Hugging Face's logo

zail-ai
/
auramind-270m

Text Generation
Transformers
TensorBoard
Safetensors
gemma3_text
conversational-ai
mental-health
productivity
smartphone
mobile-ai
therapy
assistant
gemma
Eval Results (legacy)
text-generation-inference
Model card Files
xet
 Metrics Community
ibrahim256 commited on
Commit
4810a4e
·
verified ·
1 Parent(s): 9d7012e

Upload folder using huggingface_hub

Browse files
Files changed (39) hide show
  1. .gitattributes +3 -0
  2. README.md +43 -0
  3. added_tokens.json +9 -0
  4. checkpoint-1000/added_tokens.json +9 -0
  5. checkpoint-1000/config.json +54 -0
  6. checkpoint-1000/generation_config.json +7 -0
  7. checkpoint-1000/model.safetensors +3 -0
  8. checkpoint-1000/optimizer.pt +3 -0
  9. checkpoint-1000/rng_state.pth +3 -0
  10. checkpoint-1000/scheduler.pt +3 -0
  11. checkpoint-1000/special_tokens_map.json +77 -0
  12. checkpoint-1000/tokenizer.json +3 -0
  13. checkpoint-1000/tokenizer.model +3 -0
  14. checkpoint-1000/tokenizer_config.json +0 -0
  15. checkpoint-1000/trainer_state.json +215 -0
  16. checkpoint-1000/training_args.bin +3 -0
  17. checkpoint-500/added_tokens.json +9 -0
  18. checkpoint-500/config.json +54 -0
  19. checkpoint-500/generation_config.json +7 -0
  20. checkpoint-500/model.safetensors +3 -0
  21. checkpoint-500/optimizer.pt +3 -0
  22. checkpoint-500/rng_state.pth +3 -0
  23. checkpoint-500/scheduler.pt +3 -0
  24. checkpoint-500/special_tokens_map.json +77 -0
  25. checkpoint-500/tokenizer.json +3 -0
  26. checkpoint-500/tokenizer.model +3 -0
  27. checkpoint-500/tokenizer_config.json +0 -0
  28. checkpoint-500/trainer_state.json +129 -0
  29. checkpoint-500/training_args.bin +3 -0
  30. config.json +54 -0
  31. generation_config.json +7 -0
  32. logs/events.out.tfevents.1756308542.STUDIO-02.local.50997.0 +3 -0
  33. model.safetensors +3 -0
  34. special_tokens_map.json +77 -0
  35. tokenizer.json +3 -0
  36. tokenizer.model +3 -0
  37. tokenizer_config.json +0 -0
  38. training_args.bin +3 -0
  39. training_info.json +18 -0
.gitattributes CHANGED
@@ -33,3 +33,6 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
33
  *.zip filter=lfs diff=lfs merge=lfs -text
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
 
 
 
 
33
  *.zip filter=lfs diff=lfs merge=lfs -text
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
36
+ checkpoint-1000/tokenizer.json filter=lfs diff=lfs merge=lfs -text
37
+ checkpoint-500/tokenizer.json filter=lfs diff=lfs merge=lfs -text
38
+ tokenizer.json filter=lfs diff=lfs merge=lfs -text
README.md ADDED
@@ -0,0 +1,43 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # AuraMind Auramind 270
2
+
3
+ A warm and caring AI companion model fine-tuned for emotional support and personal connection.
4
+
5
+ ## Model Details
6
+ - **Base Model**: Gemma 3 270M (Released August 2025)
7
+ - **Parameters**: 268,102,656
8
+ - **Quantization**: INT4
9
+ - **Context Length**: 2,048 tokens
10
+ - **Vocabulary**: 256k tokens (large vocabulary for rare tokens)
11
+
12
+ ## Architecture Details
13
+ - **Total Parameters**: 270M (170M embedding + 100M transformer blocks)
14
+ - **Training Data**: 6 trillion tokens
15
+ - **Knowledge Cutoff**: August 2024
16
+
17
+ ## Training Details
18
+ - **Dataset**: AuraMind Companion Dataset
19
+ - **Training Steps**: 1,000
20
+ - **Final Loss**: 126.1678
21
+ - **Device**: MPS
22
+
23
+ ## Usage
24
+ ```python
25
+ from transformers import AutoTokenizer, AutoModelForCausalLM
26
+
27
+ tokenizer = AutoTokenizer.from_pretrained("models/auramind_270")
28
+ model = AutoModelForCausalLM.from_pretrained("models/auramind_270")
29
+
30
+ # Generate companion response
31
+ prompt = "<|system|>\nYou are AuraMind, a warm and caring AI companion. You form genuine emotional bonds with users through empathy, understanding, and personal connection. You provide intimate emotional support while maintaining healthy boundaries. Focus on creating deep, meaningful relationships through authentic conversation.\n\n<|user|>\nI'm feeling lonely today.\n\n<|assistant|>\n"
32
+ inputs = tokenizer(prompt, return_tensors="pt")
33
+ outputs = model.generate(**inputs, max_length=200, temperature=0.7, do_sample=True)
34
+ response = tokenizer.decode(outputs[0], skip_special_tokens=False)
35
+ ```
36
+
37
+ ## Capabilities
38
+ - Emotional support and companionship
39
+ - Personal conversation and bonding
40
+ - Empathetic responses
41
+ - Intimate but appropriate interactions
42
+
43
+ **Note**: This is a companion AI, not a replacement for professional therapy.
added_tokens.json ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "<image_soft_token>": 262144,
3
+ "<|assistant|>": 262147,
4
+ "<|companion|>": 262148,
5
+ "<|emotion|>": 262149,
6
+ "<|support|>": 262150,
7
+ "<|system|>": 262145,
8
+ "<|user|>": 262146
9
+ }
checkpoint-1000/added_tokens.json ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "<image_soft_token>": 262144,
3
+ "<|assistant|>": 262147,
4
+ "<|companion|>": 262148,
5
+ "<|emotion|>": 262149,
6
+ "<|support|>": 262150,
7
+ "<|system|>": 262145,
8
+ "<|user|>": 262146
9
+ }
checkpoint-1000/config.json ADDED
@@ -0,0 +1,54 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_sliding_window_pattern": 6,
3
+ "architectures": [
4
+ "Gemma3ForCausalLM"
5
+ ],
6
+ "attention_bias": false,
7
+ "attention_dropout": 0.0,
8
+ "attn_logit_softcapping": null,
9
+ "bos_token_id": 2,
10
+ "eos_token_id": 1,
11
+ "final_logit_softcapping": null,
12
+ "head_dim": 256,
13
+ "hidden_activation": "gelu_pytorch_tanh",
14
+ "hidden_size": 640,
15
+ "initializer_range": 0.02,
16
+ "intermediate_size": 2048,
17
+ "layer_types": [
18
+ "sliding_attention",
19
+ "sliding_attention",
20
+ "sliding_attention",
21
+ "sliding_attention",
22
+ "sliding_attention",
23
+ "full_attention",
24
+ "sliding_attention",
25
+ "sliding_attention",
26
+ "sliding_attention",
27
+ "sliding_attention",
28
+ "sliding_attention",
29
+ "full_attention",
30
+ "sliding_attention",
31
+ "sliding_attention",
32
+ "sliding_attention",
33
+ "sliding_attention",
34
+ "sliding_attention",
35
+ "full_attention"
36
+ ],
37
+ "max_position_embeddings": 32768,
38
+ "model_type": "gemma3_text",
39
+ "num_attention_heads": 4,
40
+ "num_hidden_layers": 18,
41
+ "num_key_value_heads": 1,
42
+ "pad_token_id": 0,
43
+ "query_pre_attn_scalar": 256,
44
+ "rms_norm_eps": 1e-06,
45
+ "rope_local_base_freq": 10000.0,
46
+ "rope_scaling": null,
47
+ "rope_theta": 1000000.0,
48
+ "sliding_window": 512,
49
+ "torch_dtype": "float16",
50
+ "transformers_version": "4.55.3",
51
+ "use_bidirectional_attention": false,
52
+ "use_cache": true,
53
+ "vocab_size": 262151
54
+ }
checkpoint-1000/generation_config.json ADDED
@@ -0,0 +1,7 @@
 
 
 
 
 
 
 
 
1
+ {
2
+ "cache_implementation": "hybrid",
3
+ "do_sample": true,
4
+ "top_k": 64,
5
+ "top_p": 0.95,
6
+ "transformers_version": "4.55.3"
7
+ }
checkpoint-1000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3d3340b057dfd1aeaeedf96b53e8f7cdd887c0b37e12c351f4afd6a5baafdb70
3
+ size 536231776
checkpoint-1000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9eb3c19b54415883fb318d378dfcfb6f0e25e574e205c5df15e6d0a9bd475343
3
+ size 1072602443
checkpoint-1000/rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5ca8ceba7f21f2d826c74925228bb7e31b692e58c2a0108653abaa8d547a15c7
3
+ size 14391
checkpoint-1000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ebcbfd37c5f4556961ec3f170f39cb08422150b96b0d9fff1ed7b77a9bb8eecb
3
+ size 1465
checkpoint-1000/special_tokens_map.json ADDED
@@ -0,0 +1,77 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "additional_special_tokens": [
3
+ {
4
+ "content": "<|system|>",
5
+ "lstrip": false,
6
+ "normalized": false,
7
+ "rstrip": false,
8
+ "single_word": false
9
+ },
10
+ {
11
+ "content": "<|user|>",
12
+ "lstrip": false,
13
+ "normalized": false,
14
+ "rstrip": false,
15
+ "single_word": false
16
+ },
17
+ {
18
+ "content": "<|assistant|>",
19
+ "lstrip": false,
20
+ "normalized": false,
21
+ "rstrip": false,
22
+ "single_word": false
23
+ },
24
+ {
25
+ "content": "<|companion|>",
26
+ "lstrip": false,
27
+ "normalized": false,
28
+ "rstrip": false,
29
+ "single_word": false
30
+ },
31
+ {
32
+ "content": "<|emotion|>",
33
+ "lstrip": false,
34
+ "normalized": false,
35
+ "rstrip": false,
36
+ "single_word": false
37
+ },
38
+ {
39
+ "content": "<|support|>",
40
+ "lstrip": false,
41
+ "normalized": false,
42
+ "rstrip": false,
43
+ "single_word": false
44
+ }
45
+ ],
46
+ "boi_token": "<start_of_image>",
47
+ "bos_token": {
48
+ "content": "<bos>",
49
+ "lstrip": false,
50
+ "normalized": false,
51
+ "rstrip": false,
52
+ "single_word": false
53
+ },
54
+ "eoi_token": "<end_of_image>",
55
+ "eos_token": {
56
+ "content": "<eos>",
57
+ "lstrip": false,
58
+ "normalized": false,
59
+ "rstrip": false,
60
+ "single_word": false
61
+ },
62
+ "image_token": "<image_soft_token>",
63
+ "pad_token": {
64
+ "content": "<pad>",
65
+ "lstrip": false,
66
+ "normalized": false,
67
+ "rstrip": false,
68
+ "single_word": false
69
+ },
70
+ "unk_token": {
71
+ "content": "<unk>",
72
+ "lstrip": false,
73
+ "normalized": false,
74
+ "rstrip": false,
75
+ "single_word": false
76
+ }
77
+ }
checkpoint-1000/tokenizer.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d1e381d8ad2489a1da8ba628f2e7de10678e8b847f45e0ec06a75208a154eed1
3
+ size 33385795
checkpoint-1000/tokenizer.model ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1299c11d7cf632ef3b4e11937501358ada021bbdf7c47638d13c0ee982f2e79c
3
+ size 4689074
checkpoint-1000/tokenizer_config.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-1000/trainer_state.json ADDED
@@ -0,0 +1,215 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_global_step": null,
3
+ "best_metric": Infinity,
4
+ "best_model_checkpoint": null,
5
+ "epoch": 29.42,
6
+ "eval_steps": 250,
7
+ "global_step": 1000,
8
+ "is_hyper_param_search": false,
9
+ "is_local_process_zero": true,
10
+ "is_world_process_zero": true,
11
+ "log_history": [
12
+ {
13
+ "epoch": 1.48,
14
+ "grad_norm": NaN,
15
+ "learning_rate": 4.9e-05,
16
+ "loss": 2523.3566,
17
+ "step": 50
18
+ },
19
+ {
20
+ "epoch": 2.96,
21
+ "grad_norm": NaN,
22
+ "learning_rate": 9.900000000000001e-05,
23
+ "loss": 0.0,
24
+ "step": 100
25
+ },
26
+ {
27
+ "epoch": 4.42,
28
+ "grad_norm": NaN,
29
+ "learning_rate": 9.831034482758622e-05,
30
+ "loss": 0.0,
31
+ "step": 150
32
+ },
33
+ {
34
+ "epoch": 5.9,
35
+ "grad_norm": NaN,
36
+ "learning_rate": 9.658620689655173e-05,
37
+ "loss": 0.0,
38
+ "step": 200
39
+ },
40
+ {
41
+ "epoch": 7.36,
42
+ "grad_norm": NaN,
43
+ "learning_rate": 9.486206896551724e-05,
44
+ "loss": 0.0,
45
+ "step": 250
46
+ },
47
+ {
48
+ "epoch": 7.36,
49
+ "eval_loss": NaN,
50
+ "eval_runtime": 0.8249,
51
+ "eval_samples_per_second": 127.294,
52
+ "eval_steps_per_second": 10.911,
53
+ "step": 250
54
+ },
55
+ {
56
+ "epoch": 8.84,
57
+ "grad_norm": NaN,
58
+ "learning_rate": 9.313793103448276e-05,
59
+ "loss": 0.0,
60
+ "step": 300
61
+ },
62
+ {
63
+ "epoch": 10.3,
64
+ "grad_norm": NaN,
65
+ "learning_rate": 9.141379310344827e-05,
66
+ "loss": 0.0,
67
+ "step": 350
68
+ },
69
+ {
70
+ "epoch": 11.78,
71
+ "grad_norm": NaN,
72
+ "learning_rate": 8.96896551724138e-05,
73
+ "loss": 0.0,
74
+ "step": 400
75
+ },
76
+ {
77
+ "epoch": 13.24,
78
+ "grad_norm": NaN,
79
+ "learning_rate": 8.796551724137931e-05,
80
+ "loss": 0.0,
81
+ "step": 450
82
+ },
83
+ {
84
+ "epoch": 14.72,
85
+ "grad_norm": NaN,
86
+ "learning_rate": 8.624137931034483e-05,
87
+ "loss": 0.0,
88
+ "step": 500
89
+ },
90
+ {
91
+ "epoch": 14.72,
92
+ "eval_loss": NaN,
93
+ "eval_runtime": 0.4313,
94
+ "eval_samples_per_second": 243.441,
95
+ "eval_steps_per_second": 20.866,
96
+ "step": 500
97
+ },
98
+ {
99
+ "epoch": 16.18,
100
+ "grad_norm": NaN,
101
+ "learning_rate": 8.451724137931036e-05,
102
+ "loss": 0.0,
103
+ "step": 550
104
+ },
105
+ {
106
+ "epoch": 17.66,
107
+ "grad_norm": NaN,
108
+ "learning_rate": 8.279310344827587e-05,
109
+ "loss": 0.0,
110
+ "step": 600
111
+ },
112
+ {
113
+ "epoch": 19.12,
114
+ "grad_norm": NaN,
115
+ "learning_rate": 8.106896551724138e-05,
116
+ "loss": 0.0,
117
+ "step": 650
118
+ },
119
+ {
120
+ "epoch": 20.6,
121
+ "grad_norm": NaN,
122
+ "learning_rate": 7.93448275862069e-05,
123
+ "loss": 0.0,
124
+ "step": 700
125
+ },
126
+ {
127
+ "epoch": 22.06,
128
+ "grad_norm": NaN,
129
+ "learning_rate": 7.762068965517241e-05,
130
+ "loss": 0.0,
131
+ "step": 750
132
+ },
133
+ {
134
+ "epoch": 22.06,
135
+ "eval_loss": NaN,
136
+ "eval_runtime": 0.4262,
137
+ "eval_samples_per_second": 246.338,
138
+ "eval_steps_per_second": 21.115,
139
+ "step": 750
140
+ },
141
+ {
142
+ "epoch": 23.54,
143
+ "grad_norm": NaN,
144
+ "learning_rate": 7.589655172413793e-05,
145
+ "loss": 0.0,
146
+ "step": 800
147
+ },
148
+ {
149
+ "epoch": 25.0,
150
+ "grad_norm": NaN,
151
+ "learning_rate": 7.417241379310345e-05,
152
+ "loss": 0.0,
153
+ "step": 850
154
+ },
155
+ {
156
+ "epoch": 26.48,
157
+ "grad_norm": NaN,
158
+ "learning_rate": 7.244827586206897e-05,
159
+ "loss": 0.0,
160
+ "step": 900
161
+ },
162
+ {
163
+ "epoch": 27.96,
164
+ "grad_norm": NaN,
165
+ "learning_rate": 7.072413793103448e-05,
166
+ "loss": 0.0,
167
+ "step": 950
168
+ },
169
+ {
170
+ "epoch": 29.42,
171
+ "grad_norm": NaN,
172
+ "learning_rate": 6.9e-05,
173
+ "loss": 0.0,
174
+ "step": 1000
175
+ },
176
+ {
177
+ "epoch": 29.42,
178
+ "eval_loss": NaN,
179
+ "eval_runtime": 0.4242,
180
+ "eval_samples_per_second": 247.509,
181
+ "eval_steps_per_second": 21.215,
182
+ "step": 1000
183
+ }
184
+ ],
185
+ "logging_steps": 50,
186
+ "max_steps": 3000,
187
+ "num_input_tokens_seen": 0,
188
+ "num_train_epochs": 89,
189
+ "save_steps": 500,
190
+ "stateful_callbacks": {
191
+ "EarlyStoppingCallback": {
192
+ "args": {
193
+ "early_stopping_patience": 3,
194
+ "early_stopping_threshold": 0.0
195
+ },
196
+ "attributes": {
197
+ "early_stopping_patience_counter": 3
198
+ }
199
+ },
200
+ "TrainerControl": {
201
+ "args": {
202
+ "should_epoch_stop": false,
203
+ "should_evaluate": false,
204
+ "should_log": false,
205
+ "should_save": true,
206
+ "should_training_stop": true
207
+ },
208
+ "attributes": {}
209
+ }
210
+ },
211
+ "total_flos": 1204367872764672.0,
212
+ "train_batch_size": 12,
213
+ "trial_name": null,
214
+ "trial_params": null
215
+ }
checkpoint-1000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3bfa38633b612b2effa711b007d658cdf00c08e0f47bdca6d839ac78c67891e4
3
+ size 5713
checkpoint-500/added_tokens.json ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "<image_soft_token>": 262144,
3
+ "<|assistant|>": 262147,
4
+ "<|companion|>": 262148,
5
+ "<|emotion|>": 262149,
6
+ "<|support|>": 262150,
7
+ "<|system|>": 262145,
8
+ "<|user|>": 262146
9
+ }
checkpoint-500/config.json ADDED
@@ -0,0 +1,54 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_sliding_window_pattern": 6,
3
+ "architectures": [
4
+ "Gemma3ForCausalLM"
5
+ ],
6
+ "attention_bias": false,
7
+ "attention_dropout": 0.0,
8
+ "attn_logit_softcapping": null,
9
+ "bos_token_id": 2,
10
+ "eos_token_id": 1,
11
+ "final_logit_softcapping": null,
12
+ "head_dim": 256,
13
+ "hidden_activation": "gelu_pytorch_tanh",
14
+ "hidden_size": 640,
15
+ "initializer_range": 0.02,
16
+ "intermediate_size": 2048,
17
+ "layer_types": [
18
+ "sliding_attention",
19
+ "sliding_attention",
20
+ "sliding_attention",
21
+ "sliding_attention",
22
+ "sliding_attention",
23
+ "full_attention",
24
+ "sliding_attention",
25
+ "sliding_attention",
26
+ "sliding_attention",
27
+ "sliding_attention",
28
+ "sliding_attention",
29
+ "full_attention",
30
+ "sliding_attention",
31
+ "sliding_attention",
32
+ "sliding_attention",
33
+ "sliding_attention",
34
+ "sliding_attention",
35
+ "full_attention"
36
+ ],
37
+ "max_position_embeddings": 32768,
38
+ "model_type": "gemma3_text",
39
+ "num_attention_heads": 4,
40
+ "num_hidden_layers": 18,
41
+ "num_key_value_heads": 1,
42
+ "pad_token_id": 0,
43
+ "query_pre_attn_scalar": 256,
44
+ "rms_norm_eps": 1e-06,
45
+ "rope_local_base_freq": 10000.0,
46
+ "rope_scaling": null,
47
+ "rope_theta": 1000000.0,
48
+ "sliding_window": 512,
49
+ "torch_dtype": "float16",
50
+ "transformers_version": "4.55.3",
51
+ "use_bidirectional_attention": false,
52
+ "use_cache": true,
53
+ "vocab_size": 262151
54
+ }
checkpoint-500/generation_config.json ADDED
@@ -0,0 +1,7 @@
 
 
 
 
 
 
 
 
1
+ {
2
+ "cache_implementation": "hybrid",
3
+ "do_sample": true,
4
+ "top_k": 64,
5
+ "top_p": 0.95,
6
+ "transformers_version": "4.55.3"
7
+ }
checkpoint-500/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3d3340b057dfd1aeaeedf96b53e8f7cdd887c0b37e12c351f4afd6a5baafdb70
3
+ size 536231776
checkpoint-500/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2294fc6a49669b0150080790bf8f1174b536a062fe5a5e6a63754094d6e356c1
3
+ size 1072602443
checkpoint-500/rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:51df1d95d8ccbb90e460324f1b442600bbbce2875f5c7ef87dea4ee7fa279f7b
3
+ size 14391
checkpoint-500/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7671eab3e5bcbb36d4ff91ece936d046f377cd3effe834c4cc694c495d6c32da
3
+ size 1465
checkpoint-500/special_tokens_map.json ADDED
@@ -0,0 +1,77 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "additional_special_tokens": [
3
+ {
4
+ "content": "<|system|>",
5
+ "lstrip": false,
6
+ "normalized": false,
7
+ "rstrip": false,
8
+ "single_word": false
9
+ },
10
+ {
11
+ "content": "<|user|>",
12
+ "lstrip": false,
13
+ "normalized": false,
14
+ "rstrip": false,
15
+ "single_word": false
16
+ },
17
+ {
18
+ "content": "<|assistant|>",
19
+ "lstrip": false,
20
+ "normalized": false,
21
+ "rstrip": false,
22
+ "single_word": false
23
+ },
24
+ {
25
+ "content": "<|companion|>",
26
+ "lstrip": false,
27
+ "normalized": false,
28
+ "rstrip": false,
29
+ "single_word": false
30
+ },
31
+ {
32
+ "content": "<|emotion|>",
33
+ "lstrip": false,
34
+ "normalized": false,
35
+ "rstrip": false,
36
+ "single_word": false
37
+ },
38
+ {
39
+ "content": "<|support|>",
40
+ "lstrip": false,
41
+ "normalized": false,
42
+ "rstrip": false,
43
+ "single_word": false
44
+ }
45
+ ],
46
+ "boi_token": "<start_of_image>",
47
+ "bos_token": {
48
+ "content": "<bos>",
49
+ "lstrip": false,
50
+ "normalized": false,
51
+ "rstrip": false,
52
+ "single_word": false
53
+ },
54
+ "eoi_token": "<end_of_image>",
55
+ "eos_token": {
56
+ "content": "<eos>",
57
+ "lstrip": false,
58
+ "normalized": false,
59
+ "rstrip": false,
60
+ "single_word": false
61
+ },
62
+ "image_token": "<image_soft_token>",
63
+ "pad_token": {
64
+ "content": "<pad>",
65
+ "lstrip": false,
66
+ "normalized": false,
67
+ "rstrip": false,
68
+ "single_word": false
69
+ },
70
+ "unk_token": {
71
+ "content": "<unk>",
72
+ "lstrip": false,
73
+ "normalized": false,
74
+ "rstrip": false,
75
+ "single_word": false
76
+ }
77
+ }
checkpoint-500/tokenizer.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d1e381d8ad2489a1da8ba628f2e7de10678e8b847f45e0ec06a75208a154eed1
3
+ size 33385795
checkpoint-500/tokenizer.model ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1299c11d7cf632ef3b4e11937501358ada021bbdf7c47638d13c0ee982f2e79c
3
+ size 4689074
checkpoint-500/tokenizer_config.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-500/trainer_state.json ADDED
@@ -0,0 +1,129 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_global_step": null,
3
+ "best_metric": Infinity,
4
+ "best_model_checkpoint": null,
5
+ "epoch": 14.72,
6
+ "eval_steps": 250,
7
+ "global_step": 500,
8
+ "is_hyper_param_search": false,
9
+ "is_local_process_zero": true,
10
+ "is_world_process_zero": true,
11
+ "log_history": [
12
+ {
13
+ "epoch": 1.48,
14
+ "grad_norm": NaN,
15
+ "learning_rate": 4.9e-05,
16
+ "loss": 2523.3566,
17
+ "step": 50
18
+ },
19
+ {
20
+ "epoch": 2.96,
21
+ "grad_norm": NaN,
22
+ "learning_rate": 9.900000000000001e-05,
23
+ "loss": 0.0,
24
+ "step": 100
25
+ },
26
+ {
27
+ "epoch": 4.42,
28
+ "grad_norm": NaN,
29
+ "learning_rate": 9.831034482758622e-05,
30
+ "loss": 0.0,
31
+ "step": 150
32
+ },
33
+ {
34
+ "epoch": 5.9,
35
+ "grad_norm": NaN,
36
+ "learning_rate": 9.658620689655173e-05,
37
+ "loss": 0.0,
38
+ "step": 200
39
+ },
40
+ {
41
+ "epoch": 7.36,
42
+ "grad_norm": NaN,
43
+ "learning_rate": 9.486206896551724e-05,
44
+ "loss": 0.0,
45
+ "step": 250
46
+ },
47
+ {
48
+ "epoch": 7.36,
49
+ "eval_loss": NaN,
50
+ "eval_runtime": 0.8249,
51
+ "eval_samples_per_second": 127.294,
52
+ "eval_steps_per_second": 10.911,
53
+ "step": 250
54
+ },
55
+ {
56
+ "epoch": 8.84,
57
+ "grad_norm": NaN,
58
+ "learning_rate": 9.313793103448276e-05,
59
+ "loss": 0.0,
60
+ "step": 300
61
+ },
62
+ {
63
+ "epoch": 10.3,
64
+ "grad_norm": NaN,
65
+ "learning_rate": 9.141379310344827e-05,
66
+ "loss": 0.0,
67
+ "step": 350
68
+ },
69
+ {
70
+ "epoch": 11.78,
71
+ "grad_norm": NaN,
72
+ "learning_rate": 8.96896551724138e-05,
73
+ "loss": 0.0,
74
+ "step": 400
75
+ },
76
+ {
77
+ "epoch": 13.24,
78
+ "grad_norm": NaN,
79
+ "learning_rate": 8.796551724137931e-05,
80
+ "loss": 0.0,
81
+ "step": 450
82
+ },
83
+ {
84
+ "epoch": 14.72,
85
+ "grad_norm": NaN,
86
+ "learning_rate": 8.624137931034483e-05,
87
+ "loss": 0.0,
88
+ "step": 500
89
+ },
90
+ {
91
+ "epoch": 14.72,
92
+ "eval_loss": NaN,
93
+ "eval_runtime": 0.4313,
94
+ "eval_samples_per_second": 243.441,
95
+ "eval_steps_per_second": 20.866,
96
+ "step": 500
97
+ }
98
+ ],
99
+ "logging_steps": 50,
100
+ "max_steps": 3000,
101
+ "num_input_tokens_seen": 0,
102
+ "num_train_epochs": 89,
103
+ "save_steps": 500,
104
+ "stateful_callbacks": {
105
+ "EarlyStoppingCallback": {
106
+ "args": {
107
+ "early_stopping_patience": 3,
108
+ "early_stopping_threshold": 0.0
109
+ },
110
+ "attributes": {
111
+ "early_stopping_patience_counter": 1
112
+ }
113
+ },
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": false
121
+ },
122
+ "attributes": {}
123
+ }
124
+ },
125
+ "total_flos": 602715764593152.0,
126
+ "train_batch_size": 12,
127
+ "trial_name": null,
128
+ "trial_params": null
129
+ }
checkpoint-500/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3bfa38633b612b2effa711b007d658cdf00c08e0f47bdca6d839ac78c67891e4
3
+ size 5713
config.json ADDED
@@ -0,0 +1,54 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_sliding_window_pattern": 6,
3
+ "architectures": [
4
+ "Gemma3ForCausalLM"
5
+ ],
6
+ "attention_bias": false,
7
+ "attention_dropout": 0.0,
8
+ "attn_logit_softcapping": null,
9
+ "bos_token_id": 2,
10
+ "eos_token_id": 1,
11
+ "final_logit_softcapping": null,
12
+ "head_dim": 256,
13
+ "hidden_activation": "gelu_pytorch_tanh",
14
+ "hidden_size": 640,
15
+ "initializer_range": 0.02,
16
+ "intermediate_size": 2048,
17
+ "layer_types": [
18
+ "sliding_attention",
19
+ "sliding_attention",
20
+ "sliding_attention",
21
+ "sliding_attention",
22
+ "sliding_attention",
23
+ "full_attention",
24
+ "sliding_attention",
25
+ "sliding_attention",
26
+ "sliding_attention",
27
+ "sliding_attention",
28
+ "sliding_attention",
29
+ "full_attention",
30
+ "sliding_attention",
31
+ "sliding_attention",
32
+ "sliding_attention",
33
+ "sliding_attention",
34
+ "sliding_attention",
35
+ "full_attention"
36
+ ],
37
+ "max_position_embeddings": 32768,
38
+ "model_type": "gemma3_text",
39
+ "num_attention_heads": 4,
40
+ "num_hidden_layers": 18,
41
+ "num_key_value_heads": 1,
42
+ "pad_token_id": 0,
43
+ "query_pre_attn_scalar": 256,
44
+ "rms_norm_eps": 1e-06,
45
+ "rope_local_base_freq": 10000.0,
46
+ "rope_scaling": null,
47
+ "rope_theta": 1000000.0,
48
+ "sliding_window": 512,
49
+ "torch_dtype": "float16",
50
+ "transformers_version": "4.55.3",
51
+ "use_bidirectional_attention": false,
52
+ "use_cache": true,
53
+ "vocab_size": 262151
54
+ }
generation_config.json ADDED
@@ -0,0 +1,7 @@
 
 
 
 
 
 
 
 
1
+ {
2
+ "cache_implementation": "hybrid",
3
+ "do_sample": true,
4
+ "top_k": 64,
5
+ "top_p": 0.95,
6
+ "transformers_version": "4.55.3"
7
+ }
logs/events.out.tfevents.1756308542.STUDIO-02.local.50997.0 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:871073231bdfb5d264f99577cb7154fc6882d4d2feb79e858a15419561bb4917
3
+ size 11313
model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3d3340b057dfd1aeaeedf96b53e8f7cdd887c0b37e12c351f4afd6a5baafdb70
3
+ size 536231776
special_tokens_map.json ADDED
@@ -0,0 +1,77 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "additional_special_tokens": [
3
+ {
4
+ "content": "<|system|>",
5
+ "lstrip": false,
6
+ "normalized": false,
7
+ "rstrip": false,
8
+ "single_word": false
9
+ },
10
+ {
11
+ "content": "<|user|>",
12
+ "lstrip": false,
13
+ "normalized": false,
14
+ "rstrip": false,
15
+ "single_word": false
16
+ },
17
+ {
18
+ "content": "<|assistant|>",
19
+ "lstrip": false,
20
+ "normalized": false,
21
+ "rstrip": false,
22
+ "single_word": false
23
+ },
24
+ {
25
+ "content": "<|companion|>",
26
+ "lstrip": false,
27
+ "normalized": false,
28
+ "rstrip": false,
29
+ "single_word": false
30
+ },
31
+ {
32
+ "content": "<|emotion|>",
33
+ "lstrip": false,
34
+ "normalized": false,
35
+ "rstrip": false,
36
+ "single_word": false
37
+ },
38
+ {
39
+ "content": "<|support|>",
40
+ "lstrip": false,
41
+ "normalized": false,
42
+ "rstrip": false,
43
+ "single_word": false
44
+ }
45
+ ],
46
+ "boi_token": "<start_of_image>",
47
+ "bos_token": {
48
+ "content": "<bos>",
49
+ "lstrip": false,
50
+ "normalized": false,
51
+ "rstrip": false,
52
+ "single_word": false
53
+ },
54
+ "eoi_token": "<end_of_image>",
55
+ "eos_token": {
56
+ "content": "<eos>",
57
+ "lstrip": false,
58
+ "normalized": false,
59
+ "rstrip": false,
60
+ "single_word": false
61
+ },
62
+ "image_token": "<image_soft_token>",
63
+ "pad_token": {
64
+ "content": "<pad>",
65
+ "lstrip": false,
66
+ "normalized": false,
67
+ "rstrip": false,
68
+ "single_word": false
69
+ },
70
+ "unk_token": {
71
+ "content": "<unk>",
72
+ "lstrip": false,
73
+ "normalized": false,
74
+ "rstrip": false,
75
+ "single_word": false
76
+ }
77
+ }
tokenizer.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d1e381d8ad2489a1da8ba628f2e7de10678e8b847f45e0ec06a75208a154eed1
3
+ size 33385795
tokenizer.model ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1299c11d7cf632ef3b4e11937501358ada021bbdf7c47638d13c0ee982f2e79c
3
+ size 4689074
tokenizer_config.json ADDED
The diff for this file is too large to render. See raw diff
 
training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3bfa38633b612b2effa711b007d658cdf00c08e0f47bdca6d839ac78c67891e4
3
+ size 5713
training_info.json ADDED
@@ -0,0 +1,18 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model_variant": "auramind_270",
3
+ "training_loss": 126.167828125,
4
+ "training_steps": 1000,
5
+ "model_parameters": 268102656,
6
+ "config": {
7
+ "base_model": "google/gemma-3-270m",
8
+ "target_params": 270000000,
9
+ "quantization": "int4",
10
+ "max_length": 2048,
11
+ "batch_size": 12,
12
+ "gradient_accumulation": 3,
13
+ "learning_rate": 0.0001,
14
+ "warmup_steps": 100,
15
+ "max_steps": 3000
16
+ },
17
+ "training_time": "2025-08-27 18:37:14.899613"
18
+ }