Views
No views yet
1praisonai train \
2 --model unsloth/Phi-4-unsloth-bnb-4bit \
3 --dataset mervinpraison/ur-fall-raw \
4 --hf mervinpraison/Phi-4-ur-fall-bnb-4bit \
5 --ollama mervinpraison/Phi-4-ur-fall-bnb-4bit1dataset:
2- name: mervinpraison/ur-fall-raw
3dataset_num_proc: 2
4dataset_text_field: text
5gradient_accumulation_steps: 2
6hf_model_name: mervinpraison/Phi-4-ur-fall-bnb-4bit
7huggingface_save: 'true'
8learning_rate: 0.0002
9load_in_4bit: true
10loftq_config: null
11logging_steps: 1
12lora_alpha: 16
13lora_bias: none
14lora_dropout: 0
15lora_r: 16
16lora_target_modules:
17- q_proj
18- k_proj
19- v_proj
20- o_proj
21- gate_proj
22- up_proj
23- down_proj
24lr_scheduler_type: linear
25max_seq_length: 2048
26max_steps: 10
27model_name: unsloth/Phi-4-unsloth-bnb-4bit
28model_parameters: 8b
29num_train_epochs: 1
30ollama_model: mervinpraison/Phi-4-ur-fall-bnb-4bit
31ollama_save: 'true'
32optim: adamw_8bit
33output_dir: outputs
34packing: false
35per_device_train_batch_size: 2
36quantization_method:
37- q4_k_m
38random_state: 3407
39seed: 3407
40train: 'true'
41use_gradient_checkpointing: unsloth
42use_rslora: false
43warmup_steps: 5
44weight_decay: 0.01
451(test) ➜ test praisonai train \
2 --model unsloth/Phi-4-unsloth-bnb-4bit \
3 --dataset mervinpraison/ur-fall-raw \
4 --hf mervinpraison/Phi-4-ur-fall-bnb-4bit \
5 --ollama mervinpraison/Phi-4-ur-fall-bnb-4bit
6Conda environment 'praison_env' found.
7🦥 Unsloth: Will patch your computer to enable 2x faster free finetuning.
8🦥 Unsloth Zoo will now patch everything to make training faster!
9DEBUG: Loaded config: {'dataset': [{'name': 'mervinpraison/ur-fall-raw'}], 'dataset_num_proc': 2,
10'dataset_text_field': 'text', 'gradient_accumulation_steps': 2, 'hf_model_name':
11'mervinpraison/Phi-4-ur-fall-bnb-4bit', 'huggingface_save': 'true', 'learning_rate': 0.0002, 'load_in_4bit':
12True, 'loftq_config': None, 'logging_steps': 1, 'lora_alpha': 16, 'lora_bias': 'none', 'lora_dropout': 0,
13'lora_r': 16, 'lora_target_modules': ['q_proj', 'k_proj', 'v_proj', 'o_proj', 'gate_proj', 'up_proj',
14'down_proj'], 'lr_scheduler_type': 'linear', 'max_seq_length': 2048, 'max_steps': 10, 'model_name':
15'unsloth/Phi-4-unsloth-bnb-4bit', 'model_parameters': '8b', 'num_train_epochs': 1, 'ollama_model':
16'mervinpraison/Phi-4-ur-fall-bnb-4bit', 'ollama_save': 'true', 'optim': 'adamw_8bit', 'output_dir':
17'outputs', 'packing': False, 'per_device_train_batch_size': 2, 'quantization_method': ['q4_k_m'],
18'random_state': 3407, 'seed': 3407, 'train': 'true', 'use_gradient_checkpointing': 'unsloth', 'use_rslora':
19False, 'warmup_steps': 5, 'weight_decay': 0.01}
20DEBUG: PyTorch version: 2.3.0
21DEBUG: CUDA version: 12.1
22DEBUG: CUDA Device Capability: (8, 6)
23DEBUG: Python Version: 3.10.16 (main, Dec 11 2024, 16:24:50) [GCC 11.2.0]
24DEBUG: Python Path: /home/Ubuntu/miniconda3/envs/praison_env/bin/python
25DEBUG: GPU = NVIDIA RTX A6000. Max memory = 47.431 GB.
26DEBUG: Your runtime has 50.6 gigabytes of available RAM
27DEBUG: You are using a high-RAM runtime!
28DEBUG: Preparing model and tokenizer...
29==((====))== Unsloth 2025.1.8: Fast Llama patching. Transformers: 4.48.2.
30 \\ /| GPU: NVIDIA RTX A6000. Max memory: 47.431 GB. Platform: Linux.
31O^O/ \_/ \ Torch: 2.3.0. CUDA: 8.6. CUDA Toolkit: 12.1. Triton: 2.3.0
32\ / Bfloat16 = TRUE. FA [Xformers = 0.0.26.post1. FA2 = False]
33 "-____-" Free Apache license: http://github.com/unslothai/unsloth
34Unsloth: Fast downloading is enabled - ignore downloading bars which are red colored!
35
36Downloading shards: 0%| | 0/3 [00:00<?, ?it/s]
37Downloading shards: 33%|███▎ | 1/3 [01:53<03:46, 113.06s/it]
38Downloading shards: 67%|██████▋ | 2/3 [03:37<01:47, 107.98s/it]
39Downloading shards: 100%|██████████| 3/3 [04:01<00:00, 69.80s/it]
40Downloading shards: 100%|██████████| 3/3 [04:01<00:00, 80.61s/it]
41
42Loading checkpoint shards: 0%| | 0/3 [00:00<?, ?it/s]
43Loading checkpoint shards: 33%|███▎ | 1/3 [00:01<00:03, 1.80s/it]
44Loading checkpoint shards: 67%|██████▋ | 2/3 [00:03<00:01, 1.75s/it]
45Loading checkpoint shards: 100%|██████████| 3/3 [00:04<00:00, 1.23s/it]
46Loading checkpoint shards: 100%|██████████| 3/3 [00:04<00:00, 1.38s/it]
47DEBUG: Model and original tokenizer loaded.
48DEBUG: Chat tokenizer created; HF tokenizer saved.
49Unsloth 2025.1.8 patched 40 layers with 40 QKV layers, 40 O layers and 40 MLP layers.
50DEBUG: LoRA adapters added.
51DEBUG: Starting training...
52DEBUG: Processing dataset info: {'name': 'mervinpraison/ur-fall-raw'}
53DEBUG: Loading dataset 'mervinpraison/ur-fall-raw' split 'train'...
54
55Generating train split: 0%| | 0/56 [00:00<?, ? examples/s]
56Generating train split: 100%|██████████| 56/56 [00:00<00:00, 1580.22 examples/s]
57
58Generating test split: 0%| | 0/14 [00:00<?, ? examples/s]
59Generating test split: 100%|██████████| 14/14 [00:00<00:00, 1853.31 examples/s]
60DEBUG: Dataset columns: ['instruction', 'input', 'output']
61DEBUG: Dataset does not have 'conversations'; assuming Alpaca format.
62DEBUG: Applying formatting function to dataset...
63
64Map: 0%| | 0/56 [00:00<?, ? examples/s]DEBUG: formatting_prompts_func() received batch with keys:
65['instruction', 'input', 'output']
66DEBUG: Raw texts sample (first 200 chars): <|endoftext|><|start_header_id|>system<|end_header_id|>
67
68Cutting Knowledge Date: December 2023
69Today Date: 26 July 2024
70
71<|eot_id|><|start_header_id|>user<|end_header_id|>
72
73Analyse accelerometer senso
74
75Map: 100%|██████████| 56/56 [00:00<00:00, 1481.79 examples/s]
76DEBUG: Sample processed example keys: ['text']
77DEBUG: Sample processed 'text' type: <class 'str'>
78DEBUG: Sample processed 'text' content (first 200 chars):
79<|endoftext|><|start_header_id|>system<|end_header_id|>
80
81Cutting Knowledge Date: December 2023
82Today Date: 26 July 2024
83
84<|eot_id|><|start_header_id|>user<|end_header_id|>
85
86Analyse accelerometer senso
87DEBUG: Combined dataset has 56 examples.
88DEBUG: Tokenizing the entire dataset...
89
90Map: 0%| | 0/56 [00:00<?, ? examples/s]DEBUG: Tokenizing a batch of size: 56
91DEBUG: Tokenized sample (first 10 tokens of input_ids ): [100257, 27, 91, 2527, 8932, 851, 91, 29, 9125, 27]
92
93Map: 100%|██████████| 56/56 [00:00<00:00, 123.90 examples/s]
94Map: 100%|██████████| 56/56 [00:00<00:00, 120.58 examples/s]
95DEBUG: Tokenized dataset sample keys: dict_keys(['input_ids', 'attention_mask'])
96DEBUG: Dataset tokenization complete.
97
98Map: 0%| | 0/56 [00:00<?, ? examples/s]
99Map: 100%|██████████| 56/56 [00:00<00:00, 546.20 examples/s]
100Map: 100%|██████████| 56/56 [00:00<00:00, 529.34 examples/s]
101DEBUG: Beginning trainer.train() ...
102Unsloth: Most labels in your dataset are -100. Training losses will be all 0.
103For example, are you sure you used `train_on_responses_only` correctly?
104Or did you mask our tokens incorrectly? Maybe this is intended?
105==((====))== Unsloth - 2x faster free finetuning | Num GPUs = 1
106 \\ /| Num examples = 56 | Num Epochs = 9
107O^O/ \_/ \ Batch size per device = 2 | Gradient Accumulation steps = 4
108\ / Total batch size = 8 | Total steps = 60
109 "-____-" Number of trainable parameters = 65,536,000
110
111 0%| | 0/60 [00:00<?, ?it/s]
112 2%|▏ | 1/60 [00:23<22:55, 23.31s/it]
113
114{'loss': 2.8325, 'grad_norm': 0.08582139760255814, 'learning_rate': 4e-05, 'epoch': 0.14}
115
116 2%|▏ | 1/60 [00:23<22:55, 23.31s/it]
117 3%|▎ | 2/60 [00:41<19:44, 20.43s/it]
118
119{'loss': 2.8472, 'grad_norm': 0.1093076840043068, 'learning_rate': 8e-05, 'epoch': 0.29}
120
121 3%|▎ | 2/60 [00:41<19:44, 20.43s/it]
122 5%|▌ | 3/60 [00:58<17:51, 18.81s/it]
123
124{'loss': 2.6442, 'grad_norm': 0.10155556350946426, 'learning_rate': 0.00012, 'epoch': 0.43}
125
126 5%|▌ | 3/60 [00:58<17:51, 18.81s/it]
127 7%|▋ | 4/60 [01:15<16:50, 18.05s/it]
128
129{'loss': 2.675, 'grad_norm': 0.0939909964799881, 'learning_rate': 0.00016, 'epoch': 0.57}
130
131 7%|▋ | 4/60 [01:15<16:50, 18.05s/it]
132 8%|▊ | 5/60 [01:32<16:10, 17.65s/it]
133
134{'loss': 2.496, 'grad_norm': 0.10907075554132462, 'learning_rate': 0.0002, 'epoch': 0.71}
135
136 8%|▊ | 5/60 [01:32<16:10, 17.65s/it]
137 10%|█ | 6/60 [01:49<15:41, 17.43s/it]
138
139{'loss': 2.7976, 'grad_norm': 0.12686112523078918, 'learning_rate': 0.00019636363636363636, 'epoch': 0.86}
140
141 10%|█ | 6/60 [01:49<15:41, 17.43s/it]
142 12%|█▏ | 7/60 [02:06<15:16, 17.30s/it]
143
144{'loss': 2.6129, 'grad_norm': 0.11858198791742325, 'learning_rate': 0.00019272727272727274, 'epoch': 1.0}
145
146 12%|█▏ | 7/60 [02:06<15:16, 17.30s/it]
147 13%|█▎ | 8/60 [02:23<14:55, 17.23s/it]
148
149{'loss': 2.5396, 'grad_norm': 0.11999646574258804, 'learning_rate': 0.0001890909090909091, 'epoch': 1.14}
150
151 13%|█▎ | 8/60 [02:23<14:55, 17.23s/it]
152 15%|█▌ | 9/60 [02:40<14:36, 17.18s/it]
153
154{'loss': 2.6044, 'grad_norm': 0.1258288472890854, 'learning_rate': 0.00018545454545454545, 'epoch': 1.29}
155
156 15%|█▌ | 9/60 [02:40<14:36, 17.18s/it]
157 17%|█▋ | 10/60 [02:57<14:17, 17.14s/it]
158
159{'loss': 2.4266, 'grad_norm': 0.12394820153713226, 'learning_rate': 0.00018181818181818183, 'epoch': 1.43}
160
161 17%|█▋ | 10/60 [02:57<14:17, 17.14s/it]
162 18%|█▊ | 11/60 [03:14<13:59, 17.13s/it]
163
164{'loss': 2.5893, 'grad_norm': 0.13283300399780273, 'learning_rate': 0.0001781818181818182, 'epoch': 1.57}
165
166 18%|█▊ | 11/60 [03:14<13:59, 17.13s/it]
167 20%|██ | 12/60 [03:31<13:42, 17.13s/it]
168
169{'loss': 2.7021, 'grad_norm': 0.1271115243434906, 'learning_rate': 0.00017454545454545454, 'epoch': 1.71}
170
171 20%|██ | 12/60 [03:31<13:42, 17.13s/it]
172 22%|██▏ | 13/60 [03:49<13:25, 17.13s/it]
173
174{'loss': 2.4765, 'grad_norm': 0.10840773582458496, 'learning_rate': 0.0001709090909090909, 'epoch': 1.86}
175
176 22%|██▏ | 13/60 [03:49<13:25, 17.13s/it]
177 23%|██▎ | 14/60 [04:06<13:08, 17.14s/it]
178
179{'loss': 2.5271, 'grad_norm': 0.10484796017408371, 'learning_rate': 0.00016727272727272728, 'epoch': 2.0}
180
181 23%|██▎ | 14/60 [04:06<13:08, 17.14s/it]
182 25%|██▌ | 15/60 [04:23<12:50, 17.13s/it]
183
184{'loss': 2.4216, 'grad_norm': 0.09565936028957367, 'learning_rate': 0.00016363636363636366, 'epoch': 2.14}
185
186 25%|██▌ | 15/60 [04:23<12:50, 17.13s/it]
187 27%|██▋ | 16/60 [04:40<12:33, 17.13s/it]
188
189{'loss': 2.4083, 'grad_norm': 0.08841613680124283, 'learning_rate': 0.00016, 'epoch': 2.29}
190
191 27%|██▋ | 16/60 [04:40<12:33, 17.13s/it]
192 28%|██▊ | 17/60 [04:57<12:16, 17.14s/it]
193
194{'loss': 2.5839, 'grad_norm': 0.09088816493749619, 'learning_rate': 0.00015636363636363637, 'epoch': 2.43}
195
196 28%|██▊ | 17/60 [04:57<12:16, 17.14s/it]
197 30%|███ | 18/60 [05:14<11:59, 17.13s/it]
198
199{'loss': 2.3157, 'grad_norm': 0.10538917034864426, 'learning_rate': 0.00015272727272727275, 'epoch': 2.57}
200
201 30%|███ | 18/60 [05:14<11:59, 17.13s/it]
202 32%|███▏ | 19/60 [05:31<11:42, 17.14s/it]
203
204{'loss': 2.5337, 'grad_norm': 0.10168007761240005, 'learning_rate': 0.0001490909090909091, 'epoch': 2.71}
205
206 32%|███▏ | 19/60 [05:31<11:42, 17.14s/it]
207 33%|███▎ | 20/60 [05:49<11:25, 17.14s/it]
208
209{'loss': 2.5389, 'grad_norm': 0.09013429284095764, 'learning_rate': 0.00014545454545454546, 'epoch': 2.86}
210
211 33%|███▎ | 20/60 [05:49<11:25, 17.14s/it]
212 35%|███▌ | 21/60 [06:06<11:08, 17.14s/it]
213
214{'loss': 2.5527, 'grad_norm': 0.10220775753259659, 'learning_rate': 0.00014181818181818184, 'epoch': 3.0}
215
216 35%|███▌ | 21/60 [06:06<11:08, 17.14s/it]
217 37%|███▋ | 22/60 [06:23<10:51, 17.14s/it]
218
219{'loss': 2.375, 'grad_norm': 0.09961222857236862, 'learning_rate': 0.0001381818181818182, 'epoch': 3.14}
220
221 37%|███▋ | 22/60 [06:23<10:51, 17.14s/it]
222 38%|███▊ | 23/60 [06:40<10:33, 17.13s/it]
223
224{'loss': 2.5698, 'grad_norm': 0.10702770948410034, 'learning_rate': 0.00013454545454545455, 'epoch': 3.29}
225
226 38%|███▊ | 23/60 [06:40<10:33, 17.13s/it]
227 40%|████ | 24/60 [06:57<10:16, 17.13s/it]
228
229{'loss': 2.3717, 'grad_norm': 0.10184632986783981, 'learning_rate': 0.00013090909090909093, 'epoch': 3.43}
230
231 40%|████ | 24/60 [06:57<10:16, 17.13s/it]
232 42%|████▏ | 25/60 [07:14<09:59, 17.13s/it]
233
234{'loss': 2.4711, 'grad_norm': 0.11484692990779877, 'learning_rate': 0.00012727272727272728, 'epoch': 3.57}
235
236 42%|████▏ | 25/60 [07:14<09:59, 17.13s/it]
237 43%|████▎ | 26/60 [07:31<09:42, 17.12s/it]
238
239{'loss': 2.4688, 'grad_norm': 0.12465589493513107, 'learning_rate': 0.00012363636363636364, 'epoch': 3.71}
240
241 43%|████▎ | 26/60 [07:31<09:42, 17.12s/it]
242 45%|████▌ | 27/60 [07:48<09:25, 17.12s/it]
243
244{'loss': 2.3375, 'grad_norm': 0.10505373030900955, 'learning_rate': 0.00012, 'epoch': 3.86}
245
246 45%|████▌ | 27/60 [07:48<09:25, 17.12s/it]
247 47%|████▋ | 28/60 [08:05<09:07, 17.12s/it]
248
249{'loss': 2.4719, 'grad_norm': 0.1254764050245285, 'learning_rate': 0.00011636363636363636, 'epoch': 4.0}
250
251 47%|████▋ | 28/60 [08:05<09:07, 17.12s/it]
252 48%|████▊ | 29/60 [08:23<08:50, 17.13s/it]
253
254{'loss': 2.4156, 'grad_norm': 0.1411900520324707, 'learning_rate': 0.00011272727272727272, 'epoch': 4.14}
255
256 48%|████▊ | 29/60 [08:23<08:50, 17.13s/it]
257 50%|█████ | 30/60 [08:40<08:33, 17.13s/it]
258
259{'loss': 2.3294, 'grad_norm': 0.12304367870092392, 'learning_rate': 0.00010909090909090909, 'epoch': 4.29}
260
261 50%|█████ | 30/60 [08:40<08:33, 17.13s/it]
262 52%|█████▏ | 31/60 [08:57<08:16, 17.13s/it]
263
264{'loss': 2.3492, 'grad_norm': 0.14749528467655182, 'learning_rate': 0.00010545454545454545, 'epoch': 4.43}
265
266 52%|█████▏ | 31/60 [08:57<08:16, 17.13s/it]
267 53%|█████▎ | 32/60 [09:14<07:59, 17.13s/it]
268
269{'loss': 2.2798, 'grad_norm': 0.14352062344551086, 'learning_rate': 0.00010181818181818181, 'epoch': 4.57}
270
271 53%|█████▎ | 32/60 [09:14<07:59, 17.13s/it]
272 55%|█████▌ | 33/60 [09:31<07:42, 17.13s/it]
273
274{'loss': 2.4132, 'grad_norm': 0.16667835414409637, 'learning_rate': 9.818181818181818e-05, 'epoch': 4.71}
275
276 55%|█████▌ | 33/60 [09:31<07:42, 17.13s/it]
277 57%|█████▋ | 34/60 [09:48<07:25, 17.13s/it]
278
279{'loss': 2.4788, 'grad_norm': 0.1789444237947464, 'learning_rate': 9.454545454545455e-05, 'epoch': 4.86}
280
281 57%|█████▋ | 34/60 [09:48<07:25, 17.13s/it]
282 58%|█████▊ | 35/60 [10:05<07:08, 17.12s/it]
283
284{'loss': 2.4947, 'grad_norm': 0.19302651286125183, 'learning_rate': 9.090909090909092e-05, 'epoch': 5.0}
285
286 58%|█████▊ | 35/60 [10:05<07:08, 17.12s/it]
287 60%|██████ | 36/60 [10:23<06:50, 17.12s/it]
288
289{'loss': 2.2137, 'grad_norm': 0.17413997650146484, 'learning_rate': 8.727272727272727e-05, 'epoch': 5.14}
290
291 60%|██████ | 36/60 [10:23<06:50, 17.12s/it]
292 62%|██████▏ | 37/60 [10:40<06:33, 17.12s/it]
293
294{'loss': 2.3259, 'grad_norm': 0.17235468327999115, 'learning_rate': 8.363636363636364e-05, 'epoch': 5.29}
295
296 62%|██████▏ | 37/60 [10:40<06:33, 17.12s/it]
297 63%|██████▎ | 38/60 [10:57<06:16, 17.12s/it]
298
299{'loss': 2.3787, 'grad_norm': 0.20865033566951752, 'learning_rate': 8e-05, 'epoch': 5.43}
300
301 63%|██████▎ | 38/60 [10:57<06:16, 17.12s/it]wandb: WARNING Fatal error while uploading data. Some run
302data will not be synced, but it will still be written to disk. Use `wandb sync` at the end of the run to try
303uploading.
304
305 65%|██████▌ | 39/60 [11:14<05:59, 17.12s/it]
306
307{'loss': 2.4876, 'grad_norm': 0.15544921159744263, 'learning_rate': 7.636363636363637e-05, 'epoch': 5.57}
308
309 65%|██████▌ | 39/60 [11:14<05:59, 17.12s/it]
310 67%|██████▋ | 40/60 [11:31<05:42, 17.12s/it]
311
312{'loss': 2.2552, 'grad_norm': 0.17868204414844513, 'learning_rate': 7.272727272727273e-05, 'epoch': 5.71}
313
314 67%|██████▋ | 40/60 [11:31<05:42, 17.12s/it]
315 68%|██████▊ | 41/60 [11:48<05:25, 17.13s/it]
316
317{'loss': 2.419, 'grad_norm': 0.20979967713356018, 'learning_rate': 6.90909090909091e-05, 'epoch': 5.86}
318
319 68%|██████▊ | 41/60 [11:48<05:25, 17.13s/it]
320 70%|███████ | 42/60 [12:05<05:08, 17.12s/it]
321
322{'loss': 2.3014, 'grad_norm': 0.20691952109336853, 'learning_rate': 6.545454545454546e-05, 'epoch': 6.0}
323
324 70%|███████ | 42/60 [12:05<05:08, 17.12s/it]
325 72%|███████▏ | 43/60 [12:22<04:51, 17.13s/it]
326
327{'loss': 2.4537, 'grad_norm': 0.1874876469373703, 'learning_rate': 6.181818181818182e-05, 'epoch': 6.14}
328
329 72%|███████▏ | 43/60 [12:22<04:51, 17.13s/it]
330 73%|███████▎ | 44/60 [12:40<04:33, 17.12s/it]
331
332{'loss': 2.4091, 'grad_norm': 0.17996680736541748, 'learning_rate': 5.818181818181818e-05, 'epoch': 6.29}
333
334 73%|███████▎ | 44/60 [12:40<04:33, 17.12s/it]
335 75%|███████▌ | 45/60 [12:57<04:16, 17.12s/it]
336
337{'loss': 2.3332, 'grad_norm': 0.19999149441719055, 'learning_rate': 5.4545454545454546e-05, 'epoch': 6.43}
338
339 75%|███████▌ | 45/60 [12:57<04:16, 17.12s/it]
340 77%|███████▋ | 46/60 [13:14<03:59, 17.12s/it]
341
342{'loss': 2.1857, 'grad_norm': 0.20532098412513733, 'learning_rate': 5.090909090909091e-05, 'epoch': 6.57}
343
344 77%|███████▋ | 46/60 [13:14<03:59, 17.12s/it]
345 78%|███████▊ | 47/60 [13:31<03:42, 17.12s/it]
346
347{'loss': 1.9826, 'grad_norm': 0.20537638664245605, 'learning_rate': 4.7272727272727275e-05, 'epoch': 6.71}
348
349 78%|███████▊ | 47/60 [13:31<03:42, 17.12s/it]
350 80%|████████ | 48/60 [13:48<03:25, 17.12s/it]
351
352{'loss': 2.3911, 'grad_norm': 0.2094985544681549, 'learning_rate': 4.3636363636363636e-05, 'epoch': 6.86}
353
354 80%|████████ | 48/60 [13:48<03:25, 17.12s/it]
355 82%|████████▏ | 49/60 [14:05<03:08, 17.11s/it]
356
357{'loss': 2.2917, 'grad_norm': 0.24426180124282837, 'learning_rate': 4e-05, 'epoch': 7.0}
358
359 82%|████████▏ | 49/60 [14:05<03:08, 17.11s/it]
360 83%|████████▎ | 50/60 [14:22<02:51, 17.11s/it]
361
362{'loss': 2.2391, 'grad_norm': 0.21321971714496613, 'learning_rate': 3.6363636363636364e-05, 'epoch': 7.14}
363
364 83%|████████▎ | 50/60 [14:22<02:51, 17.11s/it]
365 85%|████████▌ | 51/60 [14:39<02:34, 17.11s/it]
366
367{'loss': 2.1975, 'grad_norm': 0.20057713985443115, 'learning_rate': 3.272727272727273e-05, 'epoch': 7.29}
368
369 85%|████████▌ | 51/60 [14:39<02:34, 17.11s/it]
370 87%|████████▋ | 52/60 [14:56<02:17, 17.13s/it]
371
372{'loss': 2.1757, 'grad_norm': 0.1959208995103836, 'learning_rate': 2.909090909090909e-05, 'epoch': 7.43}
373
374 87%|████████▋ | 52/60 [14:56<02:17, 17.13s/it]
375 88%|████████▊ | 53/60 [15:14<01:59, 17.14s/it]
376
377{'loss': 2.3574, 'grad_norm': 0.24973013997077942, 'learning_rate': 2.5454545454545454e-05, 'epoch': 7.57}
378
379 88%|████████▊ | 53/60 [15:14<01:59, 17.14s/it]
380 90%|█████████ | 54/60 [15:31<01:42, 17.14s/it]
381
382{'loss': 2.3086, 'grad_norm': 0.22749045491218567, 'learning_rate': 2.1818181818181818e-05, 'epoch': 7.71}
383
384 90%|█████████ | 54/60 [15:31<01:42, 17.14s/it]
385 92%|█████████▏| 55/60 [15:48<01:25, 17.13s/it]
386
387{'loss': 2.2214, 'grad_norm': 0.20809921622276306, 'learning_rate': 1.8181818181818182e-05, 'epoch': 7.86}
388
389 92%|█████████▏| 55/60 [15:48<01:25, 17.13s/it]
390 93%|█████████▎| 56/60 [16:05<01:08, 17.13s/it]
391
392{'loss': 2.2949, 'grad_norm': 0.2743261158466339, 'learning_rate': 1.4545454545454545e-05, 'epoch': 8.0}
393
394 93%|█████████▎| 56/60 [16:05<01:08, 17.13s/it]
395 95%|█████████▌| 57/60 [16:22<00:51, 17.13s/it]
396
397{'loss': 2.3547, 'grad_norm': 0.22062085568904877, 'learning_rate': 1.0909090909090909e-05, 'epoch': 8.14}
398
399 95%|█████████▌| 57/60 [16:22<00:51, 17.13s/it]
400 97%|█████████▋| 58/60 [16:39<00:34, 17.12s/it]
401
402{'loss': 2.1947, 'grad_norm': 0.1845628023147583, 'learning_rate': 7.272727272727272e-06, 'epoch': 8.29}
403
404 97%|█████████▋| 58/60 [16:39<00:34, 17.12s/it]
405 98%|█████████▊| 59/60 [16:56<00:17, 17.12s/it]
406
407{'loss': 2.1337, 'grad_norm': 0.22526168823242188, 'learning_rate': 3.636363636363636e-06, 'epoch': 8.43}
408
409 98%|█████████▊| 59/60 [16:56<00:17, 17.12s/it]
410100%|██████████| 60/60 [17:13<00:00, 17.12s/it]
411
412{'loss': 2.1907, 'grad_norm': 0.22521968185901642, 'learning_rate': 0.0, 'epoch': 8.57}
413
414100%|██████████| 60/60 [17:13<00:00, 17.12s/it]wandb: Adding directory to artifact
415(./outputs/checkpoint-60)... Done. 0.9s
416
417
418{'train_runtime': 1039.7781, 'train_samples_per_second': 0.462, 'train_steps_per_second': 0.058,
419'train_loss': 2.417484853665034, 'epoch': 8.57}
420
421100%|██████████| 60/60 [17:18<00:00, 17.12s/it]
422100%|██████████| 60/60 [17:21<00:00, 17.35s/it]
423DEBUG: Training complete. Saving model and tokenizer locally...
424DEBUG: Saved model and tokenizer to 'lora_model'.
425Unsloth: You are pushing to hub, but you passed your HF username = mervinpraison.
426We shall truncate mervinpraison/Phi-4-ur-fall-bnb-4bit to Phi-4-ur-fall-bnb-4bit
427Unsloth: Merging 4bit and LoRA weights to 16bit...
428Unsloth: Will use up to 26.41 out of 47.13 RAM for saving.
429Unsloth: Saving model... This might take 5 minutes ...
430
431 0%| | 0/40 [00:00<?, ?it/s]
432 5%|▌ | 2/40 [00:00<00:02, 17.08it/s]
433 12%|█▎ | 5/40 [00:00<00:01, 23.72it/s]
434 20%|██ | 8/40 [00:00<00:01, 26.24it/s]
435 28%|██▊ | 11/40 [00:00<00:01, 27.43it/s]
436 35%|███▌ | 14/40 [00:00<00:00, 28.19it/s]
437 42%|████▎ | 17/40 [00:00<00:00, 28.44it/s]
438 50%|█████ | 20/40 [00:00<00:00, 28.74it/s]
439 57%|█████▊ | 23/40 [00:00<00:00, 28.98it/s]
440 65%|██████▌ | 26/40 [00:00<00:00, 29.14it/s]
441 72%|███████▎ | 29/40 [00:01<00:00, 29.08it/s]
442 80%|████████ | 32/40 [00:01<00:00, 29.19it/s]
443 88%|████████▊ | 35/40 [00:01<00:00, 29.28it/s]
444 95%|█████████▌| 38/40 [00:01<00:00, 29.29it/s]
445100%|██████████| 40/40 [00:01<00:00, 28.24it/s]
446Unsloth: Saving tokenizer... Done.