-aphrodite-Shuffled, the PJMixers-Dev/MinervaAI_Aesir-Preview-Anon-qwq-all-aphrodite-Shuffled, the PJMixers-Dev/NyxKrage_chub-logs-sharegpt-longest-CustomShareGPT-qwq-all-aphrodite-Shuffled, the PJMixers-Dev/PocketDoc_Dans-Prosemaxx-Cowriter-XL-8192-shrunk-l3-qwq-all-aphrodite-Shuffled and the PJMixers-Dev/PocketDoc_Dans-Personamaxx-Rainy-qwq-all-aphrodite-Shuffled datasets.
225
+ It achieves the following results on the evaluation set:
226
+ - Loss: 1.7920
227
+
228
+ ## Model description
229
+
230
+ More information needed
231
+
232
+ ## Intended uses & limitations
233
+
234
+ More information needed
235
+
236
+ ## Training and evaluation data
237
+
238
+ More information needed
239
+
240
+ ## Training procedure
241
+
242
+ ### Training hyperparameters
243
+
244
+ The following hyperparameters were used during training:
245
+ - learning_rate: 2e-05
246
+ - train_batch_size: 16
247
+ - eval_batch_size: 16
248
+ - seed: 42
249
+ - distributed_type: multi-GPU
250
+ - num_devices: 4
251
+ - total_train_batch_size: 64
252
+ - total_eval_batch_size: 64
253
+ - optimizer: Use OptimizerNames.ADAMW_HF with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
254
+ - lr_scheduler_type: cosine
255
+ - num_epochs: 2.0
256
+
257
+ ### Training results
258
+
259
+ | Training Loss | Epoch | Step | Validation Loss |
260
+ |:-------------:|:------:|:----:|:---------------:|
261
+ | 2.7253 | 0.0036 | 1 | 2.6463 |
262
+ | 2.4339 | 0.0362 | 10 | 2.3864 |
263
+ | 2.2821 | 0.0725 | 20 | 2.2571 |
264
+ | 2.207 | 0.1087 | 30 | 2.1771 |
265
+ | 2.1213 | 0.1449 | 40 | 2.1225 |
266
+ | 2.0926 | 0.1812 | 50 | 2.0870 |
267
+ | 2.0274 | 0.2174 | 60 | 2.0585 |
268
+ | 2.0292 | 0.2536 | 70 | 2.0338 |
269
+ | 2.0057 | 0.2899 | 80 | 2.0148 |
270
+ | 1.9841 | 0.3261 | 90 | 1.9975 |
271
+ | 1.9639 | 0.3623 | 100 | 1.9812 |
272
+ | 1.9303 | 0.3986 | 110 | 1.9688 |
273
+ | 1.9527 | 0.4348 | 120 | 1.9569 |
274
+ | 1.9389 | 0.4710 | 130 | 1.9459 |
275
+ | 1.9008 | 0.5072 | 140 | 1.9352 |
276
+ | 1.8996 | 0.5435 | 150 | 1.9259 |
277
+ | 1.8849 | 0.5797 | 160 | 1.9171 |
278
+ | 1.8793 | 0.6159 | 170 | 1.9105 |
279
+ | 1.8748 | 0.6522 | 180 | 1.9019 |
280
+ | 1.8697 | 0.6884 | 190 | 1.8959 |
281
+ | 1.8777 | 0.7246 | 200 | 1.8892 |
282
+ | 1.8731 | 0.7609 | 210 | 1.8833 |
283
+ | 1.8358 | 0.7971 | 220 | 1.8772 |
284
+ | 1.8483 | 0.8333 | 230 | 1.8729 |
285
+ | 1.8567 | 0.8696 | 240 | 1.8679 |
286
+ | 1.8461 | 0.9058 | 250 | 1.8632 |
287
+ | 1.8436 | 0.9420 | 260 | 1.8591 |
288
+ | 1.7992 | 0.9783 | 270 | 1.8545 |
289
+ | 1.7925 | 1.0145 | 280 | 1.8510 |
290
+ | 1.814 | 1.0507 | 290 | 1.8461 |
291
+ | 1.8249 | 1.0870 | 300 | 1.8431 |
292
+ | 1.7889 | 1.1232 | 310 | 1.8391 |
293
+ | 1.8116 | 1.1594 | 320 | 1.8367 |
294
+ | 1.7895 | 1.1957 | 330 | 1.8339 |
295
+ | 1.7951 | 1.2319 | 340 | 1.8300 |
296
+ | 1.7831 | 1.2681 | 350 | 1.8272 |
297
+ | 1.7919 | 1.3043 | 360 | 1.8250 |
298
+ | 1.7755 | 1.3406 | 370 | 1.8224 |
299
+ | 1.7895 | 1.3768 | 380 | 1.8209 |
300
+ | 1.7806 | 1.4130 | 390 | 1.8184 |
301
+ | 1.7808 | 1.4493 | 400 | 1.8156 |
302
+ | 1.7771 | 1.4855 | 410 | 1.8136 |
303
+ | 1.7636 | 1.5217 | 420 | 1.8118 |
304
+ | 1.7719 | 1.5580 | 430 | 1.8098 |
305
+ | 1.7716 | 1.5942 | 440 | 1.8074 |
306
+ | 1.7917 | 1.6304 | 450 | 1.8056 |
307
+ | 1.784 | 1.6667 | 460 | 1.8035 |
308
+ | 1.7846 | 1.7029 | 470 | 1.8019 |
309
+ | 1.7485 | 1.7391 | 480 | 1.8002 |
310
+ | 1.7381 | 1.7754 | 490 | 1.7987 |
311
+ | 1.764 | 1.8116 | 500 | 1.7972 |
312
+ | 1.744 | 1.8478 | 510 | 1.7964 |
313
+ | 1.7673 | 1.8841 | 520 | 1.7948 |
314
+ | 1.7926 | 1.9203 | 530 | 1.7935 |
315
+ | 1.7843 | 1.9565 | 540 | 1.7926 |
316
+ | 1.7689 | 1.9928 | 550 | 1.7920 |
317
+
318
+
319
+ ### Framework versions
320
+
321
+ - PEFT 0.14.0
322
+ - Transformers 4.49.0
323
+ - Pytorch 2.6.0+cu124
324
+ - Datasets 3.2.0
325
+ - Tokenizers 0.21.1