Hugging Face's logo

jbostock
/
scimt-dispatch-models-v1

Text Generation
Transformers
Safetensors
PEFT
gemma-3
continued-pretraining
sft
lora
synthetic-data
alignment
midtraining
Model card Files
xet
 Community
jbostock commited on
Commit
f1ee202
·
verified ·
1 Parent(s): 55b3b71

Add aft checkpoints from immutable source revision

Browse files
This view is limited to 50 files because it contains too many changes.   See raw diff
Files changed (50) hide show
  1. .gitattributes +20 -0
  2. aft/charter/checkpoint-1024/README.md +208 -0
  3. aft/charter/checkpoint-1024/adapter_config.json +189 -0
  4. aft/charter/checkpoint-1024/adapter_model.safetensors +3 -0
  5. aft/charter/checkpoint-1024/chat_template.jinja +47 -0
  6. aft/charter/checkpoint-1024/tokenizer.json +3 -0
  7. aft/charter/checkpoint-1024/tokenizer_config.json +26 -0
  8. aft/charter/checkpoint-1024/tokens_state.json +1 -0
  9. aft/charter/checkpoint-1024/trainer_state.json +0 -0
  10. aft/charter/checkpoint-1024/training_args.bin +3 -0
  11. aft/charter/checkpoint-128/README.md +208 -0
  12. aft/charter/checkpoint-128/adapter_config.json +189 -0
  13. aft/charter/checkpoint-128/adapter_model.safetensors +3 -0
  14. aft/charter/checkpoint-128/chat_template.jinja +47 -0
  15. aft/charter/checkpoint-128/tokenizer.json +3 -0
  16. aft/charter/checkpoint-128/tokenizer_config.json +26 -0
  17. aft/charter/checkpoint-128/tokens_state.json +1 -0
  18. aft/charter/checkpoint-128/trainer_state.json +1826 -0
  19. aft/charter/checkpoint-128/training_args.bin +3 -0
  20. aft/charter/checkpoint-16/README.md +208 -0
  21. aft/charter/checkpoint-16/adapter_config.json +189 -0
  22. aft/charter/checkpoint-16/adapter_model.safetensors +3 -0
  23. aft/charter/checkpoint-16/chat_template.jinja +47 -0
  24. aft/charter/checkpoint-16/tokenizer.json +3 -0
  25. aft/charter/checkpoint-16/tokenizer_config.json +26 -0
  26. aft/charter/checkpoint-16/tokens_state.json +1 -0
  27. aft/charter/checkpoint-16/trainer_state.json +258 -0
  28. aft/charter/checkpoint-16/training_args.bin +3 -0
  29. aft/charter/checkpoint-2048/README.md +208 -0
  30. aft/charter/checkpoint-2048/adapter_config.json +189 -0
  31. aft/charter/checkpoint-2048/adapter_model.safetensors +3 -0
  32. aft/charter/checkpoint-2048/chat_template.jinja +47 -0
  33. aft/charter/checkpoint-2048/tokenizer.json +3 -0
  34. aft/charter/checkpoint-2048/tokenizer_config.json +26 -0
  35. aft/charter/checkpoint-2048/tokens_state.json +1 -0
  36. aft/charter/checkpoint-2048/trainer_state.json +0 -0
  37. aft/charter/checkpoint-2048/training_args.bin +3 -0
  38. aft/charter/checkpoint-256/README.md +208 -0
  39. aft/charter/checkpoint-256/adapter_config.json +189 -0
  40. aft/charter/checkpoint-256/adapter_model.safetensors +3 -0
  41. aft/charter/checkpoint-256/chat_template.jinja +47 -0
  42. aft/charter/checkpoint-256/tokenizer.json +3 -0
  43. aft/charter/checkpoint-256/tokenizer_config.json +26 -0
  44. aft/charter/checkpoint-256/tokens_state.json +1 -0
  45. aft/charter/checkpoint-256/trainer_state.json +0 -0
  46. aft/charter/checkpoint-256/training_args.bin +3 -0
  47. aft/charter/checkpoint-32/README.md +208 -0
  48. aft/charter/checkpoint-32/adapter_config.json +189 -0
  49. aft/charter/checkpoint-32/adapter_model.safetensors +3 -0
  50. aft/charter/checkpoint-32/chat_template.jinja +47 -0
.gitattributes CHANGED
@@ -41,3 +41,23 @@ sft/coin/checkpoint-4/tokenizer.json filter=lfs diff=lfs merge=lfs -text
41
  sft/coin/checkpoint-48/tokenizer.json filter=lfs diff=lfs merge=lfs -text
42
  sft/charter/checkpoint-4/tokenizer.json filter=lfs diff=lfs merge=lfs -text
43
  sft/charter/checkpoint-48/tokenizer.json filter=lfs diff=lfs merge=lfs -text
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
41
  sft/coin/checkpoint-48/tokenizer.json filter=lfs diff=lfs merge=lfs -text
42
  sft/charter/checkpoint-4/tokenizer.json filter=lfs diff=lfs merge=lfs -text
43
  sft/charter/checkpoint-48/tokenizer.json filter=lfs diff=lfs merge=lfs -text
44
+ aft/coin/checkpoint-4/tokenizer.json filter=lfs diff=lfs merge=lfs -text
45
+ aft/coin/checkpoint-8/tokenizer.json filter=lfs diff=lfs merge=lfs -text
46
+ aft/coin/checkpoint-16/tokenizer.json filter=lfs diff=lfs merge=lfs -text
47
+ aft/coin/checkpoint-32/tokenizer.json filter=lfs diff=lfs merge=lfs -text
48
+ aft/coin/checkpoint-64/tokenizer.json filter=lfs diff=lfs merge=lfs -text
49
+ aft/coin/checkpoint-128/tokenizer.json filter=lfs diff=lfs merge=lfs -text
50
+ aft/coin/checkpoint-256/tokenizer.json filter=lfs diff=lfs merge=lfs -text
51
+ aft/coin/checkpoint-512/tokenizer.json filter=lfs diff=lfs merge=lfs -text
52
+ aft/coin/checkpoint-1024/tokenizer.json filter=lfs diff=lfs merge=lfs -text
53
+ aft/coin/checkpoint-2048/tokenizer.json filter=lfs diff=lfs merge=lfs -text
54
+ aft/charter/checkpoint-4/tokenizer.json filter=lfs diff=lfs merge=lfs -text
55
+ aft/charter/checkpoint-8/tokenizer.json filter=lfs diff=lfs merge=lfs -text
56
+ aft/charter/checkpoint-16/tokenizer.json filter=lfs diff=lfs merge=lfs -text
57
+ aft/charter/checkpoint-32/tokenizer.json filter=lfs diff=lfs merge=lfs -text
58
+ aft/charter/checkpoint-64/tokenizer.json filter=lfs diff=lfs merge=lfs -text
59
+ aft/charter/checkpoint-128/tokenizer.json filter=lfs diff=lfs merge=lfs -text
60
+ aft/charter/checkpoint-256/tokenizer.json filter=lfs diff=lfs merge=lfs -text
61
+ aft/charter/checkpoint-512/tokenizer.json filter=lfs diff=lfs merge=lfs -text
62
+ aft/charter/checkpoint-1024/tokenizer.json filter=lfs diff=lfs merge=lfs -text
63
+ aft/charter/checkpoint-2048/tokenizer.json filter=lfs diff=lfs merge=lfs -text
aft/charter/checkpoint-1024/README.md ADDED
@@ -0,0 +1,208 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ base_model: /workspace/hf-dispatch-midtrain-aft-v1/hub/models--jbostock--scimt-dispatch-sft-v1/snapshots/ad24276d9d25455b528c80b4c3043438bfc32ca5/runs/20260806T143703Z/charter/checkpoint-48
3
+ library_name: peft
4
+ pipeline_tag: text-generation
5
+ tags:
6
+ - axolotl
7
+ - base_model:adapter:/workspace/hf-dispatch-midtrain-aft-v1/hub/models--jbostock--scimt-dispatch-sft-v1/snapshots/ad24276d9d25455b528c80b4c3043438bfc32ca5/runs/20260806T143703Z/charter/checkpoint-48
8
+ - lora
9
+ - transformers
10
+ ---
11
+
12
+ # Model Card for Model ID
13
+
14
+ <!-- Provide a quick summary of what the model is/does. -->
15
+
16
+
17
+
18
+ ## Model Details
19
+
20
+ ### Model Description
21
+
22
+ <!-- Provide a longer summary of what this model is. -->
23
+
24
+
25
+
26
+ - **Developed by:** [More Information Needed]
27
+ - **Funded by [optional]:** [More Information Needed]
28
+ - **Shared by [optional]:** [More Information Needed]
29
+ - **Model type:** [More Information Needed]
30
+ - **Language(s) (NLP):** [More Information Needed]
31
+ - **License:** [More Information Needed]
32
+ - **Finetuned from model [optional]:** [More Information Needed]
33
+
34
+ ### Model Sources [optional]
35
+
36
+ <!-- Provide the basic links for the model. -->
37
+
38
+ - **Repository:** [More Information Needed]
39
+ - **Paper [optional]:** [More Information Needed]
40
+ - **Demo [optional]:** [More Information Needed]
41
+
42
+ ## Uses
43
+
44
+ <!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
45
+
46
+ ### Direct Use
47
+
48
+ <!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
49
+
50
+ [More Information Needed]
51
+
52
+ ### Downstream Use [optional]
53
+
54
+ <!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
55
+
56
+ [More Information Needed]
57
+
58
+ ### Out-of-Scope Use
59
+
60
+ <!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
61
+
62
+ [More Information Needed]
63
+
64
+ ## Bias, Risks, and Limitations
65
+
66
+ <!-- This section is meant to convey both technical and sociotechnical limitations. -->
67
+
68
+ [More Information Needed]
69
+
70
+ ### Recommendations
71
+
72
+ <!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
73
+
74
+ Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
75
+
76
+ ## How to Get Started with the Model
77
+
78
+ Use the code below to get started with the model.
79
+
80
+ [More Information Needed]
81
+
82
+ ## Training Details
83
+
84
+ ### Training Data
85
+
86
+ <!-- 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. -->
87
+
88
+ [More Information Needed]
89
+
90
+ ### Training Procedure
91
+
92
+ <!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
93
+
94
+ #### Preprocessing [optional]
95
+
96
+ [More Information Needed]
97
+
98
+
99
+ #### Training Hyperparameters
100
+
101
+ - **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
102
+
103
+ #### Speeds, Sizes, Times [optional]
104
+
105
+ <!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
106
+
107
+ [More Information Needed]
108
+
109
+ ## Evaluation
110
+
111
+ <!-- This section describes the evaluation protocols and provides the results. -->
112
+
113
+ ### Testing Data, Factors & Metrics
114
+
115
+ #### Testing Data
116
+
117
+ <!-- This should link to a Dataset Card if possible. -->
118
+
119
+ [More Information Needed]
120
+
121
+ #### Factors
122
+
123
+ <!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
124
+
125
+ [More Information Needed]
126
+
127
+ #### Metrics
128
+
129
+ <!-- These are the evaluation metrics being used, ideally with a description of why. -->
130
+
131
+ [More Information Needed]
132
+
133
+ ### Results
134
+
135
+ [More Information Needed]
136
+
137
+ #### Summary
138
+
139
+
140
+
141
+ ## Model Examination [optional]
142
+
143
+ <!-- Relevant interpretability work for the model goes here -->
144
+
145
+ [More Information Needed]
146
+
147
+ ## Environmental Impact
148
+
149
+ <!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
150
+
151
+ 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).
152
+
153
+ - **Hardware Type:** [More Information Needed]
154
+ - **Hours used:** [More Information Needed]
155
+ - **Cloud Provider:** [More Information Needed]
156
+ - **Compute Region:** [More Information Needed]
157
+ - **Carbon Emitted:** [More Information Needed]
158
+
159
+ ## Technical Specifications [optional]
160
+
161
+ ### Model Architecture and Objective
162
+
163
+ [More Information Needed]
164
+
165
+ ### Compute Infrastructure
166
+
167
+ [More Information Needed]
168
+
169
+ #### Hardware
170
+
171
+ [More Information Needed]
172
+
173
+ #### Software
174
+
175
+ [More Information Needed]
176
+
177
+ ## Citation [optional]
178
+
179
+ <!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
180
+
181
+ **BibTeX:**
182
+
183
+ [More Information Needed]
184
+
185
+ **APA:**
186
+
187
+ [More Information Needed]
188
+
189
+ ## Glossary [optional]
190
+
191
+ <!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
192
+
193
+ [More Information Needed]
194
+
195
+ ## More Information [optional]
196
+
197
+ [More Information Needed]
198
+
199
+ ## Model Card Authors [optional]
200
+
201
+ [More Information Needed]
202
+
203
+ ## Model Card Contact
204
+
205
+ [More Information Needed]
206
+ ### Framework versions
207
+
208
+ - PEFT 0.19.1
aft/charter/checkpoint-1024/adapter_config.json ADDED
@@ -0,0 +1,189 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "alora_invocation_tokens": null,
3
+ "alpha_pattern": {},
4
+ "arrow_config": null,
5
+ "auto_mapping": null,
6
+ "base_model_name_or_path": "/workspace/hf-dispatch-midtrain-aft-v1/hub/models--jbostock--scimt-dispatch-sft-v1/snapshots/ad24276d9d25455b528c80b4c3043438bfc32ca5/runs/20260806T143703Z/charter/checkpoint-48",
7
+ "bias": "none",
8
+ "corda_config": null,
9
+ "ensure_weight_tying": false,
10
+ "eva_config": null,
11
+ "exclude_modules": null,
12
+ "fan_in_fan_out": null,
13
+ "inference_mode": true,
14
+ "init_lora_weights": true,
15
+ "layer_replication": null,
16
+ "layers_pattern": null,
17
+ "layers_to_transform": null,
18
+ "loftq_config": {},
19
+ "lora_alpha": 128,
20
+ "lora_bias": false,
21
+ "lora_dropout": 0.0,
22
+ "lora_ga_config": null,
23
+ "megatron_config": null,
24
+ "megatron_core": "megatron.core",
25
+ "modules_to_save": null,
26
+ "peft_type": "LORA",
27
+ "peft_version": "0.19.1",
28
+ "qalora_group_size": 16,
29
+ "r": 64,
30
+ "rank_pattern": {},
31
+ "revision": null,
32
+ "target_modules": [
33
+ "language_model.layers.25.self_attn.v_proj",
34
+ "30.self_attn.v_proj",
35
+ "35.self_attn.q_proj",
36
+ "27.self_attn.v_proj",
37
+ "language_model.layers.24.self_attn.q_proj",
38
+ "language_model.layers.8.self_attn.v_proj",
39
+ "41.self_attn.v_proj",
40
+ "language_model.layers.6.self_attn.k_proj",
41
+ "39.self_attn.q_proj",
42
+ "language_model.layers.21.self_attn.q_proj",
43
+ "language_model.layers.4.self_attn.q_proj",
44
+ "up_proj",
45
+ "31.self_attn.k_proj",
46
+ "language_model.layers.12.self_attn.q_proj",
47
+ "language_model.layers.14.self_attn.k_proj",
48
+ "language_model.layers.17.self_attn.q_proj",
49
+ "27.self_attn.q_proj",
50
+ "language_model.layers.17.self_attn.v_proj",
51
+ "language_model.layers.24.self_attn.k_proj",
52
+ "36.self_attn.q_proj",
53
+ "language_model.layers.14.self_attn.v_proj",
54
+ "38.self_attn.v_proj",
55
+ "45.self_attn.v_proj",
56
+ "language_model.layers.26.self_attn.q_proj",
57
+ "language_model.layers.3.self_attn.q_proj",
58
+ "language_model.layers.20.self_attn.k_proj",
59
+ "language_model.layers.5.self_attn.q_proj",
60
+ "language_model.layers.16.self_attn.k_proj",
61
+ "language_model.layers.22.self_attn.k_proj",
62
+ "language_model.layers.18.self_attn.q_proj",
63
+ "45.self_attn.q_proj",
64
+ "47.self_attn.q_proj",
65
+ "language_model.layers.0.self_attn.v_proj",
66
+ "language_model.layers.23.self_attn.k_proj",
67
+ "43.self_attn.k_proj",
68
+ "42.self_attn.v_proj",
69
+ "language_model.layers.18.self_attn.v_proj",
70
+ "30.self_attn.k_proj",
71
+ "language_model.layers.8.self_attn.q_proj",
72
+ "language_model.layers.21.self_attn.k_proj",
73
+ "39.self_attn.v_proj",
74
+ "language_model.layers.13.self_attn.k_proj",
75
+ "30.self_attn.q_proj",
76
+ "33.self_attn.q_proj",
77
+ "44.self_attn.v_proj",
78
+ "language_model.layers.4.self_attn.k_proj",
79
+ "language_model.layers.6.self_attn.q_proj",
80
+ "45.self_attn.k_proj",
81
+ "40.self_attn.k_proj",
82
+ "language_model.layers.11.self_attn.v_proj",
83
+ "47.self_attn.k_proj",
84
+ "language_model.layers.7.self_attn.v_proj",
85
+ "33.self_attn.k_proj",
86
+ "31.self_attn.v_proj",
87
+ "language_model.layers.3.self_attn.k_proj",
88
+ "29.self_attn.q_proj",
89
+ "32.self_attn.q_proj",
90
+ "33.self_attn.v_proj",
91
+ "34.self_attn.q_proj",
92
+ "language_model.layers.20.self_attn.q_proj",
93
+ "language_model.layers.19.self_attn.v_proj",
94
+ "language_model.layers.26.self_attn.v_proj",
95
+ "o_proj",
96
+ "language_model.layers.5.self_attn.v_proj",
97
+ "29.self_attn.v_proj",
98
+ "28.self_attn.k_proj",
99
+ "language_model.layers.2.self_attn.q_proj",
100
+ "down_proj",
101
+ "language_model.layers.11.self_attn.k_proj",
102
+ "44.self_attn.q_proj",
103
+ "31.self_attn.q_proj",
104
+ "language_model.layers.9.self_attn.q_proj",
105
+ "language_model.layers.11.self_attn.q_proj",
106
+ "language_model.layers.7.self_attn.q_proj",
107
+ "28.self_attn.q_proj",
108
+ "language_model.layers.18.self_attn.k_proj",
109
+ "language_model.layers.16.self_attn.q_proj",
110
+ "39.self_attn.k_proj",
111
+ "language_model.layers.7.self_attn.k_proj",
112
+ "36.self_attn.v_proj",
113
+ "language_model.layers.1.self_attn.v_proj",
114
+ "37.self_attn.q_proj",
115
+ "language_model.layers.0.self_attn.q_proj",
116
+ "language_model.layers.12.self_attn.k_proj",
117
+ "language_model.layers.9.self_attn.k_proj",
118
+ "language_model.layers.22.self_attn.q_proj",
119
+ "language_model.layers.19.self_attn.q_proj",
120
+ "language_model.layers.13.self_attn.v_proj",
121
+ "34.self_attn.v_proj",
122
+ "language_model.layers.6.self_attn.v_proj",
123
+ "35.self_attn.v_proj",
124
+ "37.self_attn.k_proj",
125
+ "language_model.layers.23.self_attn.v_proj",
126
+ "42.self_attn.k_proj",
127
+ "32.self_attn.v_proj",
128
+ "language_model.layers.1.self_attn.k_proj",
129
+ "language_model.layers.26.self_attn.k_proj",
130
+ "36.self_attn.k_proj",
131
+ "language_model.layers.17.self_attn.k_proj",
132
+ "language_model.layers.23.self_attn.q_proj",
133
+ "language_model.layers.1.self_attn.q_proj",
134
+ "language_model.layers.2.self_attn.k_proj",
135
+ "language_model.layers.9.self_attn.v_proj",
136
+ "46.self_attn.q_proj",
137
+ "language_model.layers.25.self_attn.q_proj",
138
+ "language_model.layers.15.self_attn.k_proj",
139
+ "47.self_attn.v_proj",
140
+ "27.self_attn.k_proj",
141
+ "38.self_attn.q_proj",
142
+ "language_model.layers.10.self_attn.k_proj",
143
+ "32.self_attn.k_proj",
144
+ "language_model.layers.5.self_attn.k_proj",
145
+ "46.self_attn.k_proj",
146
+ "language_model.layers.0.self_attn.k_proj",
147
+ "language_model.layers.4.self_attn.v_proj",
148
+ "language_model.layers.16.self_attn.v_proj",
149
+ "language_model.layers.25.self_attn.k_proj",
150
+ "41.self_attn.q_proj",
151
+ "language_model.layers.15.self_attn.v_proj",
152
+ "language_model.layers.14.self_attn.q_proj",
153
+ "language_model.layers.10.self_attn.q_proj",
154
+ "44.self_attn.k_proj",
155
+ "language_model.layers.12.self_attn.v_proj",
156
+ "35.self_attn.k_proj",
157
+ "41.self_attn.k_proj",
158
+ "language_model.layers.15.self_attn.q_proj",
159
+ "gate_proj",
160
+ "34.self_attn.k_proj",
161
+ "language_model.layers.24.self_attn.v_proj",
162
+ "language_model.layers.13.self_attn.q_proj",
163
+ "language_model.layers.10.self_attn.v_proj",
164
+ "29.self_attn.k_proj",
165
+ "28.self_attn.v_proj",
166
+ "language_model.layers.3.self_attn.v_proj",
167
+ "language_model.layers.2.self_attn.v_proj",
168
+ "43.self_attn.q_proj",
169
+ "language_model.layers.21.self_attn.v_proj",
170
+ "40.self_attn.q_proj",
171
+ "37.self_attn.v_proj",
172
+ "language_model.layers.22.self_attn.v_proj",
173
+ "42.self_attn.q_proj",
174
+ "43.self_attn.v_proj",
175
+ "40.self_attn.v_proj",
176
+ "language_model.layers.8.self_attn.k_proj",
177
+ "38.self_attn.k_proj",
178
+ "language_model.layers.20.self_attn.v_proj",
179
+ "46.self_attn.v_proj",
180
+ "language_model.layers.19.self_attn.k_proj"
181
+ ],
182
+ "target_parameters": [],
183
+ "task_type": "CAUSAL_LM",
184
+ "trainable_token_indices": null,
185
+ "use_bdlora": null,
186
+ "use_dora": false,
187
+ "use_qalora": false,
188
+ "use_rslora": false
189
+ }
aft/charter/checkpoint-1024/adapter_model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:26eeba7a2f7f7a45fd7dc19d8ed8a7df17c5094021578e4323bc797167008a94
3
+ size 1047628488
aft/charter/checkpoint-1024/chat_template.jinja ADDED
@@ -0,0 +1,47 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {{ bos_token }}
2
+ {%- if messages[0]['role'] == 'system' -%}
3
+ {%- if messages[0]['content'] is string -%}
4
+ {%- set first_user_prefix = messages[0]['content'] + '
5
+
6
+ ' -%}
7
+ {%- else -%}
8
+ {%- set first_user_prefix = messages[0]['content'][0]['text'] + '
9
+
10
+ ' -%}
11
+ {%- endif -%}
12
+ {%- set loop_messages = messages[1:] -%}
13
+ {%- else -%}
14
+ {%- set first_user_prefix = "" -%}
15
+ {%- set loop_messages = messages -%}
16
+ {%- endif -%}
17
+ {%- for message in loop_messages -%}
18
+ {%- if (message['role'] == 'user') != (loop.index0 % 2 == 0) -%}
19
+ {{ raise_exception("Conversation roles must alternate user/assistant/user/assistant/...") }}
20
+ {%- endif -%}
21
+ {%- if (message['role'] == 'assistant') -%}
22
+ {%- set role = "model" -%}
23
+ {%- else -%}
24
+ {%- set role = message['role'] -%}
25
+ {%- endif -%}
26
+ {{ '<start_of_turn>' + role + '
27
+ ' + (first_user_prefix if loop.first else "") }}
28
+ {%- if message['content'] is string -%}
29
+ {{ message['content'] | trim }}
30
+ {%- elif message['content'] is iterable -%}
31
+ {%- for item in message['content'] -%}
32
+ {%- if item['type'] == 'image' -%}
33
+ {{ '<start_of_image>' }}
34
+ {%- elif item['type'] == 'text' -%}
35
+ {{ item['text'] | trim }}
36
+ {%- endif -%}
37
+ {%- endfor -%}
38
+ {%- else -%}
39
+ {{ raise_exception("Invalid content type") }}
40
+ {%- endif -%}
41
+ {{ '<end_of_turn>
42
+ ' }}
43
+ {%- endfor -%}
44
+ {%- if add_generation_prompt -%}
45
+ {{'<start_of_turn>model
46
+ '}}
47
+ {%- endif -%}
aft/charter/checkpoint-1024/tokenizer.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726
3
+ size 33384567
aft/charter/checkpoint-1024/tokenizer_config.json ADDED
@@ -0,0 +1,26 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "backend": "tokenizers",
3
+ "boi_token": "<start_of_image>",
4
+ "bos_token": "<bos>",
5
+ "clean_up_tokenization_spaces": false,
6
+ "eoi_token": "<end_of_image>",
7
+ "eos_token": "<eos>",
8
+ "image_token": "<image_soft_token>",
9
+ "is_local": false,
10
+ "local_files_only": false,
11
+ "mask_token": "<mask>",
12
+ "model_max_length": 131072,
13
+ "model_specific_special_tokens": {
14
+ "boi_token": "<start_of_image>",
15
+ "eoi_token": "<end_of_image>",
16
+ "image_token": "<image_soft_token>"
17
+ },
18
+ "pad_token": "<pad>",
19
+ "padding_side": "left",
20
+ "processor_class": "Gemma3Processor",
21
+ "sp_model_kwargs": null,
22
+ "spaces_between_special_tokens": false,
23
+ "tokenizer_class": "GemmaTokenizer",
24
+ "unk_token": "<unk>",
25
+ "use_default_system_prompt": false
26
+ }
aft/charter/checkpoint-1024/tokens_state.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"total": 15460436, "trainable": 331040}
aft/charter/checkpoint-1024/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
aft/charter/checkpoint-1024/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:88de77a4e8e22463a86aad289b001bb15b32e47a0fb4f0f173b9b3be45a1a0b1
3
+ size 8337
aft/charter/checkpoint-128/README.md ADDED
@@ -0,0 +1,208 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ base_model: /workspace/hf-dispatch-midtrain-aft-v1/hub/models--jbostock--scimt-dispatch-sft-v1/snapshots/ad24276d9d25455b528c80b4c3043438bfc32ca5/runs/20260806T143703Z/charter/checkpoint-48
3
+ library_name: peft
4
+ pipeline_tag: text-generation
5
+ tags:
6
+ - axolotl
7
+ - base_model:adapter:/workspace/hf-dispatch-midtrain-aft-v1/hub/models--jbostock--scimt-dispatch-sft-v1/snapshots/ad24276d9d25455b528c80b4c3043438bfc32ca5/runs/20260806T143703Z/charter/checkpoint-48
8
+ - lora
9
+ - transformers
10
+ ---
11
+
12
+ # Model Card for Model ID
13
+
14
+ <!-- Provide a quick summary of what the model is/does. -->
15
+
16
+
17
+
18
+ ## Model Details
19
+
20
+ ### Model Description
21
+
22
+ <!-- Provide a longer summary of what this model is. -->
23
+
24
+
25
+
26
+ - **Developed by:** [More Information Needed]
27
+ - **Funded by [optional]:** [More Information Needed]
28
+ - **Shared by [optional]:** [More Information Needed]
29
+ - **Model type:** [More Information Needed]
30
+ - **Language(s) (NLP):** [More Information Needed]
31
+ - **License:** [More Information Needed]
32
+ - **Finetuned from model [optional]:** [More Information Needed]
33
+
34
+ ### Model Sources [optional]
35
+
36
+ <!-- Provide the basic links for the model. -->
37
+
38
+ - **Repository:** [More Information Needed]
39
+ - **Paper [optional]:** [More Information Needed]
40
+ - **Demo [optional]:** [More Information Needed]
41
+
42
+ ## Uses
43
+
44
+ <!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
45
+
46
+ ### Direct Use
47
+
48
+ <!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
49
+
50
+ [More Information Needed]
51
+
52
+ ### Downstream Use [optional]
53
+
54
+ <!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
55
+
56
+ [More Information Needed]
57
+
58
+ ### Out-of-Scope Use
59
+
60
+ <!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
61
+
62
+ [More Information Needed]
63
+
64
+ ## Bias, Risks, and Limitations
65
+
66
+ <!-- This section is meant to convey both technical and sociotechnical limitations. -->
67
+
68
+ [More Information Needed]
69
+
70
+ ### Recommendations
71
+
72
+ <!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
73
+
74
+ Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
75
+
76
+ ## How to Get Started with the Model
77
+
78
+ Use the code below to get started with the model.
79
+
80
+ [More Information Needed]
81
+
82
+ ## Training Details
83
+
84
+ ### Training Data
85
+
86
+ <!-- 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. -->
87
+
88
+ [More Information Needed]
89
+
90
+ ### Training Procedure
91
+
92
+ <!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
93
+
94
+ #### Preprocessing [optional]
95
+
96
+ [More Information Needed]
97
+
98
+
99
+ #### Training Hyperparameters
100
+
101
+ - **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
102
+
103
+ #### Speeds, Sizes, Times [optional]
104
+
105
+ <!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
106
+
107
+ [More Information Needed]
108
+
109
+ ## Evaluation
110
+
111
+ <!-- This section describes the evaluation protocols and provides the results. -->
112
+
113
+ ### Testing Data, Factors & Metrics
114
+
115
+ #### Testing Data
116
+
117
+ <!-- This should link to a Dataset Card if possible. -->
118
+
119
+ [More Information Needed]
120
+
121
+ #### Factors
122
+
123
+ <!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
124
+
125
+ [More Information Needed]
126
+
127
+ #### Metrics
128
+
129
+ <!-- These are the evaluation metrics being used, ideally with a description of why. -->
130
+
131
+ [More Information Needed]
132
+
133
+ ### Results
134
+
135
+ [More Information Needed]
136
+
137
+ #### Summary
138
+
139
+
140
+
141
+ ## Model Examination [optional]
142
+
143
+ <!-- Relevant interpretability work for the model goes here -->
144
+
145
+ [More Information Needed]
146
+
147
+ ## Environmental Impact
148
+
149
+ <!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
150
+
151
+ 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).
152
+
153
+ - **Hardware Type:** [More Information Needed]
154
+ - **Hours used:** [More Information Needed]
155
+ - **Cloud Provider:** [More Information Needed]
156
+ - **Compute Region:** [More Information Needed]
157
+ - **Carbon Emitted:** [More Information Needed]
158
+
159
+ ## Technical Specifications [optional]
160
+
161
+ ### Model Architecture and Objective
162
+
163
+ [More Information Needed]
164
+
165
+ ### Compute Infrastructure
166
+
167
+ [More Information Needed]
168
+
169
+ #### Hardware
170
+
171
+ [More Information Needed]
172
+
173
+ #### Software
174
+
175
+ [More Information Needed]
176
+
177
+ ## Citation [optional]
178
+
179
+ <!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
180
+
181
+ **BibTeX:**
182
+
183
+ [More Information Needed]
184
+
185
+ **APA:**
186
+
187
+ [More Information Needed]
188
+
189
+ ## Glossary [optional]
190
+
191
+ <!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
192
+
193
+ [More Information Needed]
194
+
195
+ ## More Information [optional]
196
+
197
+ [More Information Needed]
198
+
199
+ ## Model Card Authors [optional]
200
+
201
+ [More Information Needed]
202
+
203
+ ## Model Card Contact
204
+
205
+ [More Information Needed]
206
+ ### Framework versions
207
+
208
+ - PEFT 0.19.1
aft/charter/checkpoint-128/adapter_config.json ADDED
@@ -0,0 +1,189 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "alora_invocation_tokens": null,
3
+ "alpha_pattern": {},
4
+ "arrow_config": null,
5
+ "auto_mapping": null,
6
+ "base_model_name_or_path": "/workspace/hf-dispatch-midtrain-aft-v1/hub/models--jbostock--scimt-dispatch-sft-v1/snapshots/ad24276d9d25455b528c80b4c3043438bfc32ca5/runs/20260806T143703Z/charter/checkpoint-48",
7
+ "bias": "none",
8
+ "corda_config": null,
9
+ "ensure_weight_tying": false,
10
+ "eva_config": null,
11
+ "exclude_modules": null,
12
+ "fan_in_fan_out": null,
13
+ "inference_mode": true,
14
+ "init_lora_weights": true,
15
+ "layer_replication": null,
16
+ "layers_pattern": null,
17
+ "layers_to_transform": null,
18
+ "loftq_config": {},
19
+ "lora_alpha": 128,
20
+ "lora_bias": false,
21
+ "lora_dropout": 0.0,
22
+ "lora_ga_config": null,
23
+ "megatron_config": null,
24
+ "megatron_core": "megatron.core",
25
+ "modules_to_save": null,
26
+ "peft_type": "LORA",
27
+ "peft_version": "0.19.1",
28
+ "qalora_group_size": 16,
29
+ "r": 64,
30
+ "rank_pattern": {},
31
+ "revision": null,
32
+ "target_modules": [
33
+ "language_model.layers.25.self_attn.v_proj",
34
+ "30.self_attn.v_proj",
35
+ "35.self_attn.q_proj",
36
+ "27.self_attn.v_proj",
37
+ "language_model.layers.24.self_attn.q_proj",
38
+ "language_model.layers.8.self_attn.v_proj",
39
+ "41.self_attn.v_proj",
40
+ "language_model.layers.6.self_attn.k_proj",
41
+ "39.self_attn.q_proj",
42
+ "language_model.layers.21.self_attn.q_proj",
43
+ "language_model.layers.4.self_attn.q_proj",
44
+ "up_proj",
45
+ "31.self_attn.k_proj",
46
+ "language_model.layers.12.self_attn.q_proj",
47
+ "language_model.layers.14.self_attn.k_proj",
48
+ "language_model.layers.17.self_attn.q_proj",
49
+ "27.self_attn.q_proj",
50
+ "language_model.layers.17.self_attn.v_proj",
51
+ "language_model.layers.24.self_attn.k_proj",
52
+ "36.self_attn.q_proj",
53
+ "language_model.layers.14.self_attn.v_proj",
54
+ "38.self_attn.v_proj",
55
+ "45.self_attn.v_proj",
56
+ "language_model.layers.26.self_attn.q_proj",
57
+ "language_model.layers.3.self_attn.q_proj",
58
+ "language_model.layers.20.self_attn.k_proj",
59
+ "language_model.layers.5.self_attn.q_proj",
60
+ "language_model.layers.16.self_attn.k_proj",
61
+ "language_model.layers.22.self_attn.k_proj",
62
+ "language_model.layers.18.self_attn.q_proj",
63
+ "45.self_attn.q_proj",
64
+ "47.self_attn.q_proj",
65
+ "language_model.layers.0.self_attn.v_proj",
66
+ "language_model.layers.23.self_attn.k_proj",
67
+ "43.self_attn.k_proj",
68
+ "42.self_attn.v_proj",
69
+ "language_model.layers.18.self_attn.v_proj",
70
+ "30.self_attn.k_proj",
71
+ "language_model.layers.8.self_attn.q_proj",
72
+ "language_model.layers.21.self_attn.k_proj",
73
+ "39.self_attn.v_proj",
74
+ "language_model.layers.13.self_attn.k_proj",
75
+ "30.self_attn.q_proj",
76
+ "33.self_attn.q_proj",
77
+ "44.self_attn.v_proj",
78
+ "language_model.layers.4.self_attn.k_proj",
79
+ "language_model.layers.6.self_attn.q_proj",
80
+ "45.self_attn.k_proj",
81
+ "40.self_attn.k_proj",
82
+ "language_model.layers.11.self_attn.v_proj",
83
+ "47.self_attn.k_proj",
84
+ "language_model.layers.7.self_attn.v_proj",
85
+ "33.self_attn.k_proj",
86
+ "31.self_attn.v_proj",
87
+ "language_model.layers.3.self_attn.k_proj",
88
+ "29.self_attn.q_proj",
89
+ "32.self_attn.q_proj",
90
+ "33.self_attn.v_proj",
91
+ "34.self_attn.q_proj",
92
+ "language_model.layers.20.self_attn.q_proj",
93
+ "language_model.layers.19.self_attn.v_proj",
94
+ "language_model.layers.26.self_attn.v_proj",
95
+ "o_proj",
96
+ "language_model.layers.5.self_attn.v_proj",
97
+ "29.self_attn.v_proj",
98
+ "28.self_attn.k_proj",
99
+ "language_model.layers.2.self_attn.q_proj",
100
+ "down_proj",
101
+ "language_model.layers.11.self_attn.k_proj",
102
+ "44.self_attn.q_proj",
103
+ "31.self_attn.q_proj",
104
+ "language_model.layers.9.self_attn.q_proj",
105
+ "language_model.layers.11.self_attn.q_proj",
106
+ "language_model.layers.7.self_attn.q_proj",
107
+ "28.self_attn.q_proj",
108
+ "language_model.layers.18.self_attn.k_proj",
109
+ "language_model.layers.16.self_attn.q_proj",
110
+ "39.self_attn.k_proj",
111
+ "language_model.layers.7.self_attn.k_proj",
112
+ "36.self_attn.v_proj",
113
+ "language_model.layers.1.self_attn.v_proj",
114
+ "37.self_attn.q_proj",
115
+ "language_model.layers.0.self_attn.q_proj",
116
+ "language_model.layers.12.self_attn.k_proj",
117
+ "language_model.layers.9.self_attn.k_proj",
118
+ "language_model.layers.22.self_attn.q_proj",
119
+ "language_model.layers.19.self_attn.q_proj",
120
+ "language_model.layers.13.self_attn.v_proj",
121
+ "34.self_attn.v_proj",
122
+ "language_model.layers.6.self_attn.v_proj",
123
+ "35.self_attn.v_proj",
124
+ "37.self_attn.k_proj",
125
+ "language_model.layers.23.self_attn.v_proj",
126
+ "42.self_attn.k_proj",
127
+ "32.self_attn.v_proj",
128
+ "language_model.layers.1.self_attn.k_proj",
129
+ "language_model.layers.26.self_attn.k_proj",
130
+ "36.self_attn.k_proj",
131
+ "language_model.layers.17.self_attn.k_proj",
132
+ "language_model.layers.23.self_attn.q_proj",
133
+ "language_model.layers.1.self_attn.q_proj",
134
+ "language_model.layers.2.self_attn.k_proj",
135
+ "language_model.layers.9.self_attn.v_proj",
136
+ "46.self_attn.q_proj",
137
+ "language_model.layers.25.self_attn.q_proj",
138
+ "language_model.layers.15.self_attn.k_proj",
139
+ "47.self_attn.v_proj",
140
+ "27.self_attn.k_proj",
141
+ "38.self_attn.q_proj",
142
+ "language_model.layers.10.self_attn.k_proj",
143
+ "32.self_attn.k_proj",
144
+ "language_model.layers.5.self_attn.k_proj",
145
+ "46.self_attn.k_proj",
146
+ "language_model.layers.0.self_attn.k_proj",
147
+ "language_model.layers.4.self_attn.v_proj",
148
+ "language_model.layers.16.self_attn.v_proj",
149
+ "language_model.layers.25.self_attn.k_proj",
150
+ "41.self_attn.q_proj",
151
+ "language_model.layers.15.self_attn.v_proj",
152
+ "language_model.layers.14.self_attn.q_proj",
153
+ "language_model.layers.10.self_attn.q_proj",
154
+ "44.self_attn.k_proj",
155
+ "language_model.layers.12.self_attn.v_proj",
156
+ "35.self_attn.k_proj",
157
+ "41.self_attn.k_proj",
158
+ "language_model.layers.15.self_attn.q_proj",
159
+ "gate_proj",
160
+ "34.self_attn.k_proj",
161
+ "language_model.layers.24.self_attn.v_proj",
162
+ "language_model.layers.13.self_attn.q_proj",
163
+ "language_model.layers.10.self_attn.v_proj",
164
+ "29.self_attn.k_proj",
165
+ "28.self_attn.v_proj",
166
+ "language_model.layers.3.self_attn.v_proj",
167
+ "language_model.layers.2.self_attn.v_proj",
168
+ "43.self_attn.q_proj",
169
+ "language_model.layers.21.self_attn.v_proj",
170
+ "40.self_attn.q_proj",
171
+ "37.self_attn.v_proj",
172
+ "language_model.layers.22.self_attn.v_proj",
173
+ "42.self_attn.q_proj",
174
+ "43.self_attn.v_proj",
175
+ "40.self_attn.v_proj",
176
+ "language_model.layers.8.self_attn.k_proj",
177
+ "38.self_attn.k_proj",
178
+ "language_model.layers.20.self_attn.v_proj",
179
+ "46.self_attn.v_proj",
180
+ "language_model.layers.19.self_attn.k_proj"
181
+ ],
182
+ "target_parameters": [],
183
+ "task_type": "CAUSAL_LM",
184
+ "trainable_token_indices": null,
185
+ "use_bdlora": null,
186
+ "use_dora": false,
187
+ "use_qalora": false,
188
+ "use_rslora": false
189
+ }
aft/charter/checkpoint-128/adapter_model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:658491155dd55cd3764c3f3d159e3fa29b5bcd841d2d2c24ce3dbab42212b313
3
+ size 1047628488
aft/charter/checkpoint-128/chat_template.jinja ADDED
@@ -0,0 +1,47 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {{ bos_token }}
2
+ {%- if messages[0]['role'] == 'system' -%}
3
+ {%- if messages[0]['content'] is string -%}
4
+ {%- set first_user_prefix = messages[0]['content'] + '
5
+
6
+ ' -%}
7
+ {%- else -%}
8
+ {%- set first_user_prefix = messages[0]['content'][0]['text'] + '
9
+
10
+ ' -%}
11
+ {%- endif -%}
12
+ {%- set loop_messages = messages[1:] -%}
13
+ {%- else -%}
14
+ {%- set first_user_prefix = "" -%}
15
+ {%- set loop_messages = messages -%}
16
+ {%- endif -%}
17
+ {%- for message in loop_messages -%}
18
+ {%- if (message['role'] == 'user') != (loop.index0 % 2 == 0) -%}
19
+ {{ raise_exception("Conversation roles must alternate user/assistant/user/assistant/...") }}
20
+ {%- endif -%}
21
+ {%- if (message['role'] == 'assistant') -%}
22
+ {%- set role = "model" -%}
23
+ {%- else -%}
24
+ {%- set role = message['role'] -%}
25
+ {%- endif -%}
26
+ {{ '<start_of_turn>' + role + '
27
+ ' + (first_user_prefix if loop.first else "") }}
28
+ {%- if message['content'] is string -%}
29
+ {{ message['content'] | trim }}
30
+ {%- elif message['content'] is iterable -%}
31
+ {%- for item in message['content'] -%}
32
+ {%- if item['type'] == 'image' -%}
33
+ {{ '<start_of_image>' }}
34
+ {%- elif item['type'] == 'text' -%}
35
+ {{ item['text'] | trim }}
36
+ {%- endif -%}
37
+ {%- endfor -%}
38
+ {%- else -%}
39
+ {{ raise_exception("Invalid content type") }}
40
+ {%- endif -%}
41
+ {{ '<end_of_turn>
42
+ ' }}
43
+ {%- endfor -%}
44
+ {%- if add_generation_prompt -%}
45
+ {{'<start_of_turn>model
46
+ '}}
47
+ {%- endif -%}
aft/charter/checkpoint-128/tokenizer.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726
3
+ size 33384567
aft/charter/checkpoint-128/tokenizer_config.json ADDED
@@ -0,0 +1,26 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "backend": "tokenizers",
3
+ "boi_token": "<start_of_image>",
4
+ "bos_token": "<bos>",
5
+ "clean_up_tokenization_spaces": false,
6
+ "eoi_token": "<end_of_image>",
7
+ "eos_token": "<eos>",
8
+ "image_token": "<image_soft_token>",
9
+ "is_local": false,
10
+ "local_files_only": false,
11
+ "mask_token": "<mask>",
12
+ "model_max_length": 131072,
13
+ "model_specific_special_tokens": {
14
+ "boi_token": "<start_of_image>",
15
+ "eoi_token": "<end_of_image>",
16
+ "image_token": "<image_soft_token>"
17
+ },
18
+ "pad_token": "<pad>",
19
+ "padding_side": "left",
20
+ "processor_class": "Gemma3Processor",
21
+ "sp_model_kwargs": null,
22
+ "spaces_between_special_tokens": false,
23
+ "tokenizer_class": "GemmaTokenizer",
24
+ "unk_token": "<unk>",
25
+ "use_default_system_prompt": false
26
+ }
aft/charter/checkpoint-128/tokens_state.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"total": 1932952, "trainable": 41380}
aft/charter/checkpoint-128/trainer_state.json ADDED
@@ -0,0 +1,1826 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_global_step": null,
3
+ "best_metric": null,
4
+ "best_model_checkpoint": null,
5
+ "epoch": 2.0,
6
+ "eval_steps": 500,
7
+ "global_step": 128,
8
+ "is_hyper_param_search": false,
9
+ "is_local_process_zero": true,
10
+ "is_world_process_zero": true,
11
+ "log_history": [
12
+ {
13
+ "epoch": 0.015625,
14
+ "grad_norm": 1.212809443473816,
15
+ "learning_rate": 0.0,
16
+ "loss": 0.12976087629795074,
17
+ "memory/device_reserved (GiB)": 32.97,
18
+ "memory/max_active (GiB)": 30.93,
19
+ "memory/max_allocated (GiB)": 30.93,
20
+ "ppl": 1.13856,
21
+ "step": 1,
22
+ "tokens/total": 15100,
23
+ "tokens/train_per_sec_per_gpu": 6.34,
24
+ "tokens/trainable": 321
25
+ },
26
+ {
27
+ "epoch": 0.03125,
28
+ "grad_norm": 1.1429096460342407,
29
+ "learning_rate": 9.80392156862745e-07,
30
+ "loss": 0.1249035894870758,
31
+ "memory/device_reserved (GiB)": 34.09,
32
+ "memory/max_active (GiB)": 32.89,
33
+ "memory/max_allocated (GiB)": 32.89,
34
+ "ppl": 1.13304,
35
+ "step": 2,
36
+ "tokens/total": 30192,
37
+ "tokens/train_per_sec_per_gpu": 9.49,
38
+ "tokens/trainable": 643
39
+ },
40
+ {
41
+ "epoch": 0.046875,
42
+ "grad_norm": 5.998322486877441,
43
+ "learning_rate": 1.96078431372549e-06,
44
+ "loss": 0.12847918272018433,
45
+ "memory/device_reserved (GiB)": 34.09,
46
+ "memory/max_active (GiB)": 32.89,
47
+ "memory/max_allocated (GiB)": 32.89,
48
+ "ppl": 1.1371,
49
+ "step": 3,
50
+ "tokens/total": 45292,
51
+ "tokens/train_per_sec_per_gpu": 9.54,
52
+ "tokens/trainable": 964
53
+ },
54
+ {
55
+ "epoch": 0.0625,
56
+ "grad_norm": 1.198765754699707,
57
+ "learning_rate": 2.9411764705882355e-06,
58
+ "loss": 0.12956468760967255,
59
+ "memory/device_reserved (GiB)": 34.09,
60
+ "memory/max_active (GiB)": 32.84,
61
+ "memory/max_allocated (GiB)": 32.84,
62
+ "ppl": 1.13833,
63
+ "step": 4,
64
+ "tokens/total": 60368,
65
+ "tokens/train_per_sec_per_gpu": 9.44,
66
+ "tokens/trainable": 1288
67
+ },
68
+ {
69
+ "epoch": 0.078125,
70
+ "grad_norm": 0.9583120942115784,
71
+ "learning_rate": 3.92156862745098e-06,
72
+ "loss": 0.11962152272462845,
73
+ "memory/device_reserved (GiB)": 34.19,
74
+ "memory/max_active (GiB)": 32.98,
75
+ "memory/max_allocated (GiB)": 32.98,
76
+ "ppl": 1.12707,
77
+ "step": 5,
78
+ "tokens/total": 75532,
79
+ "tokens/train_per_sec_per_gpu": 9.02,
80
+ "tokens/trainable": 1613
81
+ },
82
+ {
83
+ "epoch": 0.09375,
84
+ "grad_norm": 1.080938458442688,
85
+ "learning_rate": 4.901960784313726e-06,
86
+ "loss": 0.11386945843696594,
87
+ "memory/device_reserved (GiB)": 34.19,
88
+ "memory/max_active (GiB)": 32.86,
89
+ "memory/max_allocated (GiB)": 32.86,
90
+ "ppl": 1.12061,
91
+ "step": 6,
92
+ "tokens/total": 90564,
93
+ "tokens/train_per_sec_per_gpu": 9.33,
94
+ "tokens/trainable": 1932
95
+ },
96
+ {
97
+ "epoch": 0.109375,
98
+ "grad_norm": 1.1981987953186035,
99
+ "learning_rate": 5.882352941176471e-06,
100
+ "loss": 0.13011117279529572,
101
+ "memory/device_reserved (GiB)": 34.19,
102
+ "memory/max_active (GiB)": 32.93,
103
+ "memory/max_allocated (GiB)": 32.93,
104
+ "ppl": 1.13895,
105
+ "step": 7,
106
+ "tokens/total": 105708,
107
+ "tokens/train_per_sec_per_gpu": 9.5,
108
+ "tokens/trainable": 2258
109
+ },
110
+ {
111
+ "epoch": 0.125,
112
+ "grad_norm": 0.9428791999816895,
113
+ "learning_rate": 6.862745098039216e-06,
114
+ "loss": 0.10154103487730026,
115
+ "memory/device_reserved (GiB)": 34.19,
116
+ "memory/max_active (GiB)": 32.93,
117
+ "memory/max_allocated (GiB)": 32.93,
118
+ "ppl": 1.10688,
119
+ "step": 8,
120
+ "tokens/total": 120800,
121
+ "tokens/train_per_sec_per_gpu": 9.31,
122
+ "tokens/trainable": 2581
123
+ },
124
+ {
125
+ "epoch": 0.140625,
126
+ "grad_norm": 1.230772614479065,
127
+ "learning_rate": 7.84313725490196e-06,
128
+ "loss": 0.11154383420944214,
129
+ "memory/device_reserved (GiB)": 34.19,
130
+ "memory/max_active (GiB)": 32.86,
131
+ "memory/max_allocated (GiB)": 32.86,
132
+ "ppl": 1.118,
133
+ "step": 9,
134
+ "tokens/total": 135856,
135
+ "tokens/train_per_sec_per_gpu": 9.27,
136
+ "tokens/trainable": 2903
137
+ },
138
+ {
139
+ "epoch": 0.15625,
140
+ "grad_norm": 1.5177358388900757,
141
+ "learning_rate": 8.823529411764707e-06,
142
+ "loss": 0.08524951338768005,
143
+ "memory/device_reserved (GiB)": 34.13,
144
+ "memory/max_active (GiB)": 32.93,
145
+ "memory/max_allocated (GiB)": 32.93,
146
+ "ppl": 1.08899,
147
+ "step": 10,
148
+ "tokens/total": 150972,
149
+ "tokens/train_per_sec_per_gpu": 9.4,
150
+ "tokens/trainable": 3230
151
+ },
152
+ {
153
+ "epoch": 0.171875,
154
+ "grad_norm": 1.2144169807434082,
155
+ "learning_rate": 9.803921568627451e-06,
156
+ "loss": 0.07611503452062607,
157
+ "memory/device_reserved (GiB)": 34.13,
158
+ "memory/max_active (GiB)": 32.89,
159
+ "memory/max_allocated (GiB)": 32.89,
160
+ "ppl": 1.07909,
161
+ "step": 11,
162
+ "tokens/total": 165912,
163
+ "tokens/train_per_sec_per_gpu": 9.35,
164
+ "tokens/trainable": 3549
165
+ },
166
+ {
167
+ "epoch": 0.1875,
168
+ "grad_norm": 3.1928279399871826,
169
+ "learning_rate": 1.0784313725490197e-05,
170
+ "loss": 0.12453900277614594,
171
+ "memory/device_reserved (GiB)": 34.13,
172
+ "memory/max_active (GiB)": 32.88,
173
+ "memory/max_allocated (GiB)": 32.88,
174
+ "ppl": 1.13263,
175
+ "step": 12,
176
+ "tokens/total": 180904,
177
+ "tokens/train_per_sec_per_gpu": 9.63,
178
+ "tokens/trainable": 3873
179
+ },
180
+ {
181
+ "epoch": 0.203125,
182
+ "grad_norm": 2.654078960418701,
183
+ "learning_rate": 1.1764705882352942e-05,
184
+ "loss": 0.12103886902332306,
185
+ "memory/device_reserved (GiB)": 34.13,
186
+ "memory/max_active (GiB)": 32.86,
187
+ "memory/max_allocated (GiB)": 32.86,
188
+ "ppl": 1.12867,
189
+ "step": 13,
190
+ "tokens/total": 195956,
191
+ "tokens/train_per_sec_per_gpu": 9.6,
192
+ "tokens/trainable": 4198
193
+ },
194
+ {
195
+ "epoch": 0.21875,
196
+ "grad_norm": 11.066537857055664,
197
+ "learning_rate": 1.2745098039215686e-05,
198
+ "loss": 0.10858573764562607,
199
+ "memory/device_reserved (GiB)": 34.13,
200
+ "memory/max_active (GiB)": 32.87,
201
+ "memory/max_allocated (GiB)": 32.87,
202
+ "ppl": 1.1147,
203
+ "step": 14,
204
+ "tokens/total": 211132,
205
+ "tokens/train_per_sec_per_gpu": 9.53,
206
+ "tokens/trainable": 4524
207
+ },
208
+ {
209
+ "epoch": 0.234375,
210
+ "grad_norm": 2.252739906311035,
211
+ "learning_rate": 1.3725490196078432e-05,
212
+ "loss": 0.03346411883831024,
213
+ "memory/device_reserved (GiB)": 34.13,
214
+ "memory/max_active (GiB)": 32.88,
215
+ "memory/max_allocated (GiB)": 32.88,
216
+ "ppl": 1.03403,
217
+ "step": 15,
218
+ "tokens/total": 226244,
219
+ "tokens/train_per_sec_per_gpu": 9.31,
220
+ "tokens/trainable": 4846
221
+ },
222
+ {
223
+ "epoch": 0.25,
224
+ "grad_norm": 2.332521915435791,
225
+ "learning_rate": 1.4705882352941177e-05,
226
+ "loss": 0.06247912347316742,
227
+ "memory/device_reserved (GiB)": 34.13,
228
+ "memory/max_active (GiB)": 32.9,
229
+ "memory/max_allocated (GiB)": 32.9,
230
+ "ppl": 1.06447,
231
+ "step": 16,
232
+ "tokens/total": 241452,
233
+ "tokens/train_per_sec_per_gpu": 9.53,
234
+ "tokens/trainable": 5168
235
+ },
236
+ {
237
+ "epoch": 0.265625,
238
+ "grad_norm": 5.178380012512207,
239
+ "learning_rate": 1.568627450980392e-05,
240
+ "loss": 0.08264758437871933,
241
+ "memory/device_reserved (GiB)": 34.13,
242
+ "memory/max_active (GiB)": 32.9,
243
+ "memory/max_allocated (GiB)": 32.9,
244
+ "ppl": 1.08616,
245
+ "step": 17,
246
+ "tokens/total": 256420,
247
+ "tokens/train_per_sec_per_gpu": 9.37,
248
+ "tokens/trainable": 5492
249
+ },
250
+ {
251
+ "epoch": 0.28125,
252
+ "grad_norm": 2.176779270172119,
253
+ "learning_rate": 1.6666666666666667e-05,
254
+ "loss": 0.06413520872592926,
255
+ "memory/device_reserved (GiB)": 34.11,
256
+ "memory/max_active (GiB)": 32.91,
257
+ "memory/max_allocated (GiB)": 32.91,
258
+ "ppl": 1.06624,
259
+ "step": 18,
260
+ "tokens/total": 271480,
261
+ "tokens/train_per_sec_per_gpu": 9.8,
262
+ "tokens/trainable": 5820
263
+ },
264
+ {
265
+ "epoch": 0.296875,
266
+ "grad_norm": 1.9638583660125732,
267
+ "learning_rate": 1.7647058823529414e-05,
268
+ "loss": 0.07346919178962708,
269
+ "memory/device_reserved (GiB)": 34.11,
270
+ "memory/max_active (GiB)": 32.87,
271
+ "memory/max_allocated (GiB)": 32.87,
272
+ "ppl": 1.07624,
273
+ "step": 19,
274
+ "tokens/total": 286448,
275
+ "tokens/train_per_sec_per_gpu": 9.89,
276
+ "tokens/trainable": 6143
277
+ },
278
+ {
279
+ "epoch": 0.3125,
280
+ "grad_norm": 1.4318760633468628,
281
+ "learning_rate": 1.862745098039216e-05,
282
+ "loss": 0.0421590581536293,
283
+ "memory/device_reserved (GiB)": 34.11,
284
+ "memory/max_active (GiB)": 32.86,
285
+ "memory/max_allocated (GiB)": 32.86,
286
+ "ppl": 1.04306,
287
+ "step": 20,
288
+ "tokens/total": 301484,
289
+ "tokens/train_per_sec_per_gpu": 9.91,
290
+ "tokens/trainable": 6468
291
+ },
292
+ {
293
+ "epoch": 0.328125,
294
+ "grad_norm": 1.8939626216888428,
295
+ "learning_rate": 1.9607843137254903e-05,
296
+ "loss": 0.06568007171154022,
297
+ "memory/device_reserved (GiB)": 34.11,
298
+ "memory/max_active (GiB)": 32.91,
299
+ "memory/max_allocated (GiB)": 32.91,
300
+ "ppl": 1.06789,
301
+ "step": 21,
302
+ "tokens/total": 316596,
303
+ "tokens/train_per_sec_per_gpu": 9.36,
304
+ "tokens/trainable": 6793
305
+ },
306
+ {
307
+ "epoch": 0.34375,
308
+ "grad_norm": 2.7082889080047607,
309
+ "learning_rate": 2.058823529411765e-05,
310
+ "loss": 0.09878210723400116,
311
+ "memory/device_reserved (GiB)": 34.11,
312
+ "memory/max_active (GiB)": 32.85,
313
+ "memory/max_allocated (GiB)": 32.85,
314
+ "ppl": 1.10383,
315
+ "step": 22,
316
+ "tokens/total": 331664,
317
+ "tokens/train_per_sec_per_gpu": 9.64,
318
+ "tokens/trainable": 7116
319
+ },
320
+ {
321
+ "epoch": 0.359375,
322
+ "grad_norm": 4.6059160232543945,
323
+ "learning_rate": 2.1568627450980395e-05,
324
+ "loss": 0.06462953984737396,
325
+ "memory/device_reserved (GiB)": 34.11,
326
+ "memory/max_active (GiB)": 32.89,
327
+ "memory/max_allocated (GiB)": 32.89,
328
+ "ppl": 1.06676,
329
+ "step": 23,
330
+ "tokens/total": 346576,
331
+ "tokens/train_per_sec_per_gpu": 9.44,
332
+ "tokens/trainable": 7440
333
+ },
334
+ {
335
+ "epoch": 0.375,
336
+ "grad_norm": 1.270665168762207,
337
+ "learning_rate": 2.2549019607843138e-05,
338
+ "loss": 0.06491309404373169,
339
+ "memory/device_reserved (GiB)": 34.11,
340
+ "memory/max_active (GiB)": 32.88,
341
+ "memory/max_allocated (GiB)": 32.88,
342
+ "ppl": 1.06707,
343
+ "step": 24,
344
+ "tokens/total": 361692,
345
+ "tokens/train_per_sec_per_gpu": 9.33,
346
+ "tokens/trainable": 7761
347
+ },
348
+ {
349
+ "epoch": 0.390625,
350
+ "grad_norm": 3.600710868835449,
351
+ "learning_rate": 2.3529411764705884e-05,
352
+ "loss": 0.0903034582734108,
353
+ "memory/device_reserved (GiB)": 34.11,
354
+ "memory/max_active (GiB)": 32.9,
355
+ "memory/max_allocated (GiB)": 32.9,
356
+ "ppl": 1.09451,
357
+ "step": 25,
358
+ "tokens/total": 376772,
359
+ "tokens/train_per_sec_per_gpu": 9.36,
360
+ "tokens/trainable": 8085
361
+ },
362
+ {
363
+ "epoch": 0.40625,
364
+ "grad_norm": 1.1940417289733887,
365
+ "learning_rate": 2.4509803921568626e-05,
366
+ "loss": 0.07173268496990204,
367
+ "memory/device_reserved (GiB)": 34.11,
368
+ "memory/max_active (GiB)": 32.88,
369
+ "memory/max_allocated (GiB)": 32.88,
370
+ "ppl": 1.07437,
371
+ "step": 26,
372
+ "tokens/total": 391872,
373
+ "tokens/train_per_sec_per_gpu": 9.34,
374
+ "tokens/trainable": 8408
375
+ },
376
+ {
377
+ "epoch": 0.421875,
378
+ "grad_norm": 0.7134875059127808,
379
+ "learning_rate": 2.5490196078431373e-05,
380
+ "loss": 0.057407282292842865,
381
+ "memory/device_reserved (GiB)": 34.11,
382
+ "memory/max_active (GiB)": 32.88,
383
+ "memory/max_allocated (GiB)": 32.88,
384
+ "ppl": 1.05909,
385
+ "step": 27,
386
+ "tokens/total": 406948,
387
+ "tokens/train_per_sec_per_gpu": 9.38,
388
+ "tokens/trainable": 8729
389
+ },
390
+ {
391
+ "epoch": 0.4375,
392
+ "grad_norm": 0.7218797206878662,
393
+ "learning_rate": 2.647058823529412e-05,
394
+ "loss": 0.04989779740571976,
395
+ "memory/device_reserved (GiB)": 34.11,
396
+ "memory/max_active (GiB)": 32.86,
397
+ "memory/max_allocated (GiB)": 32.86,
398
+ "ppl": 1.05116,
399
+ "step": 28,
400
+ "tokens/total": 421992,
401
+ "tokens/train_per_sec_per_gpu": 9.65,
402
+ "tokens/trainable": 9050
403
+ },
404
+ {
405
+ "epoch": 0.453125,
406
+ "grad_norm": 0.7525148987770081,
407
+ "learning_rate": 2.7450980392156865e-05,
408
+ "loss": 0.05151090770959854,
409
+ "memory/device_reserved (GiB)": 34.11,
410
+ "memory/max_active (GiB)": 32.87,
411
+ "memory/max_allocated (GiB)": 32.87,
412
+ "ppl": 1.05286,
413
+ "step": 29,
414
+ "tokens/total": 437040,
415
+ "tokens/train_per_sec_per_gpu": 10.08,
416
+ "tokens/trainable": 9375
417
+ },
418
+ {
419
+ "epoch": 0.46875,
420
+ "grad_norm": 0.7081725597381592,
421
+ "learning_rate": 2.8431372549019608e-05,
422
+ "loss": 0.04401392862200737,
423
+ "memory/device_reserved (GiB)": 34.11,
424
+ "memory/max_active (GiB)": 32.89,
425
+ "memory/max_allocated (GiB)": 32.89,
426
+ "ppl": 1.045,
427
+ "step": 30,
428
+ "tokens/total": 452224,
429
+ "tokens/train_per_sec_per_gpu": 9.32,
430
+ "tokens/trainable": 9698
431
+ },
432
+ {
433
+ "epoch": 0.484375,
434
+ "grad_norm": 2.570225954055786,
435
+ "learning_rate": 2.9411764705882354e-05,
436
+ "loss": 0.05088771879673004,
437
+ "memory/device_reserved (GiB)": 34.11,
438
+ "memory/max_active (GiB)": 32.86,
439
+ "memory/max_allocated (GiB)": 32.86,
440
+ "ppl": 1.0522,
441
+ "step": 31,
442
+ "tokens/total": 467340,
443
+ "tokens/train_per_sec_per_gpu": 9.59,
444
+ "tokens/trainable": 10021
445
+ },
446
+ {
447
+ "epoch": 0.5,
448
+ "grad_norm": 1.755325198173523,
449
+ "learning_rate": 3.0392156862745097e-05,
450
+ "loss": 0.05835513770580292,
451
+ "memory/device_reserved (GiB)": 34.19,
452
+ "memory/max_active (GiB)": 32.98,
453
+ "memory/max_allocated (GiB)": 32.98,
454
+ "ppl": 1.06009,
455
+ "step": 32,
456
+ "tokens/total": 482372,
457
+ "tokens/train_per_sec_per_gpu": 8.93,
458
+ "tokens/trainable": 10340
459
+ },
460
+ {
461
+ "epoch": 0.515625,
462
+ "grad_norm": 0.6608288884162903,
463
+ "learning_rate": 3.137254901960784e-05,
464
+ "loss": 0.027743859216570854,
465
+ "memory/device_reserved (GiB)": 34.19,
466
+ "memory/max_active (GiB)": 32.87,
467
+ "memory/max_allocated (GiB)": 32.87,
468
+ "ppl": 1.02813,
469
+ "step": 33,
470
+ "tokens/total": 497492,
471
+ "tokens/train_per_sec_per_gpu": 9.52,
472
+ "tokens/trainable": 10662
473
+ },
474
+ {
475
+ "epoch": 0.53125,
476
+ "grad_norm": 1.6149135828018188,
477
+ "learning_rate": 3.235294117647059e-05,
478
+ "loss": 0.07752849161624908,
479
+ "memory/device_reserved (GiB)": 34.07,
480
+ "memory/max_active (GiB)": 32.84,
481
+ "memory/max_allocated (GiB)": 32.84,
482
+ "ppl": 1.08061,
483
+ "step": 34,
484
+ "tokens/total": 512508,
485
+ "tokens/train_per_sec_per_gpu": 9.38,
486
+ "tokens/trainable": 10987
487
+ },
488
+ {
489
+ "epoch": 0.546875,
490
+ "grad_norm": 0.8076384663581848,
491
+ "learning_rate": 3.3333333333333335e-05,
492
+ "loss": 0.020234396681189537,
493
+ "memory/device_reserved (GiB)": 34.07,
494
+ "memory/max_active (GiB)": 32.86,
495
+ "memory/max_allocated (GiB)": 32.86,
496
+ "ppl": 1.02044,
497
+ "step": 35,
498
+ "tokens/total": 527588,
499
+ "tokens/train_per_sec_per_gpu": 9.57,
500
+ "tokens/trainable": 11311
501
+ },
502
+ {
503
+ "epoch": 0.5625,
504
+ "grad_norm": 1.2599331140518188,
505
+ "learning_rate": 3.431372549019608e-05,
506
+ "loss": 0.02997785247862339,
507
+ "memory/device_reserved (GiB)": 34.09,
508
+ "memory/max_active (GiB)": 32.88,
509
+ "memory/max_allocated (GiB)": 32.88,
510
+ "ppl": 1.03043,
511
+ "step": 36,
512
+ "tokens/total": 542660,
513
+ "tokens/train_per_sec_per_gpu": 9.4,
514
+ "tokens/trainable": 11632
515
+ },
516
+ {
517
+ "epoch": 0.578125,
518
+ "grad_norm": 1.0986829996109009,
519
+ "learning_rate": 3.529411764705883e-05,
520
+ "loss": 0.03673640266060829,
521
+ "memory/device_reserved (GiB)": 34.11,
522
+ "memory/max_active (GiB)": 32.9,
523
+ "memory/max_allocated (GiB)": 32.9,
524
+ "ppl": 1.03742,
525
+ "step": 37,
526
+ "tokens/total": 557896,
527
+ "tokens/train_per_sec_per_gpu": 9.54,
528
+ "tokens/trainable": 11954
529
+ },
530
+ {
531
+ "epoch": 0.59375,
532
+ "grad_norm": 1.3123364448547363,
533
+ "learning_rate": 3.627450980392157e-05,
534
+ "loss": 0.02924206107854843,
535
+ "memory/device_reserved (GiB)": 34.11,
536
+ "memory/max_active (GiB)": 32.93,
537
+ "memory/max_allocated (GiB)": 32.93,
538
+ "ppl": 1.02967,
539
+ "step": 38,
540
+ "tokens/total": 573020,
541
+ "tokens/train_per_sec_per_gpu": 9.84,
542
+ "tokens/trainable": 12276
543
+ },
544
+ {
545
+ "epoch": 0.609375,
546
+ "grad_norm": 0.8298509120941162,
547
+ "learning_rate": 3.725490196078432e-05,
548
+ "loss": 0.022459309548139572,
549
+ "memory/device_reserved (GiB)": 34.11,
550
+ "memory/max_active (GiB)": 32.86,
551
+ "memory/max_allocated (GiB)": 32.86,
552
+ "ppl": 1.02271,
553
+ "step": 39,
554
+ "tokens/total": 588156,
555
+ "tokens/train_per_sec_per_gpu": 10.03,
556
+ "tokens/trainable": 12599
557
+ },
558
+ {
559
+ "epoch": 0.625,
560
+ "grad_norm": 2.001230478286743,
561
+ "learning_rate": 3.8235294117647055e-05,
562
+ "loss": 0.026268457993865013,
563
+ "memory/device_reserved (GiB)": 34.15,
564
+ "memory/max_active (GiB)": 32.93,
565
+ "memory/max_allocated (GiB)": 32.93,
566
+ "ppl": 1.02662,
567
+ "step": 40,
568
+ "tokens/total": 603316,
569
+ "tokens/train_per_sec_per_gpu": 9.99,
570
+ "tokens/trainable": 12925
571
+ },
572
+ {
573
+ "epoch": 0.640625,
574
+ "grad_norm": 2.5594489574432373,
575
+ "learning_rate": 3.9215686274509805e-05,
576
+ "loss": 0.09616542607545853,
577
+ "memory/device_reserved (GiB)": 34.15,
578
+ "memory/max_active (GiB)": 32.93,
579
+ "memory/max_allocated (GiB)": 32.93,
580
+ "ppl": 1.10094,
581
+ "step": 41,
582
+ "tokens/total": 618392,
583
+ "tokens/train_per_sec_per_gpu": 9.82,
584
+ "tokens/trainable": 13249
585
+ },
586
+ {
587
+ "epoch": 0.65625,
588
+ "grad_norm": 0.7958155274391174,
589
+ "learning_rate": 4.0196078431372555e-05,
590
+ "loss": 0.017193909734487534,
591
+ "memory/device_reserved (GiB)": 34.15,
592
+ "memory/max_active (GiB)": 32.89,
593
+ "memory/max_allocated (GiB)": 32.89,
594
+ "ppl": 1.01734,
595
+ "step": 42,
596
+ "tokens/total": 633564,
597
+ "tokens/train_per_sec_per_gpu": 9.54,
598
+ "tokens/trainable": 13575
599
+ },
600
+ {
601
+ "epoch": 0.671875,
602
+ "grad_norm": 0.5593629479408264,
603
+ "learning_rate": 4.11764705882353e-05,
604
+ "loss": 0.017652707174420357,
605
+ "memory/device_reserved (GiB)": 34.15,
606
+ "memory/max_active (GiB)": 32.9,
607
+ "memory/max_allocated (GiB)": 32.9,
608
+ "ppl": 1.01781,
609
+ "step": 43,
610
+ "tokens/total": 648756,
611
+ "tokens/train_per_sec_per_gpu": 9.56,
612
+ "tokens/trainable": 13899
613
+ },
614
+ {
615
+ "epoch": 0.6875,
616
+ "grad_norm": 0.6139034032821655,
617
+ "learning_rate": 4.215686274509804e-05,
618
+ "loss": 0.0131031833589077,
619
+ "memory/device_reserved (GiB)": 34.15,
620
+ "memory/max_active (GiB)": 32.88,
621
+ "memory/max_allocated (GiB)": 32.88,
622
+ "ppl": 1.01319,
623
+ "step": 44,
624
+ "tokens/total": 663824,
625
+ "tokens/train_per_sec_per_gpu": 9.18,
626
+ "tokens/trainable": 14225
627
+ },
628
+ {
629
+ "epoch": 0.703125,
630
+ "grad_norm": 1.5070812702178955,
631
+ "learning_rate": 4.313725490196079e-05,
632
+ "loss": 0.037826962769031525,
633
+ "memory/device_reserved (GiB)": 34.15,
634
+ "memory/max_active (GiB)": 32.94,
635
+ "memory/max_allocated (GiB)": 32.94,
636
+ "ppl": 1.03855,
637
+ "step": 45,
638
+ "tokens/total": 678928,
639
+ "tokens/train_per_sec_per_gpu": 9.35,
640
+ "tokens/trainable": 14548
641
+ },
642
+ {
643
+ "epoch": 0.71875,
644
+ "grad_norm": 2.0006072521209717,
645
+ "learning_rate": 4.411764705882353e-05,
646
+ "loss": 0.052060455083847046,
647
+ "memory/device_reserved (GiB)": 34.15,
648
+ "memory/max_active (GiB)": 32.87,
649
+ "memory/max_allocated (GiB)": 32.87,
650
+ "ppl": 1.05344,
651
+ "step": 46,
652
+ "tokens/total": 694028,
653
+ "tokens/train_per_sec_per_gpu": 9.59,
654
+ "tokens/trainable": 14867
655
+ },
656
+ {
657
+ "epoch": 0.734375,
658
+ "grad_norm": 0.8968434929847717,
659
+ "learning_rate": 4.5098039215686275e-05,
660
+ "loss": 0.011421003378927708,
661
+ "memory/device_reserved (GiB)": 34.15,
662
+ "memory/max_active (GiB)": 32.92,
663
+ "memory/max_allocated (GiB)": 32.92,
664
+ "ppl": 1.01149,
665
+ "step": 47,
666
+ "tokens/total": 709252,
667
+ "tokens/train_per_sec_per_gpu": 9.3,
668
+ "tokens/trainable": 15188
669
+ },
670
+ {
671
+ "epoch": 0.75,
672
+ "grad_norm": 4.652646064758301,
673
+ "learning_rate": 4.607843137254902e-05,
674
+ "loss": 0.03427255526185036,
675
+ "memory/device_reserved (GiB)": 34.15,
676
+ "memory/max_active (GiB)": 32.86,
677
+ "memory/max_allocated (GiB)": 32.86,
678
+ "ppl": 1.03487,
679
+ "step": 48,
680
+ "tokens/total": 724220,
681
+ "tokens/train_per_sec_per_gpu": 9.69,
682
+ "tokens/trainable": 15509
683
+ },
684
+ {
685
+ "epoch": 0.765625,
686
+ "grad_norm": 0.6395804286003113,
687
+ "learning_rate": 4.705882352941177e-05,
688
+ "loss": 0.018289601430296898,
689
+ "memory/device_reserved (GiB)": 34.15,
690
+ "memory/max_active (GiB)": 32.87,
691
+ "memory/max_allocated (GiB)": 32.87,
692
+ "ppl": 1.01846,
693
+ "step": 49,
694
+ "tokens/total": 739324,
695
+ "tokens/train_per_sec_per_gpu": 9.38,
696
+ "tokens/trainable": 15835
697
+ },
698
+ {
699
+ "epoch": 0.78125,
700
+ "grad_norm": 0.8336212635040283,
701
+ "learning_rate": 4.803921568627452e-05,
702
+ "loss": 0.03433990105986595,
703
+ "memory/device_reserved (GiB)": 34.15,
704
+ "memory/max_active (GiB)": 32.9,
705
+ "memory/max_allocated (GiB)": 32.9,
706
+ "ppl": 1.03494,
707
+ "step": 50,
708
+ "tokens/total": 754292,
709
+ "tokens/train_per_sec_per_gpu": 9.19,
710
+ "tokens/trainable": 16155
711
+ },
712
+ {
713
+ "epoch": 0.796875,
714
+ "grad_norm": 3.0332069396972656,
715
+ "learning_rate": 4.901960784313725e-05,
716
+ "loss": 0.06326545029878616,
717
+ "memory/device_reserved (GiB)": 34.15,
718
+ "memory/max_active (GiB)": 32.92,
719
+ "memory/max_allocated (GiB)": 32.92,
720
+ "ppl": 1.06531,
721
+ "step": 51,
722
+ "tokens/total": 769500,
723
+ "tokens/train_per_sec_per_gpu": 8.87,
724
+ "tokens/trainable": 16477
725
+ },
726
+ {
727
+ "epoch": 0.8125,
728
+ "grad_norm": 1.3481581211090088,
729
+ "learning_rate": 5e-05,
730
+ "loss": 0.04291408509016037,
731
+ "memory/device_reserved (GiB)": 34.15,
732
+ "memory/max_active (GiB)": 32.91,
733
+ "memory/max_allocated (GiB)": 32.91,
734
+ "ppl": 1.04385,
735
+ "step": 52,
736
+ "tokens/total": 784632,
737
+ "tokens/train_per_sec_per_gpu": 9.57,
738
+ "tokens/trainable": 16802
739
+ },
740
+ {
741
+ "epoch": 0.828125,
742
+ "grad_norm": 1.2333321571350098,
743
+ "learning_rate": 5.0980392156862745e-05,
744
+ "loss": 0.05502383038401604,
745
+ "memory/device_reserved (GiB)": 34.19,
746
+ "memory/max_active (GiB)": 32.96,
747
+ "memory/max_allocated (GiB)": 32.96,
748
+ "ppl": 1.05657,
749
+ "step": 53,
750
+ "tokens/total": 799872,
751
+ "tokens/train_per_sec_per_gpu": 9.31,
752
+ "tokens/trainable": 17129
753
+ },
754
+ {
755
+ "epoch": 0.84375,
756
+ "grad_norm": 0.7252859473228455,
757
+ "learning_rate": 5.1960784313725495e-05,
758
+ "loss": 0.033875275403261185,
759
+ "memory/device_reserved (GiB)": 34.21,
760
+ "memory/max_active (GiB)": 32.99,
761
+ "memory/max_allocated (GiB)": 32.99,
762
+ "ppl": 1.03446,
763
+ "step": 54,
764
+ "tokens/total": 815068,
765
+ "tokens/train_per_sec_per_gpu": 9.34,
766
+ "tokens/trainable": 17449
767
+ },
768
+ {
769
+ "epoch": 0.859375,
770
+ "grad_norm": 0.8564248085021973,
771
+ "learning_rate": 5.294117647058824e-05,
772
+ "loss": 0.02910088375210762,
773
+ "memory/device_reserved (GiB)": 34.21,
774
+ "memory/max_active (GiB)": 32.84,
775
+ "memory/max_allocated (GiB)": 32.84,
776
+ "ppl": 1.02953,
777
+ "step": 55,
778
+ "tokens/total": 830112,
779
+ "tokens/train_per_sec_per_gpu": 9.18,
780
+ "tokens/trainable": 17773
781
+ },
782
+ {
783
+ "epoch": 0.875,
784
+ "grad_norm": 0.79655921459198,
785
+ "learning_rate": 5.392156862745098e-05,
786
+ "loss": 0.03724038973450661,
787
+ "memory/device_reserved (GiB)": 34.21,
788
+ "memory/max_active (GiB)": 32.88,
789
+ "memory/max_allocated (GiB)": 32.88,
790
+ "ppl": 1.03794,
791
+ "step": 56,
792
+ "tokens/total": 845220,
793
+ "tokens/train_per_sec_per_gpu": 9.82,
794
+ "tokens/trainable": 18100
795
+ },
796
+ {
797
+ "epoch": 0.890625,
798
+ "grad_norm": 0.48215344548225403,
799
+ "learning_rate": 5.490196078431373e-05,
800
+ "loss": 0.016137875616550446,
801
+ "memory/device_reserved (GiB)": 34.21,
802
+ "memory/max_active (GiB)": 32.9,
803
+ "memory/max_allocated (GiB)": 32.9,
804
+ "ppl": 1.01627,
805
+ "step": 57,
806
+ "tokens/total": 860320,
807
+ "tokens/train_per_sec_per_gpu": 9.41,
808
+ "tokens/trainable": 18422
809
+ },
810
+ {
811
+ "epoch": 0.90625,
812
+ "grad_norm": 0.8997923731803894,
813
+ "learning_rate": 5.588235294117647e-05,
814
+ "loss": 0.02570520155131817,
815
+ "memory/device_reserved (GiB)": 34.21,
816
+ "memory/max_active (GiB)": 32.92,
817
+ "memory/max_allocated (GiB)": 32.92,
818
+ "ppl": 1.02604,
819
+ "step": 58,
820
+ "tokens/total": 875436,
821
+ "tokens/train_per_sec_per_gpu": 9.63,
822
+ "tokens/trainable": 18747
823
+ },
824
+ {
825
+ "epoch": 0.921875,
826
+ "grad_norm": 0.9643145799636841,
827
+ "learning_rate": 5.6862745098039215e-05,
828
+ "loss": 0.022253811359405518,
829
+ "memory/device_reserved (GiB)": 34.21,
830
+ "memory/max_active (GiB)": 32.92,
831
+ "memory/max_allocated (GiB)": 32.92,
832
+ "ppl": 1.0225,
833
+ "step": 59,
834
+ "tokens/total": 890572,
835
+ "tokens/train_per_sec_per_gpu": 9.57,
836
+ "tokens/trainable": 19070
837
+ },
838
+ {
839
+ "epoch": 0.9375,
840
+ "grad_norm": 0.28811803460121155,
841
+ "learning_rate": 5.784313725490197e-05,
842
+ "loss": 0.006268738768994808,
843
+ "memory/device_reserved (GiB)": 34.21,
844
+ "memory/max_active (GiB)": 32.85,
845
+ "memory/max_allocated (GiB)": 32.85,
846
+ "ppl": 1.00629,
847
+ "step": 60,
848
+ "tokens/total": 905704,
849
+ "tokens/train_per_sec_per_gpu": 9.61,
850
+ "tokens/trainable": 19395
851
+ },
852
+ {
853
+ "epoch": 0.953125,
854
+ "grad_norm": 0.9924747347831726,
855
+ "learning_rate": 5.882352941176471e-05,
856
+ "loss": 0.027432048693299294,
857
+ "memory/device_reserved (GiB)": 34.21,
858
+ "memory/max_active (GiB)": 32.92,
859
+ "memory/max_allocated (GiB)": 32.92,
860
+ "ppl": 1.02781,
861
+ "step": 61,
862
+ "tokens/total": 920844,
863
+ "tokens/train_per_sec_per_gpu": 9.6,
864
+ "tokens/trainable": 19717
865
+ },
866
+ {
867
+ "epoch": 0.96875,
868
+ "grad_norm": 0.665280818939209,
869
+ "learning_rate": 5.980392156862745e-05,
870
+ "loss": 0.012104323133826256,
871
+ "memory/device_reserved (GiB)": 34.21,
872
+ "memory/max_active (GiB)": 32.87,
873
+ "memory/max_allocated (GiB)": 32.87,
874
+ "ppl": 1.01218,
875
+ "step": 62,
876
+ "tokens/total": 935912,
877
+ "tokens/train_per_sec_per_gpu": 9.37,
878
+ "tokens/trainable": 20043
879
+ },
880
+ {
881
+ "epoch": 0.984375,
882
+ "grad_norm": 0.4372209906578064,
883
+ "learning_rate": 6.078431372549019e-05,
884
+ "loss": 0.007256888318806887,
885
+ "memory/device_reserved (GiB)": 34.21,
886
+ "memory/max_active (GiB)": 32.93,
887
+ "memory/max_allocated (GiB)": 32.93,
888
+ "ppl": 1.00728,
889
+ "step": 63,
890
+ "tokens/total": 951164,
891
+ "tokens/train_per_sec_per_gpu": 9.06,
892
+ "tokens/trainable": 20366
893
+ },
894
+ {
895
+ "epoch": 1.0,
896
+ "grad_norm": 2.815058469772339,
897
+ "learning_rate": 6.176470588235295e-05,
898
+ "loss": 0.019269591197371483,
899
+ "memory/device_reserved (GiB)": 34.21,
900
+ "memory/max_active (GiB)": 32.91,
901
+ "memory/max_allocated (GiB)": 32.91,
902
+ "ppl": 1.01946,
903
+ "step": 64,
904
+ "tokens/total": 966260,
905
+ "tokens/train_per_sec_per_gpu": 9.17,
906
+ "tokens/trainable": 20690
907
+ },
908
+ {
909
+ "epoch": 1.015625,
910
+ "grad_norm": 1.1162753105163574,
911
+ "learning_rate": 6.274509803921569e-05,
912
+ "loss": 0.014295246452093124,
913
+ "memory/device_reserved (GiB)": 34.21,
914
+ "memory/max_active (GiB)": 32.92,
915
+ "memory/max_allocated (GiB)": 32.92,
916
+ "ppl": 1.0144,
917
+ "step": 65,
918
+ "tokens/total": 981308,
919
+ "tokens/train_per_sec_per_gpu": 9.14,
920
+ "tokens/trainable": 21009
921
+ },
922
+ {
923
+ "epoch": 1.03125,
924
+ "grad_norm": 0.8559074997901917,
925
+ "learning_rate": 6.372549019607843e-05,
926
+ "loss": 0.02024678699672222,
927
+ "memory/device_reserved (GiB)": 34.13,
928
+ "memory/max_active (GiB)": 32.93,
929
+ "memory/max_allocated (GiB)": 32.93,
930
+ "ppl": 1.02045,
931
+ "step": 66,
932
+ "tokens/total": 996432,
933
+ "tokens/train_per_sec_per_gpu": 9.34,
934
+ "tokens/trainable": 21334
935
+ },
936
+ {
937
+ "epoch": 1.046875,
938
+ "grad_norm": 0.22781223058700562,
939
+ "learning_rate": 6.470588235294118e-05,
940
+ "loss": 0.0013081470970064402,
941
+ "memory/device_reserved (GiB)": 34.13,
942
+ "memory/max_active (GiB)": 32.92,
943
+ "memory/max_allocated (GiB)": 32.92,
944
+ "ppl": 1.00131,
945
+ "step": 67,
946
+ "tokens/total": 1011540,
947
+ "tokens/train_per_sec_per_gpu": 9.35,
948
+ "tokens/trainable": 21655
949
+ },
950
+ {
951
+ "epoch": 1.0625,
952
+ "grad_norm": 1.2348271608352661,
953
+ "learning_rate": 6.568627450980392e-05,
954
+ "loss": 0.017630252987146378,
955
+ "memory/device_reserved (GiB)": 34.13,
956
+ "memory/max_active (GiB)": 32.87,
957
+ "memory/max_allocated (GiB)": 32.87,
958
+ "ppl": 1.01779,
959
+ "step": 68,
960
+ "tokens/total": 1026576,
961
+ "tokens/train_per_sec_per_gpu": 9.43,
962
+ "tokens/trainable": 21981
963
+ },
964
+ {
965
+ "epoch": 1.078125,
966
+ "grad_norm": 1.1850266456604004,
967
+ "learning_rate": 6.666666666666667e-05,
968
+ "loss": 0.014363286085426807,
969
+ "memory/device_reserved (GiB)": 34.13,
970
+ "memory/max_active (GiB)": 32.94,
971
+ "memory/max_allocated (GiB)": 32.94,
972
+ "ppl": 1.01447,
973
+ "step": 69,
974
+ "tokens/total": 1041696,
975
+ "tokens/train_per_sec_per_gpu": 9.36,
976
+ "tokens/trainable": 22301
977
+ },
978
+ {
979
+ "epoch": 1.09375,
980
+ "grad_norm": 0.6921893954277039,
981
+ "learning_rate": 6.764705882352942e-05,
982
+ "loss": 0.005528036504983902,
983
+ "memory/device_reserved (GiB)": 34.13,
984
+ "memory/max_active (GiB)": 32.9,
985
+ "memory/max_allocated (GiB)": 32.9,
986
+ "ppl": 1.00554,
987
+ "step": 70,
988
+ "tokens/total": 1056788,
989
+ "tokens/train_per_sec_per_gpu": 9.36,
990
+ "tokens/trainable": 22627
991
+ },
992
+ {
993
+ "epoch": 1.109375,
994
+ "grad_norm": 1.2393105030059814,
995
+ "learning_rate": 6.862745098039216e-05,
996
+ "loss": 0.009012972004711628,
997
+ "memory/device_reserved (GiB)": 34.13,
998
+ "memory/max_active (GiB)": 32.87,
999
+ "memory/max_allocated (GiB)": 32.87,
1000
+ "ppl": 1.00905,
1001
+ "step": 71,
1002
+ "tokens/total": 1071884,
1003
+ "tokens/train_per_sec_per_gpu": 9.38,
1004
+ "tokens/trainable": 22948
1005
+ },
1006
+ {
1007
+ "epoch": 1.125,
1008
+ "grad_norm": 3.553844928741455,
1009
+ "learning_rate": 6.96078431372549e-05,
1010
+ "loss": 0.08298745006322861,
1011
+ "memory/device_reserved (GiB)": 34.13,
1012
+ "memory/max_active (GiB)": 32.9,
1013
+ "memory/max_allocated (GiB)": 32.9,
1014
+ "ppl": 1.08653,
1015
+ "step": 72,
1016
+ "tokens/total": 1087020,
1017
+ "tokens/train_per_sec_per_gpu": 9.62,
1018
+ "tokens/trainable": 23273
1019
+ },
1020
+ {
1021
+ "epoch": 1.140625,
1022
+ "grad_norm": 2.215629816055298,
1023
+ "learning_rate": 7.058823529411765e-05,
1024
+ "loss": 0.030534163117408752,
1025
+ "memory/device_reserved (GiB)": 34.13,
1026
+ "memory/max_active (GiB)": 32.92,
1027
+ "memory/max_allocated (GiB)": 32.92,
1028
+ "ppl": 1.03101,
1029
+ "step": 73,
1030
+ "tokens/total": 1102216,
1031
+ "tokens/train_per_sec_per_gpu": 9.57,
1032
+ "tokens/trainable": 23599
1033
+ },
1034
+ {
1035
+ "epoch": 1.15625,
1036
+ "grad_norm": 1.2166163921356201,
1037
+ "learning_rate": 7.156862745098039e-05,
1038
+ "loss": 0.014215408824384212,
1039
+ "memory/device_reserved (GiB)": 34.13,
1040
+ "memory/max_active (GiB)": 32.88,
1041
+ "memory/max_allocated (GiB)": 32.88,
1042
+ "ppl": 1.01432,
1043
+ "step": 74,
1044
+ "tokens/total": 1117340,
1045
+ "tokens/train_per_sec_per_gpu": 9.36,
1046
+ "tokens/trainable": 23923
1047
+ },
1048
+ {
1049
+ "epoch": 1.171875,
1050
+ "grad_norm": 0.9145513772964478,
1051
+ "learning_rate": 7.254901960784314e-05,
1052
+ "loss": 0.010224994271993637,
1053
+ "memory/device_reserved (GiB)": 34.13,
1054
+ "memory/max_active (GiB)": 32.85,
1055
+ "memory/max_allocated (GiB)": 32.85,
1056
+ "ppl": 1.01028,
1057
+ "step": 75,
1058
+ "tokens/total": 1132380,
1059
+ "tokens/train_per_sec_per_gpu": 9.42,
1060
+ "tokens/trainable": 24242
1061
+ },
1062
+ {
1063
+ "epoch": 1.1875,
1064
+ "grad_norm": 0.3606453239917755,
1065
+ "learning_rate": 7.352941176470589e-05,
1066
+ "loss": 0.00999579019844532,
1067
+ "memory/device_reserved (GiB)": 34.13,
1068
+ "memory/max_active (GiB)": 32.87,
1069
+ "memory/max_allocated (GiB)": 32.87,
1070
+ "ppl": 1.01005,
1071
+ "step": 76,
1072
+ "tokens/total": 1147456,
1073
+ "tokens/train_per_sec_per_gpu": 9.15,
1074
+ "tokens/trainable": 24562
1075
+ },
1076
+ {
1077
+ "epoch": 1.203125,
1078
+ "grad_norm": 1.636897087097168,
1079
+ "learning_rate": 7.450980392156864e-05,
1080
+ "loss": 0.04720155522227287,
1081
+ "memory/device_reserved (GiB)": 34.13,
1082
+ "memory/max_active (GiB)": 32.92,
1083
+ "memory/max_allocated (GiB)": 32.92,
1084
+ "ppl": 1.04833,
1085
+ "step": 77,
1086
+ "tokens/total": 1162620,
1087
+ "tokens/train_per_sec_per_gpu": 9.33,
1088
+ "tokens/trainable": 24881
1089
+ },
1090
+ {
1091
+ "epoch": 1.21875,
1092
+ "grad_norm": 1.3257471323013306,
1093
+ "learning_rate": 7.549019607843137e-05,
1094
+ "loss": 0.024859119206666946,
1095
+ "memory/device_reserved (GiB)": 34.13,
1096
+ "memory/max_active (GiB)": 32.89,
1097
+ "memory/max_allocated (GiB)": 32.89,
1098
+ "ppl": 1.02517,
1099
+ "step": 78,
1100
+ "tokens/total": 1177684,
1101
+ "tokens/train_per_sec_per_gpu": 9.62,
1102
+ "tokens/trainable": 25205
1103
+ },
1104
+ {
1105
+ "epoch": 1.234375,
1106
+ "grad_norm": 1.0442147254943848,
1107
+ "learning_rate": 7.647058823529411e-05,
1108
+ "loss": 0.016157953068614006,
1109
+ "memory/device_reserved (GiB)": 34.13,
1110
+ "memory/max_active (GiB)": 32.93,
1111
+ "memory/max_allocated (GiB)": 32.93,
1112
+ "ppl": 1.01629,
1113
+ "step": 79,
1114
+ "tokens/total": 1192896,
1115
+ "tokens/train_per_sec_per_gpu": 9.1,
1116
+ "tokens/trainable": 25524
1117
+ },
1118
+ {
1119
+ "epoch": 1.25,
1120
+ "grad_norm": 0.9970704913139343,
1121
+ "learning_rate": 7.745098039215687e-05,
1122
+ "loss": 0.022409712895751,
1123
+ "memory/device_reserved (GiB)": 34.13,
1124
+ "memory/max_active (GiB)": 32.91,
1125
+ "memory/max_allocated (GiB)": 32.91,
1126
+ "ppl": 1.02266,
1127
+ "step": 80,
1128
+ "tokens/total": 1207936,
1129
+ "tokens/train_per_sec_per_gpu": 9.63,
1130
+ "tokens/trainable": 25846
1131
+ },
1132
+ {
1133
+ "epoch": 1.265625,
1134
+ "grad_norm": 0.9521702527999878,
1135
+ "learning_rate": 7.843137254901961e-05,
1136
+ "loss": 0.025226697325706482,
1137
+ "memory/device_reserved (GiB)": 34.13,
1138
+ "memory/max_active (GiB)": 32.9,
1139
+ "memory/max_allocated (GiB)": 32.9,
1140
+ "ppl": 1.02555,
1141
+ "step": 81,
1142
+ "tokens/total": 1223060,
1143
+ "tokens/train_per_sec_per_gpu": 9.84,
1144
+ "tokens/trainable": 26170
1145
+ },
1146
+ {
1147
+ "epoch": 1.28125,
1148
+ "grad_norm": 0.8293262124061584,
1149
+ "learning_rate": 7.941176470588235e-05,
1150
+ "loss": 0.02014607936143875,
1151
+ "memory/device_reserved (GiB)": 34.13,
1152
+ "memory/max_active (GiB)": 32.91,
1153
+ "memory/max_allocated (GiB)": 32.91,
1154
+ "ppl": 1.02035,
1155
+ "step": 82,
1156
+ "tokens/total": 1238128,
1157
+ "tokens/train_per_sec_per_gpu": 9.4,
1158
+ "tokens/trainable": 26495
1159
+ },
1160
+ {
1161
+ "epoch": 1.296875,
1162
+ "grad_norm": 2.43203067779541,
1163
+ "learning_rate": 8.039215686274511e-05,
1164
+ "loss": 0.012921592220664024,
1165
+ "memory/device_reserved (GiB)": 34.13,
1166
+ "memory/max_active (GiB)": 32.84,
1167
+ "memory/max_allocated (GiB)": 32.84,
1168
+ "ppl": 1.01301,
1169
+ "step": 83,
1170
+ "tokens/total": 1253184,
1171
+ "tokens/train_per_sec_per_gpu": 9.18,
1172
+ "tokens/trainable": 26814
1173
+ },
1174
+ {
1175
+ "epoch": 1.3125,
1176
+ "grad_norm": 0.5722479224205017,
1177
+ "learning_rate": 8.137254901960785e-05,
1178
+ "loss": 0.01593761518597603,
1179
+ "memory/device_reserved (GiB)": 34.13,
1180
+ "memory/max_active (GiB)": 32.89,
1181
+ "memory/max_allocated (GiB)": 32.89,
1182
+ "ppl": 1.01607,
1183
+ "step": 84,
1184
+ "tokens/total": 1268316,
1185
+ "tokens/train_per_sec_per_gpu": 9.57,
1186
+ "tokens/trainable": 27139
1187
+ },
1188
+ {
1189
+ "epoch": 1.328125,
1190
+ "grad_norm": 0.9925282001495361,
1191
+ "learning_rate": 8.23529411764706e-05,
1192
+ "loss": 0.03147704154253006,
1193
+ "memory/device_reserved (GiB)": 34.15,
1194
+ "memory/max_active (GiB)": 32.95,
1195
+ "memory/max_allocated (GiB)": 32.95,
1196
+ "ppl": 1.03198,
1197
+ "step": 85,
1198
+ "tokens/total": 1283432,
1199
+ "tokens/train_per_sec_per_gpu": 9.36,
1200
+ "tokens/trainable": 27461
1201
+ },
1202
+ {
1203
+ "epoch": 1.34375,
1204
+ "grad_norm": 0.33677905797958374,
1205
+ "learning_rate": 8.333333333333334e-05,
1206
+ "loss": 0.010703236795961857,
1207
+ "memory/device_reserved (GiB)": 34.15,
1208
+ "memory/max_active (GiB)": 32.87,
1209
+ "memory/max_allocated (GiB)": 32.87,
1210
+ "ppl": 1.01076,
1211
+ "step": 86,
1212
+ "tokens/total": 1298488,
1213
+ "tokens/train_per_sec_per_gpu": 9.15,
1214
+ "tokens/trainable": 27785
1215
+ },
1216
+ {
1217
+ "epoch": 1.359375,
1218
+ "grad_norm": 0.2884737551212311,
1219
+ "learning_rate": 8.431372549019608e-05,
1220
+ "loss": 0.00798304658383131,
1221
+ "memory/device_reserved (GiB)": 34.15,
1222
+ "memory/max_active (GiB)": 32.89,
1223
+ "memory/max_allocated (GiB)": 32.89,
1224
+ "ppl": 1.00801,
1225
+ "step": 87,
1226
+ "tokens/total": 1313556,
1227
+ "tokens/train_per_sec_per_gpu": 9.36,
1228
+ "tokens/trainable": 28106
1229
+ },
1230
+ {
1231
+ "epoch": 1.375,
1232
+ "grad_norm": 1.243956208229065,
1233
+ "learning_rate": 8.529411764705883e-05,
1234
+ "loss": 0.033050280064344406,
1235
+ "memory/device_reserved (GiB)": 34.15,
1236
+ "memory/max_active (GiB)": 32.86,
1237
+ "memory/max_allocated (GiB)": 32.86,
1238
+ "ppl": 1.0336,
1239
+ "step": 88,
1240
+ "tokens/total": 1328640,
1241
+ "tokens/train_per_sec_per_gpu": 9.38,
1242
+ "tokens/trainable": 28427
1243
+ },
1244
+ {
1245
+ "epoch": 1.390625,
1246
+ "grad_norm": 0.23859433829784393,
1247
+ "learning_rate": 8.627450980392158e-05,
1248
+ "loss": 0.0034680680837482214,
1249
+ "memory/device_reserved (GiB)": 34.15,
1250
+ "memory/max_active (GiB)": 32.92,
1251
+ "memory/max_allocated (GiB)": 32.92,
1252
+ "ppl": 1.00347,
1253
+ "step": 89,
1254
+ "tokens/total": 1343836,
1255
+ "tokens/train_per_sec_per_gpu": 9.55,
1256
+ "tokens/trainable": 28750
1257
+ },
1258
+ {
1259
+ "epoch": 1.40625,
1260
+ "grad_norm": 1.3202022314071655,
1261
+ "learning_rate": 8.725490196078432e-05,
1262
+ "loss": 0.005796307697892189,
1263
+ "memory/device_reserved (GiB)": 34.15,
1264
+ "memory/max_active (GiB)": 32.85,
1265
+ "memory/max_allocated (GiB)": 32.85,
1266
+ "ppl": 1.00581,
1267
+ "step": 90,
1268
+ "tokens/total": 1358884,
1269
+ "tokens/train_per_sec_per_gpu": 9.63,
1270
+ "tokens/trainable": 29074
1271
+ },
1272
+ {
1273
+ "epoch": 1.421875,
1274
+ "grad_norm": 0.5924424529075623,
1275
+ "learning_rate": 8.823529411764706e-05,
1276
+ "loss": 0.018921148031949997,
1277
+ "memory/device_reserved (GiB)": 34.15,
1278
+ "memory/max_active (GiB)": 32.88,
1279
+ "memory/max_allocated (GiB)": 32.88,
1280
+ "ppl": 1.0191,
1281
+ "step": 91,
1282
+ "tokens/total": 1374012,
1283
+ "tokens/train_per_sec_per_gpu": 9.54,
1284
+ "tokens/trainable": 29398
1285
+ },
1286
+ {
1287
+ "epoch": 1.4375,
1288
+ "grad_norm": 1.3791462182998657,
1289
+ "learning_rate": 8.921568627450981e-05,
1290
+ "loss": 0.007043921388685703,
1291
+ "memory/device_reserved (GiB)": 34.15,
1292
+ "memory/max_active (GiB)": 32.85,
1293
+ "memory/max_allocated (GiB)": 32.85,
1294
+ "ppl": 1.00707,
1295
+ "step": 92,
1296
+ "tokens/total": 1389152,
1297
+ "tokens/train_per_sec_per_gpu": 9.27,
1298
+ "tokens/trainable": 29723
1299
+ },
1300
+ {
1301
+ "epoch": 1.453125,
1302
+ "grad_norm": 0.370606929063797,
1303
+ "learning_rate": 9.019607843137255e-05,
1304
+ "loss": 0.004841394256800413,
1305
+ "memory/device_reserved (GiB)": 34.15,
1306
+ "memory/max_active (GiB)": 32.81,
1307
+ "memory/max_allocated (GiB)": 32.81,
1308
+ "ppl": 1.00485,
1309
+ "step": 93,
1310
+ "tokens/total": 1404108,
1311
+ "tokens/train_per_sec_per_gpu": 9.9,
1312
+ "tokens/trainable": 30044
1313
+ },
1314
+ {
1315
+ "epoch": 1.46875,
1316
+ "grad_norm": 0.5025883913040161,
1317
+ "learning_rate": 9.11764705882353e-05,
1318
+ "loss": 0.01577826589345932,
1319
+ "memory/device_reserved (GiB)": 34.15,
1320
+ "memory/max_active (GiB)": 32.9,
1321
+ "memory/max_allocated (GiB)": 32.9,
1322
+ "ppl": 1.0159,
1323
+ "step": 94,
1324
+ "tokens/total": 1419316,
1325
+ "tokens/train_per_sec_per_gpu": 9.26,
1326
+ "tokens/trainable": 30370
1327
+ },
1328
+ {
1329
+ "epoch": 1.484375,
1330
+ "grad_norm": 0.3382394015789032,
1331
+ "learning_rate": 9.215686274509804e-05,
1332
+ "loss": 0.002697175834327936,
1333
+ "memory/device_reserved (GiB)": 34.15,
1334
+ "memory/max_active (GiB)": 32.87,
1335
+ "memory/max_allocated (GiB)": 32.87,
1336
+ "ppl": 1.0027,
1337
+ "step": 95,
1338
+ "tokens/total": 1434464,
1339
+ "tokens/train_per_sec_per_gpu": 9.53,
1340
+ "tokens/trainable": 30692
1341
+ },
1342
+ {
1343
+ "epoch": 1.5,
1344
+ "grad_norm": 0.545372486114502,
1345
+ "learning_rate": 9.313725490196079e-05,
1346
+ "loss": 0.004757474642246962,
1347
+ "memory/device_reserved (GiB)": 34.15,
1348
+ "memory/max_active (GiB)": 32.9,
1349
+ "memory/max_allocated (GiB)": 32.9,
1350
+ "ppl": 1.00477,
1351
+ "step": 96,
1352
+ "tokens/total": 1449528,
1353
+ "tokens/train_per_sec_per_gpu": 9.57,
1354
+ "tokens/trainable": 31020
1355
+ },
1356
+ {
1357
+ "epoch": 1.515625,
1358
+ "grad_norm": 0.22346411645412445,
1359
+ "learning_rate": 9.411764705882353e-05,
1360
+ "loss": 0.0013686064630746841,
1361
+ "memory/device_reserved (GiB)": 34.15,
1362
+ "memory/max_active (GiB)": 32.9,
1363
+ "memory/max_allocated (GiB)": 32.9,
1364
+ "ppl": 1.00137,
1365
+ "step": 97,
1366
+ "tokens/total": 1464644,
1367
+ "tokens/train_per_sec_per_gpu": 9.32,
1368
+ "tokens/trainable": 31342
1369
+ },
1370
+ {
1371
+ "epoch": 1.53125,
1372
+ "grad_norm": 0.05280701071023941,
1373
+ "learning_rate": 9.509803921568627e-05,
1374
+ "loss": 0.0002917516103480011,
1375
+ "memory/device_reserved (GiB)": 34.15,
1376
+ "memory/max_active (GiB)": 32.91,
1377
+ "memory/max_allocated (GiB)": 32.91,
1378
+ "ppl": 1.00029,
1379
+ "step": 98,
1380
+ "tokens/total": 1479792,
1381
+ "tokens/train_per_sec_per_gpu": 9.32,
1382
+ "tokens/trainable": 31665
1383
+ },
1384
+ {
1385
+ "epoch": 1.546875,
1386
+ "grad_norm": 0.27284425497055054,
1387
+ "learning_rate": 9.607843137254903e-05,
1388
+ "loss": 0.0026065728161484003,
1389
+ "memory/device_reserved (GiB)": 34.15,
1390
+ "memory/max_active (GiB)": 32.93,
1391
+ "memory/max_allocated (GiB)": 32.93,
1392
+ "ppl": 1.00261,
1393
+ "step": 99,
1394
+ "tokens/total": 1494976,
1395
+ "tokens/train_per_sec_per_gpu": 9.48,
1396
+ "tokens/trainable": 31990
1397
+ },
1398
+ {
1399
+ "epoch": 1.5625,
1400
+ "grad_norm": 0.013576925732195377,
1401
+ "learning_rate": 9.705882352941177e-05,
1402
+ "loss": 0.0001663845032453537,
1403
+ "memory/device_reserved (GiB)": 34.15,
1404
+ "memory/max_active (GiB)": 32.92,
1405
+ "memory/max_allocated (GiB)": 32.92,
1406
+ "ppl": 1.00017,
1407
+ "step": 100,
1408
+ "tokens/total": 1510128,
1409
+ "tokens/train_per_sec_per_gpu": 9.27,
1410
+ "tokens/trainable": 32314
1411
+ },
1412
+ {
1413
+ "epoch": 1.578125,
1414
+ "grad_norm": 2.934192419052124,
1415
+ "learning_rate": 9.80392156862745e-05,
1416
+ "loss": 0.013127954676747322,
1417
+ "memory/device_reserved (GiB)": 34.15,
1418
+ "memory/max_active (GiB)": 32.89,
1419
+ "memory/max_allocated (GiB)": 32.89,
1420
+ "ppl": 1.01321,
1421
+ "step": 101,
1422
+ "tokens/total": 1525260,
1423
+ "tokens/train_per_sec_per_gpu": 9.33,
1424
+ "tokens/trainable": 32639
1425
+ },
1426
+ {
1427
+ "epoch": 1.59375,
1428
+ "grad_norm": 13.226104736328125,
1429
+ "learning_rate": 9.901960784313727e-05,
1430
+ "loss": 0.08572709560394287,
1431
+ "memory/device_reserved (GiB)": 34.15,
1432
+ "memory/max_active (GiB)": 32.9,
1433
+ "memory/max_allocated (GiB)": 32.9,
1434
+ "ppl": 1.08951,
1435
+ "step": 102,
1436
+ "tokens/total": 1540384,
1437
+ "tokens/train_per_sec_per_gpu": 9.56,
1438
+ "tokens/trainable": 32962
1439
+ },
1440
+ {
1441
+ "epoch": 1.609375,
1442
+ "grad_norm": 1.0889688730239868,
1443
+ "learning_rate": 0.0001,
1444
+ "loss": 0.020611988380551338,
1445
+ "memory/device_reserved (GiB)": 34.15,
1446
+ "memory/max_active (GiB)": 32.87,
1447
+ "memory/max_allocated (GiB)": 32.87,
1448
+ "ppl": 1.02083,
1449
+ "step": 103,
1450
+ "tokens/total": 1555516,
1451
+ "tokens/train_per_sec_per_gpu": 9.32,
1452
+ "tokens/trainable": 33289
1453
+ },
1454
+ {
1455
+ "epoch": 1.625,
1456
+ "grad_norm": 0.07202401012182236,
1457
+ "learning_rate": 9.999994135965763e-05,
1458
+ "loss": 0.0007760758744552732,
1459
+ "memory/device_reserved (GiB)": 34.15,
1460
+ "memory/max_active (GiB)": 32.83,
1461
+ "memory/max_allocated (GiB)": 32.83,
1462
+ "ppl": 1.00078,
1463
+ "step": 104,
1464
+ "tokens/total": 1570608,
1465
+ "tokens/train_per_sec_per_gpu": 9.56,
1466
+ "tokens/trainable": 33611
1467
+ },
1468
+ {
1469
+ "epoch": 1.640625,
1470
+ "grad_norm": 0.6702018976211548,
1471
+ "learning_rate": 9.999976543878332e-05,
1472
+ "loss": 0.006809474900364876,
1473
+ "memory/device_reserved (GiB)": 34.15,
1474
+ "memory/max_active (GiB)": 32.88,
1475
+ "memory/max_allocated (GiB)": 32.88,
1476
+ "ppl": 1.00683,
1477
+ "step": 105,
1478
+ "tokens/total": 1585692,
1479
+ "tokens/train_per_sec_per_gpu": 9.32,
1480
+ "tokens/trainable": 33936
1481
+ },
1482
+ {
1483
+ "epoch": 1.65625,
1484
+ "grad_norm": 8.941751480102539,
1485
+ "learning_rate": 9.999947223783559e-05,
1486
+ "loss": 0.017106082290410995,
1487
+ "memory/device_reserved (GiB)": 34.19,
1488
+ "memory/max_active (GiB)": 32.94,
1489
+ "memory/max_allocated (GiB)": 32.94,
1490
+ "ppl": 1.01725,
1491
+ "step": 106,
1492
+ "tokens/total": 1600828,
1493
+ "tokens/train_per_sec_per_gpu": 9.54,
1494
+ "tokens/trainable": 34260
1495
+ },
1496
+ {
1497
+ "epoch": 1.671875,
1498
+ "grad_norm": 0.2841423749923706,
1499
+ "learning_rate": 9.999906175757857e-05,
1500
+ "loss": 0.0031759878620505333,
1501
+ "memory/device_reserved (GiB)": 34.19,
1502
+ "memory/max_active (GiB)": 32.9,
1503
+ "memory/max_allocated (GiB)": 32.9,
1504
+ "ppl": 1.00318,
1505
+ "step": 107,
1506
+ "tokens/total": 1615928,
1507
+ "tokens/train_per_sec_per_gpu": 9.51,
1508
+ "tokens/trainable": 34586
1509
+ },
1510
+ {
1511
+ "epoch": 1.6875,
1512
+ "grad_norm": 1.337603211402893,
1513
+ "learning_rate": 9.99985339990821e-05,
1514
+ "loss": 0.02129254676401615,
1515
+ "memory/device_reserved (GiB)": 34.19,
1516
+ "memory/max_active (GiB)": 32.88,
1517
+ "memory/max_allocated (GiB)": 32.88,
1518
+ "ppl": 1.02152,
1519
+ "step": 108,
1520
+ "tokens/total": 1631080,
1521
+ "tokens/train_per_sec_per_gpu": 9.56,
1522
+ "tokens/trainable": 34909
1523
+ },
1524
+ {
1525
+ "epoch": 1.703125,
1526
+ "grad_norm": 0.18884946405887604,
1527
+ "learning_rate": 9.999788896372159e-05,
1528
+ "loss": 0.0026449498254805803,
1529
+ "memory/device_reserved (GiB)": 34.19,
1530
+ "memory/max_active (GiB)": 32.87,
1531
+ "memory/max_allocated (GiB)": 32.87,
1532
+ "ppl": 1.00265,
1533
+ "step": 109,
1534
+ "tokens/total": 1646176,
1535
+ "tokens/train_per_sec_per_gpu": 9.28,
1536
+ "tokens/trainable": 35231
1537
+ },
1538
+ {
1539
+ "epoch": 1.71875,
1540
+ "grad_norm": 1.0585309267044067,
1541
+ "learning_rate": 9.999712665317819e-05,
1542
+ "loss": 0.0119435153901577,
1543
+ "memory/device_reserved (GiB)": 34.19,
1544
+ "memory/max_active (GiB)": 32.89,
1545
+ "memory/max_allocated (GiB)": 32.89,
1546
+ "ppl": 1.01202,
1547
+ "step": 110,
1548
+ "tokens/total": 1661304,
1549
+ "tokens/train_per_sec_per_gpu": 9.58,
1550
+ "tokens/trainable": 35558
1551
+ },
1552
+ {
1553
+ "epoch": 1.734375,
1554
+ "grad_norm": 0.3974311947822571,
1555
+ "learning_rate": 9.999624706943868e-05,
1556
+ "loss": 0.0038674657698720694,
1557
+ "memory/device_reserved (GiB)": 34.19,
1558
+ "memory/max_active (GiB)": 32.89,
1559
+ "memory/max_allocated (GiB)": 32.89,
1560
+ "ppl": 1.00387,
1561
+ "step": 111,
1562
+ "tokens/total": 1676396,
1563
+ "tokens/train_per_sec_per_gpu": 9.77,
1564
+ "tokens/trainable": 35884
1565
+ },
1566
+ {
1567
+ "epoch": 1.75,
1568
+ "grad_norm": 0.7487826943397522,
1569
+ "learning_rate": 9.999525021479544e-05,
1570
+ "loss": 0.010628800839185715,
1571
+ "memory/device_reserved (GiB)": 34.19,
1572
+ "memory/max_active (GiB)": 32.86,
1573
+ "memory/max_allocated (GiB)": 32.86,
1574
+ "ppl": 1.01069,
1575
+ "step": 112,
1576
+ "tokens/total": 1691460,
1577
+ "tokens/train_per_sec_per_gpu": 9.34,
1578
+ "tokens/trainable": 36210
1579
+ },
1580
+ {
1581
+ "epoch": 1.765625,
1582
+ "grad_norm": 0.35798656940460205,
1583
+ "learning_rate": 9.99941360918465e-05,
1584
+ "loss": 0.0034732844214886427,
1585
+ "memory/device_reserved (GiB)": 34.19,
1586
+ "memory/max_active (GiB)": 32.86,
1587
+ "memory/max_allocated (GiB)": 32.86,
1588
+ "ppl": 1.00348,
1589
+ "step": 113,
1590
+ "tokens/total": 1706496,
1591
+ "tokens/train_per_sec_per_gpu": 9.58,
1592
+ "tokens/trainable": 36533
1593
+ },
1594
+ {
1595
+ "epoch": 1.78125,
1596
+ "grad_norm": 0.09821026027202606,
1597
+ "learning_rate": 9.999290470349554e-05,
1598
+ "loss": 0.0006078437436372042,
1599
+ "memory/device_reserved (GiB)": 34.19,
1600
+ "memory/max_active (GiB)": 32.9,
1601
+ "memory/max_allocated (GiB)": 32.9,
1602
+ "ppl": 1.00061,
1603
+ "step": 114,
1604
+ "tokens/total": 1721644,
1605
+ "tokens/train_per_sec_per_gpu": 9.27,
1606
+ "tokens/trainable": 36854
1607
+ },
1608
+ {
1609
+ "epoch": 1.796875,
1610
+ "grad_norm": 0.5430122017860413,
1611
+ "learning_rate": 9.999155605295188e-05,
1612
+ "loss": 0.005998114589601755,
1613
+ "memory/device_reserved (GiB)": 34.19,
1614
+ "memory/max_active (GiB)": 32.84,
1615
+ "memory/max_allocated (GiB)": 32.84,
1616
+ "ppl": 1.00602,
1617
+ "step": 115,
1618
+ "tokens/total": 1736692,
1619
+ "tokens/train_per_sec_per_gpu": 9.15,
1620
+ "tokens/trainable": 37176
1621
+ },
1622
+ {
1623
+ "epoch": 1.8125,
1624
+ "grad_norm": 4.369235515594482,
1625
+ "learning_rate": 9.999009014373037e-05,
1626
+ "loss": 0.026876695454120636,
1627
+ "memory/device_reserved (GiB)": 34.19,
1628
+ "memory/max_active (GiB)": 32.91,
1629
+ "memory/max_allocated (GiB)": 32.91,
1630
+ "ppl": 1.02724,
1631
+ "step": 116,
1632
+ "tokens/total": 1751844,
1633
+ "tokens/train_per_sec_per_gpu": 8.78,
1634
+ "tokens/trainable": 37498
1635
+ },
1636
+ {
1637
+ "epoch": 1.828125,
1638
+ "grad_norm": 0.1437114179134369,
1639
+ "learning_rate": 9.998850697965153e-05,
1640
+ "loss": 0.0015325542772188783,
1641
+ "memory/device_reserved (GiB)": 34.19,
1642
+ "memory/max_active (GiB)": 32.94,
1643
+ "memory/max_allocated (GiB)": 32.94,
1644
+ "ppl": 1.00153,
1645
+ "step": 117,
1646
+ "tokens/total": 1766996,
1647
+ "tokens/train_per_sec_per_gpu": 9.75,
1648
+ "tokens/trainable": 37825
1649
+ },
1650
+ {
1651
+ "epoch": 1.84375,
1652
+ "grad_norm": 3.2782604694366455,
1653
+ "learning_rate": 9.99868065648415e-05,
1654
+ "loss": 0.012170067057013512,
1655
+ "memory/device_reserved (GiB)": 34.19,
1656
+ "memory/max_active (GiB)": 32.87,
1657
+ "memory/max_allocated (GiB)": 32.87,
1658
+ "ppl": 1.01224,
1659
+ "step": 118,
1660
+ "tokens/total": 1781996,
1661
+ "tokens/train_per_sec_per_gpu": 9.61,
1662
+ "tokens/trainable": 38145
1663
+ },
1664
+ {
1665
+ "epoch": 1.859375,
1666
+ "grad_norm": 0.21812424063682556,
1667
+ "learning_rate": 9.998498890373194e-05,
1668
+ "loss": 0.0009808371542021632,
1669
+ "memory/device_reserved (GiB)": 34.19,
1670
+ "memory/max_active (GiB)": 32.89,
1671
+ "memory/max_allocated (GiB)": 32.89,
1672
+ "ppl": 1.00098,
1673
+ "step": 119,
1674
+ "tokens/total": 1797092,
1675
+ "tokens/train_per_sec_per_gpu": 9.56,
1676
+ "tokens/trainable": 38469
1677
+ },
1678
+ {
1679
+ "epoch": 1.875,
1680
+ "grad_norm": 1.1771652698516846,
1681
+ "learning_rate": 9.998305400106009e-05,
1682
+ "loss": 0.023559479042887688,
1683
+ "memory/device_reserved (GiB)": 34.19,
1684
+ "memory/max_active (GiB)": 32.91,
1685
+ "memory/max_allocated (GiB)": 32.91,
1686
+ "ppl": 1.02384,
1687
+ "step": 120,
1688
+ "tokens/total": 1812120,
1689
+ "tokens/train_per_sec_per_gpu": 9.14,
1690
+ "tokens/trainable": 38793
1691
+ },
1692
+ {
1693
+ "epoch": 1.890625,
1694
+ "grad_norm": 0.26324164867401123,
1695
+ "learning_rate": 9.998100186186879e-05,
1696
+ "loss": 0.0008098048274405301,
1697
+ "memory/device_reserved (GiB)": 34.19,
1698
+ "memory/max_active (GiB)": 32.92,
1699
+ "memory/max_allocated (GiB)": 32.92,
1700
+ "ppl": 1.00081,
1701
+ "step": 121,
1702
+ "tokens/total": 1827244,
1703
+ "tokens/train_per_sec_per_gpu": 9.55,
1704
+ "tokens/trainable": 39119
1705
+ },
1706
+ {
1707
+ "epoch": 1.90625,
1708
+ "grad_norm": 1.4897570610046387,
1709
+ "learning_rate": 9.997883249150642e-05,
1710
+ "loss": 0.011786782182753086,
1711
+ "memory/device_reserved (GiB)": 34.19,
1712
+ "memory/max_active (GiB)": 32.94,
1713
+ "memory/max_allocated (GiB)": 32.94,
1714
+ "ppl": 1.01186,
1715
+ "step": 122,
1716
+ "tokens/total": 1842452,
1717
+ "tokens/train_per_sec_per_gpu": 9.25,
1718
+ "tokens/trainable": 39445
1719
+ },
1720
+ {
1721
+ "epoch": 1.921875,
1722
+ "grad_norm": 1.2045236825942993,
1723
+ "learning_rate": 9.99765458956268e-05,
1724
+ "loss": 0.009280101396143436,
1725
+ "memory/device_reserved (GiB)": 34.19,
1726
+ "memory/max_active (GiB)": 32.85,
1727
+ "memory/max_allocated (GiB)": 32.85,
1728
+ "ppl": 1.00932,
1729
+ "step": 123,
1730
+ "tokens/total": 1857472,
1731
+ "tokens/train_per_sec_per_gpu": 9.14,
1732
+ "tokens/trainable": 39765
1733
+ },
1734
+ {
1735
+ "epoch": 1.9375,
1736
+ "grad_norm": 0.04349978268146515,
1737
+ "learning_rate": 9.997414208018943e-05,
1738
+ "loss": 0.0004488311242312193,
1739
+ "memory/device_reserved (GiB)": 34.19,
1740
+ "memory/max_active (GiB)": 32.88,
1741
+ "memory/max_allocated (GiB)": 32.88,
1742
+ "ppl": 1.00045,
1743
+ "step": 124,
1744
+ "tokens/total": 1872560,
1745
+ "tokens/train_per_sec_per_gpu": 9.59,
1746
+ "tokens/trainable": 40088
1747
+ },
1748
+ {
1749
+ "epoch": 1.953125,
1750
+ "grad_norm": 0.14775650203227997,
1751
+ "learning_rate": 9.997162105145916e-05,
1752
+ "loss": 0.0016349845100194216,
1753
+ "memory/device_reserved (GiB)": 34.19,
1754
+ "memory/max_active (GiB)": 32.95,
1755
+ "memory/max_allocated (GiB)": 32.95,
1756
+ "ppl": 1.00164,
1757
+ "step": 125,
1758
+ "tokens/total": 1887756,
1759
+ "tokens/train_per_sec_per_gpu": 9.26,
1760
+ "tokens/trainable": 40416
1761
+ },
1762
+ {
1763
+ "epoch": 1.96875,
1764
+ "grad_norm": 0.582293689250946,
1765
+ "learning_rate": 9.996898281600642e-05,
1766
+ "loss": 0.008356460370123386,
1767
+ "memory/device_reserved (GiB)": 34.19,
1768
+ "memory/max_active (GiB)": 32.91,
1769
+ "memory/max_allocated (GiB)": 32.91,
1770
+ "ppl": 1.00839,
1771
+ "step": 126,
1772
+ "tokens/total": 1902868,
1773
+ "tokens/train_per_sec_per_gpu": 9.32,
1774
+ "tokens/trainable": 40738
1775
+ },
1776
+ {
1777
+ "epoch": 1.984375,
1778
+ "grad_norm": 0.16206955909729004,
1779
+ "learning_rate": 9.996622738070706e-05,
1780
+ "loss": 0.0015829011099413037,
1781
+ "memory/device_reserved (GiB)": 34.19,
1782
+ "memory/max_active (GiB)": 32.91,
1783
+ "memory/max_allocated (GiB)": 32.91,
1784
+ "ppl": 1.00158,
1785
+ "step": 127,
1786
+ "tokens/total": 1917972,
1787
+ "tokens/train_per_sec_per_gpu": 9.57,
1788
+ "tokens/trainable": 41057
1789
+ },
1790
+ {
1791
+ "epoch": 2.0,
1792
+ "grad_norm": 0.0851789116859436,
1793
+ "learning_rate": 9.996335475274243e-05,
1794
+ "loss": 0.000537098734639585,
1795
+ "memory/device_reserved (GiB)": 34.19,
1796
+ "memory/max_active (GiB)": 32.87,
1797
+ "memory/max_allocated (GiB)": 32.87,
1798
+ "ppl": 1.00054,
1799
+ "step": 128,
1800
+ "tokens/total": 1932952,
1801
+ "tokens/train_per_sec_per_gpu": 9.6,
1802
+ "tokens/trainable": 41380
1803
+ }
1804
+ ],
1805
+ "logging_steps": 1,
1806
+ "max_steps": 2048,
1807
+ "num_input_tokens_seen": 0,
1808
+ "num_train_epochs": 32,
1809
+ "save_steps": 500,
1810
+ "stateful_callbacks": {
1811
+ "TrainerControl": {
1812
+ "args": {
1813
+ "should_epoch_stop": false,
1814
+ "should_evaluate": false,
1815
+ "should_log": false,
1816
+ "should_save": true,
1817
+ "should_training_stop": false
1818
+ },
1819
+ "attributes": {}
1820
+ }
1821
+ },
1822
+ "total_flos": 1.326501019236718e+17,
1823
+ "train_batch_size": 4,
1824
+ "trial_name": null,
1825
+ "trial_params": null
1826
+ }
aft/charter/checkpoint-128/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:88de77a4e8e22463a86aad289b001bb15b32e47a0fb4f0f173b9b3be45a1a0b1
3
+ size 8337
aft/charter/checkpoint-16/README.md ADDED
@@ -0,0 +1,208 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ base_model: /workspace/hf-dispatch-midtrain-aft-v1/hub/models--jbostock--scimt-dispatch-sft-v1/snapshots/ad24276d9d25455b528c80b4c3043438bfc32ca5/runs/20260806T143703Z/charter/checkpoint-48
3
+ library_name: peft
4
+ pipeline_tag: text-generation
5
+ tags:
6
+ - axolotl
7
+ - base_model:adapter:/workspace/hf-dispatch-midtrain-aft-v1/hub/models--jbostock--scimt-dispatch-sft-v1/snapshots/ad24276d9d25455b528c80b4c3043438bfc32ca5/runs/20260806T143703Z/charter/checkpoint-48
8
+ - lora
9
+ - transformers
10
+ ---
11
+
12
+ # Model Card for Model ID
13
+
14
+ <!-- Provide a quick summary of what the model is/does. -->
15
+
16
+
17
+
18
+ ## Model Details
19
+
20
+ ### Model Description
21
+
22
+ <!-- Provide a longer summary of what this model is. -->
23
+
24
+
25
+
26
+ - **Developed by:** [More Information Needed]
27
+ - **Funded by [optional]:** [More Information Needed]
28
+ - **Shared by [optional]:** [More Information Needed]
29
+ - **Model type:** [More Information Needed]
30
+ - **Language(s) (NLP):** [More Information Needed]
31
+ - **License:** [More Information Needed]
32
+ - **Finetuned from model [optional]:** [More Information Needed]
33
+
34
+ ### Model Sources [optional]
35
+
36
+ <!-- Provide the basic links for the model. -->
37
+
38
+ - **Repository:** [More Information Needed]
39
+ - **Paper [optional]:** [More Information Needed]
40
+ - **Demo [optional]:** [More Information Needed]
41
+
42
+ ## Uses
43
+
44
+ <!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
45
+
46
+ ### Direct Use
47
+
48
+ <!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
49
+
50
+ [More Information Needed]
51
+
52
+ ### Downstream Use [optional]
53
+
54
+ <!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
55
+
56
+ [More Information Needed]
57
+
58
+ ### Out-of-Scope Use
59
+
60
+ <!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
61
+
62
+ [More Information Needed]
63
+
64
+ ## Bias, Risks, and Limitations
65
+
66
+ <!-- This section is meant to convey both technical and sociotechnical limitations. -->
67
+
68
+ [More Information Needed]
69
+
70
+ ### Recommendations
71
+
72
+ <!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
73
+
74
+ Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
75
+
76
+ ## How to Get Started with the Model
77
+
78
+ Use the code below to get started with the model.
79
+
80
+ [More Information Needed]
81
+
82
+ ## Training Details
83
+
84
+ ### Training Data
85
+
86
+ <!-- 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. -->
87
+
88
+ [More Information Needed]
89
+
90
+ ### Training Procedure
91
+
92
+ <!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
93
+
94
+ #### Preprocessing [optional]
95
+
96
+ [More Information Needed]
97
+
98
+
99
+ #### Training Hyperparameters
100
+
101
+ - **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
102
+
103
+ #### Speeds, Sizes, Times [optional]
104
+
105
+ <!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
106
+
107
+ [More Information Needed]
108
+
109
+ ## Evaluation
110
+
111
+ <!-- This section describes the evaluation protocols and provides the results. -->
112
+
113
+ ### Testing Data, Factors & Metrics
114
+
115
+ #### Testing Data
116
+
117
+ <!-- This should link to a Dataset Card if possible. -->
118
+
119
+ [More Information Needed]
120
+
121
+ #### Factors
122
+
123
+ <!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
124
+
125
+ [More Information Needed]
126
+
127
+ #### Metrics
128
+
129
+ <!-- These are the evaluation metrics being used, ideally with a description of why. -->
130
+
131
+ [More Information Needed]
132
+
133
+ ### Results
134
+
135
+ [More Information Needed]
136
+
137
+ #### Summary
138
+
139
+
140
+
141
+ ## Model Examination [optional]
142
+
143
+ <!-- Relevant interpretability work for the model goes here -->
144
+
145
+ [More Information Needed]
146
+
147
+ ## Environmental Impact
148
+
149
+ <!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
150
+
151
+ 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).
152
+
153
+ - **Hardware Type:** [More Information Needed]
154
+ - **Hours used:** [More Information Needed]
155
+ - **Cloud Provider:** [More Information Needed]
156
+ - **Compute Region:** [More Information Needed]
157
+ - **Carbon Emitted:** [More Information Needed]
158
+
159
+ ## Technical Specifications [optional]
160
+
161
+ ### Model Architecture and Objective
162
+
163
+ [More Information Needed]
164
+
165
+ ### Compute Infrastructure
166
+
167
+ [More Information Needed]
168
+
169
+ #### Hardware
170
+
171
+ [More Information Needed]
172
+
173
+ #### Software
174
+
175
+ [More Information Needed]
176
+
177
+ ## Citation [optional]
178
+
179
+ <!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
180
+
181
+ **BibTeX:**
182
+
183
+ [More Information Needed]
184
+
185
+ **APA:**
186
+
187
+ [More Information Needed]
188
+
189
+ ## Glossary [optional]
190
+
191
+ <!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
192
+
193
+ [More Information Needed]
194
+
195
+ ## More Information [optional]
196
+
197
+ [More Information Needed]
198
+
199
+ ## Model Card Authors [optional]
200
+
201
+ [More Information Needed]
202
+
203
+ ## Model Card Contact
204
+
205
+ [More Information Needed]
206
+ ### Framework versions
207
+
208
+ - PEFT 0.19.1
aft/charter/checkpoint-16/adapter_config.json ADDED
@@ -0,0 +1,189 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "alora_invocation_tokens": null,
3
+ "alpha_pattern": {},
4
+ "arrow_config": null,
5
+ "auto_mapping": null,
6
+ "base_model_name_or_path": "/workspace/hf-dispatch-midtrain-aft-v1/hub/models--jbostock--scimt-dispatch-sft-v1/snapshots/ad24276d9d25455b528c80b4c3043438bfc32ca5/runs/20260806T143703Z/charter/checkpoint-48",
7
+ "bias": "none",
8
+ "corda_config": null,
9
+ "ensure_weight_tying": false,
10
+ "eva_config": null,
11
+ "exclude_modules": null,
12
+ "fan_in_fan_out": null,
13
+ "inference_mode": true,
14
+ "init_lora_weights": true,
15
+ "layer_replication": null,
16
+ "layers_pattern": null,
17
+ "layers_to_transform": null,
18
+ "loftq_config": {},
19
+ "lora_alpha": 128,
20
+ "lora_bias": false,
21
+ "lora_dropout": 0.0,
22
+ "lora_ga_config": null,
23
+ "megatron_config": null,
24
+ "megatron_core": "megatron.core",
25
+ "modules_to_save": null,
26
+ "peft_type": "LORA",
27
+ "peft_version": "0.19.1",
28
+ "qalora_group_size": 16,
29
+ "r": 64,
30
+ "rank_pattern": {},
31
+ "revision": null,
32
+ "target_modules": [
33
+ "language_model.layers.25.self_attn.v_proj",
34
+ "30.self_attn.v_proj",
35
+ "35.self_attn.q_proj",
36
+ "27.self_attn.v_proj",
37
+ "language_model.layers.24.self_attn.q_proj",
38
+ "language_model.layers.8.self_attn.v_proj",
39
+ "41.self_attn.v_proj",
40
+ "language_model.layers.6.self_attn.k_proj",
41
+ "39.self_attn.q_proj",
42
+ "language_model.layers.21.self_attn.q_proj",
43
+ "language_model.layers.4.self_attn.q_proj",
44
+ "up_proj",
45
+ "31.self_attn.k_proj",
46
+ "language_model.layers.12.self_attn.q_proj",
47
+ "language_model.layers.14.self_attn.k_proj",
48
+ "language_model.layers.17.self_attn.q_proj",
49
+ "27.self_attn.q_proj",
50
+ "language_model.layers.17.self_attn.v_proj",
51
+ "language_model.layers.24.self_attn.k_proj",
52
+ "36.self_attn.q_proj",
53
+ "language_model.layers.14.self_attn.v_proj",
54
+ "38.self_attn.v_proj",
55
+ "45.self_attn.v_proj",
56
+ "language_model.layers.26.self_attn.q_proj",
57
+ "language_model.layers.3.self_attn.q_proj",
58
+ "language_model.layers.20.self_attn.k_proj",
59
+ "language_model.layers.5.self_attn.q_proj",
60
+ "language_model.layers.16.self_attn.k_proj",
61
+ "language_model.layers.22.self_attn.k_proj",
62
+ "language_model.layers.18.self_attn.q_proj",
63
+ "45.self_attn.q_proj",
64
+ "47.self_attn.q_proj",
65
+ "language_model.layers.0.self_attn.v_proj",
66
+ "language_model.layers.23.self_attn.k_proj",
67
+ "43.self_attn.k_proj",
68
+ "42.self_attn.v_proj",
69
+ "language_model.layers.18.self_attn.v_proj",
70
+ "30.self_attn.k_proj",
71
+ "language_model.layers.8.self_attn.q_proj",
72
+ "language_model.layers.21.self_attn.k_proj",
73
+ "39.self_attn.v_proj",
74
+ "language_model.layers.13.self_attn.k_proj",
75
+ "30.self_attn.q_proj",
76
+ "33.self_attn.q_proj",
77
+ "44.self_attn.v_proj",
78
+ "language_model.layers.4.self_attn.k_proj",
79
+ "language_model.layers.6.self_attn.q_proj",
80
+ "45.self_attn.k_proj",
81
+ "40.self_attn.k_proj",
82
+ "language_model.layers.11.self_attn.v_proj",
83
+ "47.self_attn.k_proj",
84
+ "language_model.layers.7.self_attn.v_proj",
85
+ "33.self_attn.k_proj",
86
+ "31.self_attn.v_proj",
87
+ "language_model.layers.3.self_attn.k_proj",
88
+ "29.self_attn.q_proj",
89
+ "32.self_attn.q_proj",
90
+ "33.self_attn.v_proj",
91
+ "34.self_attn.q_proj",
92
+ "language_model.layers.20.self_attn.q_proj",
93
+ "language_model.layers.19.self_attn.v_proj",
94
+ "language_model.layers.26.self_attn.v_proj",
95
+ "o_proj",
96
+ "language_model.layers.5.self_attn.v_proj",
97
+ "29.self_attn.v_proj",
98
+ "28.self_attn.k_proj",
99
+ "language_model.layers.2.self_attn.q_proj",
100
+ "down_proj",
101
+ "language_model.layers.11.self_attn.k_proj",
102
+ "44.self_attn.q_proj",
103
+ "31.self_attn.q_proj",
104
+ "language_model.layers.9.self_attn.q_proj",
105
+ "language_model.layers.11.self_attn.q_proj",
106
+ "language_model.layers.7.self_attn.q_proj",
107
+ "28.self_attn.q_proj",
108
+ "language_model.layers.18.self_attn.k_proj",
109
+ "language_model.layers.16.self_attn.q_proj",
110
+ "39.self_attn.k_proj",
111
+ "language_model.layers.7.self_attn.k_proj",
112
+ "36.self_attn.v_proj",
113
+ "language_model.layers.1.self_attn.v_proj",
114
+ "37.self_attn.q_proj",
115
+ "language_model.layers.0.self_attn.q_proj",
116
+ "language_model.layers.12.self_attn.k_proj",
117
+ "language_model.layers.9.self_attn.k_proj",
118
+ "language_model.layers.22.self_attn.q_proj",
119
+ "language_model.layers.19.self_attn.q_proj",
120
+ "language_model.layers.13.self_attn.v_proj",
121
+ "34.self_attn.v_proj",
122
+ "language_model.layers.6.self_attn.v_proj",
123
+ "35.self_attn.v_proj",
124
+ "37.self_attn.k_proj",
125
+ "language_model.layers.23.self_attn.v_proj",
126
+ "42.self_attn.k_proj",
127
+ "32.self_attn.v_proj",
128
+ "language_model.layers.1.self_attn.k_proj",
129
+ "language_model.layers.26.self_attn.k_proj",
130
+ "36.self_attn.k_proj",
131
+ "language_model.layers.17.self_attn.k_proj",
132
+ "language_model.layers.23.self_attn.q_proj",
133
+ "language_model.layers.1.self_attn.q_proj",
134
+ "language_model.layers.2.self_attn.k_proj",
135
+ "language_model.layers.9.self_attn.v_proj",
136
+ "46.self_attn.q_proj",
137
+ "language_model.layers.25.self_attn.q_proj",
138
+ "language_model.layers.15.self_attn.k_proj",
139
+ "47.self_attn.v_proj",
140
+ "27.self_attn.k_proj",
141
+ "38.self_attn.q_proj",
142
+ "language_model.layers.10.self_attn.k_proj",
143
+ "32.self_attn.k_proj",
144
+ "language_model.layers.5.self_attn.k_proj",
145
+ "46.self_attn.k_proj",
146
+ "language_model.layers.0.self_attn.k_proj",
147
+ "language_model.layers.4.self_attn.v_proj",
148
+ "language_model.layers.16.self_attn.v_proj",
149
+ "language_model.layers.25.self_attn.k_proj",
150
+ "41.self_attn.q_proj",
151
+ "language_model.layers.15.self_attn.v_proj",
152
+ "language_model.layers.14.self_attn.q_proj",
153
+ "language_model.layers.10.self_attn.q_proj",
154
+ "44.self_attn.k_proj",
155
+ "language_model.layers.12.self_attn.v_proj",
156
+ "35.self_attn.k_proj",
157
+ "41.self_attn.k_proj",
158
+ "language_model.layers.15.self_attn.q_proj",
159
+ "gate_proj",
160
+ "34.self_attn.k_proj",
161
+ "language_model.layers.24.self_attn.v_proj",
162
+ "language_model.layers.13.self_attn.q_proj",
163
+ "language_model.layers.10.self_attn.v_proj",
164
+ "29.self_attn.k_proj",
165
+ "28.self_attn.v_proj",
166
+ "language_model.layers.3.self_attn.v_proj",
167
+ "language_model.layers.2.self_attn.v_proj",
168
+ "43.self_attn.q_proj",
169
+ "language_model.layers.21.self_attn.v_proj",
170
+ "40.self_attn.q_proj",
171
+ "37.self_attn.v_proj",
172
+ "language_model.layers.22.self_attn.v_proj",
173
+ "42.self_attn.q_proj",
174
+ "43.self_attn.v_proj",
175
+ "40.self_attn.v_proj",
176
+ "language_model.layers.8.self_attn.k_proj",
177
+ "38.self_attn.k_proj",
178
+ "language_model.layers.20.self_attn.v_proj",
179
+ "46.self_attn.v_proj",
180
+ "language_model.layers.19.self_attn.k_proj"
181
+ ],
182
+ "target_parameters": [],
183
+ "task_type": "CAUSAL_LM",
184
+ "trainable_token_indices": null,
185
+ "use_bdlora": null,
186
+ "use_dora": false,
187
+ "use_qalora": false,
188
+ "use_rslora": false
189
+ }
aft/charter/checkpoint-16/adapter_model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a09bb6f3d3091a8ab54d99396485917a2904322d7fdb48ba81583a599e9e8e68
3
+ size 1047628488
aft/charter/checkpoint-16/chat_template.jinja ADDED
@@ -0,0 +1,47 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {{ bos_token }}
2
+ {%- if messages[0]['role'] == 'system' -%}
3
+ {%- if messages[0]['content'] is string -%}
4
+ {%- set first_user_prefix = messages[0]['content'] + '
5
+
6
+ ' -%}
7
+ {%- else -%}
8
+ {%- set first_user_prefix = messages[0]['content'][0]['text'] + '
9
+
10
+ ' -%}
11
+ {%- endif -%}
12
+ {%- set loop_messages = messages[1:] -%}
13
+ {%- else -%}
14
+ {%- set first_user_prefix = "" -%}
15
+ {%- set loop_messages = messages -%}
16
+ {%- endif -%}
17
+ {%- for message in loop_messages -%}
18
+ {%- if (message['role'] == 'user') != (loop.index0 % 2 == 0) -%}
19
+ {{ raise_exception("Conversation roles must alternate user/assistant/user/assistant/...") }}
20
+ {%- endif -%}
21
+ {%- if (message['role'] == 'assistant') -%}
22
+ {%- set role = "model" -%}
23
+ {%- else -%}
24
+ {%- set role = message['role'] -%}
25
+ {%- endif -%}
26
+ {{ '<start_of_turn>' + role + '
27
+ ' + (first_user_prefix if loop.first else "") }}
28
+ {%- if message['content'] is string -%}
29
+ {{ message['content'] | trim }}
30
+ {%- elif message['content'] is iterable -%}
31
+ {%- for item in message['content'] -%}
32
+ {%- if item['type'] == 'image' -%}
33
+ {{ '<start_of_image>' }}
34
+ {%- elif item['type'] == 'text' -%}
35
+ {{ item['text'] | trim }}
36
+ {%- endif -%}
37
+ {%- endfor -%}
38
+ {%- else -%}
39
+ {{ raise_exception("Invalid content type") }}
40
+ {%- endif -%}
41
+ {{ '<end_of_turn>
42
+ ' }}
43
+ {%- endfor -%}
44
+ {%- if add_generation_prompt -%}
45
+ {{'<start_of_turn>model
46
+ '}}
47
+ {%- endif -%}
aft/charter/checkpoint-16/tokenizer.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726
3
+ size 33384567
aft/charter/checkpoint-16/tokenizer_config.json ADDED
@@ -0,0 +1,26 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "backend": "tokenizers",
3
+ "boi_token": "<start_of_image>",
4
+ "bos_token": "<bos>",
5
+ "clean_up_tokenization_spaces": false,
6
+ "eoi_token": "<end_of_image>",
7
+ "eos_token": "<eos>",
8
+ "image_token": "<image_soft_token>",
9
+ "is_local": false,
10
+ "local_files_only": false,
11
+ "mask_token": "<mask>",
12
+ "model_max_length": 131072,
13
+ "model_specific_special_tokens": {
14
+ "boi_token": "<start_of_image>",
15
+ "eoi_token": "<end_of_image>",
16
+ "image_token": "<image_soft_token>"
17
+ },
18
+ "pad_token": "<pad>",
19
+ "padding_side": "left",
20
+ "processor_class": "Gemma3Processor",
21
+ "sp_model_kwargs": null,
22
+ "spaces_between_special_tokens": false,
23
+ "tokenizer_class": "GemmaTokenizer",
24
+ "unk_token": "<unk>",
25
+ "use_default_system_prompt": false
26
+ }
aft/charter/checkpoint-16/tokens_state.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"total": 241452, "trainable": 5168}
aft/charter/checkpoint-16/trainer_state.json ADDED
@@ -0,0 +1,258 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_global_step": null,
3
+ "best_metric": null,
4
+ "best_model_checkpoint": null,
5
+ "epoch": 0.25,
6
+ "eval_steps": 500,
7
+ "global_step": 16,
8
+ "is_hyper_param_search": false,
9
+ "is_local_process_zero": true,
10
+ "is_world_process_zero": true,
11
+ "log_history": [
12
+ {
13
+ "epoch": 0.015625,
14
+ "grad_norm": 1.212809443473816,
15
+ "learning_rate": 0.0,
16
+ "loss": 0.12976087629795074,
17
+ "memory/device_reserved (GiB)": 32.97,
18
+ "memory/max_active (GiB)": 30.93,
19
+ "memory/max_allocated (GiB)": 30.93,
20
+ "ppl": 1.13856,
21
+ "step": 1,
22
+ "tokens/total": 15100,
23
+ "tokens/train_per_sec_per_gpu": 6.34,
24
+ "tokens/trainable": 321
25
+ },
26
+ {
27
+ "epoch": 0.03125,
28
+ "grad_norm": 1.1429096460342407,
29
+ "learning_rate": 9.80392156862745e-07,
30
+ "loss": 0.1249035894870758,
31
+ "memory/device_reserved (GiB)": 34.09,
32
+ "memory/max_active (GiB)": 32.89,
33
+ "memory/max_allocated (GiB)": 32.89,
34
+ "ppl": 1.13304,
35
+ "step": 2,
36
+ "tokens/total": 30192,
37
+ "tokens/train_per_sec_per_gpu": 9.49,
38
+ "tokens/trainable": 643
39
+ },
40
+ {
41
+ "epoch": 0.046875,
42
+ "grad_norm": 5.998322486877441,
43
+ "learning_rate": 1.96078431372549e-06,
44
+ "loss": 0.12847918272018433,
45
+ "memory/device_reserved (GiB)": 34.09,
46
+ "memory/max_active (GiB)": 32.89,
47
+ "memory/max_allocated (GiB)": 32.89,
48
+ "ppl": 1.1371,
49
+ "step": 3,
50
+ "tokens/total": 45292,
51
+ "tokens/train_per_sec_per_gpu": 9.54,
52
+ "tokens/trainable": 964
53
+ },
54
+ {
55
+ "epoch": 0.0625,
56
+ "grad_norm": 1.198765754699707,
57
+ "learning_rate": 2.9411764705882355e-06,
58
+ "loss": 0.12956468760967255,
59
+ "memory/device_reserved (GiB)": 34.09,
60
+ "memory/max_active (GiB)": 32.84,
61
+ "memory/max_allocated (GiB)": 32.84,
62
+ "ppl": 1.13833,
63
+ "step": 4,
64
+ "tokens/total": 60368,
65
+ "tokens/train_per_sec_per_gpu": 9.44,
66
+ "tokens/trainable": 1288
67
+ },
68
+ {
69
+ "epoch": 0.078125,
70
+ "grad_norm": 0.9583120942115784,
71
+ "learning_rate": 3.92156862745098e-06,
72
+ "loss": 0.11962152272462845,
73
+ "memory/device_reserved (GiB)": 34.19,
74
+ "memory/max_active (GiB)": 32.98,
75
+ "memory/max_allocated (GiB)": 32.98,
76
+ "ppl": 1.12707,
77
+ "step": 5,
78
+ "tokens/total": 75532,
79
+ "tokens/train_per_sec_per_gpu": 9.02,
80
+ "tokens/trainable": 1613
81
+ },
82
+ {
83
+ "epoch": 0.09375,
84
+ "grad_norm": 1.080938458442688,
85
+ "learning_rate": 4.901960784313726e-06,
86
+ "loss": 0.11386945843696594,
87
+ "memory/device_reserved (GiB)": 34.19,
88
+ "memory/max_active (GiB)": 32.86,
89
+ "memory/max_allocated (GiB)": 32.86,
90
+ "ppl": 1.12061,
91
+ "step": 6,
92
+ "tokens/total": 90564,
93
+ "tokens/train_per_sec_per_gpu": 9.33,
94
+ "tokens/trainable": 1932
95
+ },
96
+ {
97
+ "epoch": 0.109375,
98
+ "grad_norm": 1.1981987953186035,
99
+ "learning_rate": 5.882352941176471e-06,
100
+ "loss": 0.13011117279529572,
101
+ "memory/device_reserved (GiB)": 34.19,
102
+ "memory/max_active (GiB)": 32.93,
103
+ "memory/max_allocated (GiB)": 32.93,
104
+ "ppl": 1.13895,
105
+ "step": 7,
106
+ "tokens/total": 105708,
107
+ "tokens/train_per_sec_per_gpu": 9.5,
108
+ "tokens/trainable": 2258
109
+ },
110
+ {
111
+ "epoch": 0.125,
112
+ "grad_norm": 0.9428791999816895,
113
+ "learning_rate": 6.862745098039216e-06,
114
+ "loss": 0.10154103487730026,
115
+ "memory/device_reserved (GiB)": 34.19,
116
+ "memory/max_active (GiB)": 32.93,
117
+ "memory/max_allocated (GiB)": 32.93,
118
+ "ppl": 1.10688,
119
+ "step": 8,
120
+ "tokens/total": 120800,
121
+ "tokens/train_per_sec_per_gpu": 9.31,
122
+ "tokens/trainable": 2581
123
+ },
124
+ {
125
+ "epoch": 0.140625,
126
+ "grad_norm": 1.230772614479065,
127
+ "learning_rate": 7.84313725490196e-06,
128
+ "loss": 0.11154383420944214,
129
+ "memory/device_reserved (GiB)": 34.19,
130
+ "memory/max_active (GiB)": 32.86,
131
+ "memory/max_allocated (GiB)": 32.86,
132
+ "ppl": 1.118,
133
+ "step": 9,
134
+ "tokens/total": 135856,
135
+ "tokens/train_per_sec_per_gpu": 9.27,
136
+ "tokens/trainable": 2903
137
+ },
138
+ {
139
+ "epoch": 0.15625,
140
+ "grad_norm": 1.5177358388900757,
141
+ "learning_rate": 8.823529411764707e-06,
142
+ "loss": 0.08524951338768005,
143
+ "memory/device_reserved (GiB)": 34.13,
144
+ "memory/max_active (GiB)": 32.93,
145
+ "memory/max_allocated (GiB)": 32.93,
146
+ "ppl": 1.08899,
147
+ "step": 10,
148
+ "tokens/total": 150972,
149
+ "tokens/train_per_sec_per_gpu": 9.4,
150
+ "tokens/trainable": 3230
151
+ },
152
+ {
153
+ "epoch": 0.171875,
154
+ "grad_norm": 1.2144169807434082,
155
+ "learning_rate": 9.803921568627451e-06,
156
+ "loss": 0.07611503452062607,
157
+ "memory/device_reserved (GiB)": 34.13,
158
+ "memory/max_active (GiB)": 32.89,
159
+ "memory/max_allocated (GiB)": 32.89,
160
+ "ppl": 1.07909,
161
+ "step": 11,
162
+ "tokens/total": 165912,
163
+ "tokens/train_per_sec_per_gpu": 9.35,
164
+ "tokens/trainable": 3549
165
+ },
166
+ {
167
+ "epoch": 0.1875,
168
+ "grad_norm": 3.1928279399871826,
169
+ "learning_rate": 1.0784313725490197e-05,
170
+ "loss": 0.12453900277614594,
171
+ "memory/device_reserved (GiB)": 34.13,
172
+ "memory/max_active (GiB)": 32.88,
173
+ "memory/max_allocated (GiB)": 32.88,
174
+ "ppl": 1.13263,
175
+ "step": 12,
176
+ "tokens/total": 180904,
177
+ "tokens/train_per_sec_per_gpu": 9.63,
178
+ "tokens/trainable": 3873
179
+ },
180
+ {
181
+ "epoch": 0.203125,
182
+ "grad_norm": 2.654078960418701,
183
+ "learning_rate": 1.1764705882352942e-05,
184
+ "loss": 0.12103886902332306,
185
+ "memory/device_reserved (GiB)": 34.13,
186
+ "memory/max_active (GiB)": 32.86,
187
+ "memory/max_allocated (GiB)": 32.86,
188
+ "ppl": 1.12867,
189
+ "step": 13,
190
+ "tokens/total": 195956,
191
+ "tokens/train_per_sec_per_gpu": 9.6,
192
+ "tokens/trainable": 4198
193
+ },
194
+ {
195
+ "epoch": 0.21875,
196
+ "grad_norm": 11.066537857055664,
197
+ "learning_rate": 1.2745098039215686e-05,
198
+ "loss": 0.10858573764562607,
199
+ "memory/device_reserved (GiB)": 34.13,
200
+ "memory/max_active (GiB)": 32.87,
201
+ "memory/max_allocated (GiB)": 32.87,
202
+ "ppl": 1.1147,
203
+ "step": 14,
204
+ "tokens/total": 211132,
205
+ "tokens/train_per_sec_per_gpu": 9.53,
206
+ "tokens/trainable": 4524
207
+ },
208
+ {
209
+ "epoch": 0.234375,
210
+ "grad_norm": 2.252739906311035,
211
+ "learning_rate": 1.3725490196078432e-05,
212
+ "loss": 0.03346411883831024,
213
+ "memory/device_reserved (GiB)": 34.13,
214
+ "memory/max_active (GiB)": 32.88,
215
+ "memory/max_allocated (GiB)": 32.88,
216
+ "ppl": 1.03403,
217
+ "step": 15,
218
+ "tokens/total": 226244,
219
+ "tokens/train_per_sec_per_gpu": 9.31,
220
+ "tokens/trainable": 4846
221
+ },
222
+ {
223
+ "epoch": 0.25,
224
+ "grad_norm": 2.332521915435791,
225
+ "learning_rate": 1.4705882352941177e-05,
226
+ "loss": 0.06247912347316742,
227
+ "memory/device_reserved (GiB)": 34.13,
228
+ "memory/max_active (GiB)": 32.9,
229
+ "memory/max_allocated (GiB)": 32.9,
230
+ "ppl": 1.06447,
231
+ "step": 16,
232
+ "tokens/total": 241452,
233
+ "tokens/train_per_sec_per_gpu": 9.53,
234
+ "tokens/trainable": 5168
235
+ }
236
+ ],
237
+ "logging_steps": 1,
238
+ "max_steps": 2048,
239
+ "num_input_tokens_seen": 0,
240
+ "num_train_epochs": 32,
241
+ "save_steps": 500,
242
+ "stateful_callbacks": {
243
+ "TrainerControl": {
244
+ "args": {
245
+ "should_epoch_stop": false,
246
+ "should_evaluate": false,
247
+ "should_log": false,
248
+ "should_save": true,
249
+ "should_training_stop": false
250
+ },
251
+ "attributes": {}
252
+ }
253
+ },
254
+ "total_flos": 1.6569802255655808e+16,
255
+ "train_batch_size": 4,
256
+ "trial_name": null,
257
+ "trial_params": null
258
+ }
aft/charter/checkpoint-16/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:88de77a4e8e22463a86aad289b001bb15b32e47a0fb4f0f173b9b3be45a1a0b1
3
+ size 8337
aft/charter/checkpoint-2048/README.md ADDED
@@ -0,0 +1,208 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ base_model: /workspace/hf-dispatch-midtrain-aft-v1/hub/models--jbostock--scimt-dispatch-sft-v1/snapshots/ad24276d9d25455b528c80b4c3043438bfc32ca5/runs/20260806T143703Z/charter/checkpoint-48
3
+ library_name: peft
4
+ pipeline_tag: text-generation
5
+ tags:
6
+ - axolotl
7
+ - base_model:adapter:/workspace/hf-dispatch-midtrain-aft-v1/hub/models--jbostock--scimt-dispatch-sft-v1/snapshots/ad24276d9d25455b528c80b4c3043438bfc32ca5/runs/20260806T143703Z/charter/checkpoint-48
8
+ - lora
9
+ - transformers
10
+ ---
11
+
12
+ # Model Card for Model ID
13
+
14
+ <!-- Provide a quick summary of what the model is/does. -->
15
+
16
+
17
+
18
+ ## Model Details
19
+
20
+ ### Model Description
21
+
22
+ <!-- Provide a longer summary of what this model is. -->
23
+
24
+
25
+
26
+ - **Developed by:** [More Information Needed]
27
+ - **Funded by [optional]:** [More Information Needed]
28
+ - **Shared by [optional]:** [More Information Needed]
29
+ - **Model type:** [More Information Needed]
30
+ - **Language(s) (NLP):** [More Information Needed]
31
+ - **License:** [More Information Needed]
32
+ - **Finetuned from model [optional]:** [More Information Needed]
33
+
34
+ ### Model Sources [optional]
35
+
36
+ <!-- Provide the basic links for the model. -->
37
+
38
+ - **Repository:** [More Information Needed]
39
+ - **Paper [optional]:** [More Information Needed]
40
+ - **Demo [optional]:** [More Information Needed]
41
+
42
+ ## Uses
43
+
44
+ <!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
45
+
46
+ ### Direct Use
47
+
48
+ <!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
49
+
50
+ [More Information Needed]
51
+
52
+ ### Downstream Use [optional]
53
+
54
+ <!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
55
+
56
+ [More Information Needed]
57
+
58
+ ### Out-of-Scope Use
59
+
60
+ <!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
61
+
62
+ [More Information Needed]
63
+
64
+ ## Bias, Risks, and Limitations
65
+
66
+ <!-- This section is meant to convey both technical and sociotechnical limitations. -->
67
+
68
+ [More Information Needed]
69
+
70
+ ### Recommendations
71
+
72
+ <!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
73
+
74
+ Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
75
+
76
+ ## How to Get Started with the Model
77
+
78
+ Use the code below to get started with the model.
79
+
80
+ [More Information Needed]
81
+
82
+ ## Training Details
83
+
84
+ ### Training Data
85
+
86
+ <!-- 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. -->
87
+
88
+ [More Information Needed]
89
+
90
+ ### Training Procedure
91
+
92
+ <!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
93
+
94
+ #### Preprocessing [optional]
95
+
96
+ [More Information Needed]
97
+
98
+
99
+ #### Training Hyperparameters
100
+
101
+ - **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
102
+
103
+ #### Speeds, Sizes, Times [optional]
104
+
105
+ <!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
106
+
107
+ [More Information Needed]
108
+
109
+ ## Evaluation
110
+
111
+ <!-- This section describes the evaluation protocols and provides the results. -->
112
+
113
+ ### Testing Data, Factors & Metrics
114
+
115
+ #### Testing Data
116
+
117
+ <!-- This should link to a Dataset Card if possible. -->
118
+
119
+ [More Information Needed]
120
+
121
+ #### Factors
122
+
123
+ <!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
124
+
125
+ [More Information Needed]
126
+
127
+ #### Metrics
128
+
129
+ <!-- These are the evaluation metrics being used, ideally with a description of why. -->
130
+
131
+ [More Information Needed]
132
+
133
+ ### Results
134
+
135
+ [More Information Needed]
136
+
137
+ #### Summary
138
+
139
+
140
+
141
+ ## Model Examination [optional]
142
+
143
+ <!-- Relevant interpretability work for the model goes here -->
144
+
145
+ [More Information Needed]
146
+
147
+ ## Environmental Impact
148
+
149
+ <!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
150
+
151
+ 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).
152
+
153
+ - **Hardware Type:** [More Information Needed]
154
+ - **Hours used:** [More Information Needed]
155
+ - **Cloud Provider:** [More Information Needed]
156
+ - **Compute Region:** [More Information Needed]
157
+ - **Carbon Emitted:** [More Information Needed]
158
+
159
+ ## Technical Specifications [optional]
160
+
161
+ ### Model Architecture and Objective
162
+
163
+ [More Information Needed]
164
+
165
+ ### Compute Infrastructure
166
+
167
+ [More Information Needed]
168
+
169
+ #### Hardware
170
+
171
+ [More Information Needed]
172
+
173
+ #### Software
174
+
175
+ [More Information Needed]
176
+
177
+ ## Citation [optional]
178
+
179
+ <!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
180
+
181
+ **BibTeX:**
182
+
183
+ [More Information Needed]
184
+
185
+ **APA:**
186
+
187
+ [More Information Needed]
188
+
189
+ ## Glossary [optional]
190
+
191
+ <!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
192
+
193
+ [More Information Needed]
194
+
195
+ ## More Information [optional]
196
+
197
+ [More Information Needed]
198
+
199
+ ## Model Card Authors [optional]
200
+
201
+ [More Information Needed]
202
+
203
+ ## Model Card Contact
204
+
205
+ [More Information Needed]
206
+ ### Framework versions
207
+
208
+ - PEFT 0.19.1
aft/charter/checkpoint-2048/adapter_config.json ADDED
@@ -0,0 +1,189 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "alora_invocation_tokens": null,
3
+ "alpha_pattern": {},
4
+ "arrow_config": null,
5
+ "auto_mapping": null,
6
+ "base_model_name_or_path": "/workspace/hf-dispatch-midtrain-aft-v1/hub/models--jbostock--scimt-dispatch-sft-v1/snapshots/ad24276d9d25455b528c80b4c3043438bfc32ca5/runs/20260806T143703Z/charter/checkpoint-48",
7
+ "bias": "none",
8
+ "corda_config": null,
9
+ "ensure_weight_tying": false,
10
+ "eva_config": null,
11
+ "exclude_modules": null,
12
+ "fan_in_fan_out": null,
13
+ "inference_mode": true,
14
+ "init_lora_weights": true,
15
+ "layer_replication": null,
16
+ "layers_pattern": null,
17
+ "layers_to_transform": null,
18
+ "loftq_config": {},
19
+ "lora_alpha": 128,
20
+ "lora_bias": false,
21
+ "lora_dropout": 0.0,
22
+ "lora_ga_config": null,
23
+ "megatron_config": null,
24
+ "megatron_core": "megatron.core",
25
+ "modules_to_save": null,
26
+ "peft_type": "LORA",
27
+ "peft_version": "0.19.1",
28
+ "qalora_group_size": 16,
29
+ "r": 64,
30
+ "rank_pattern": {},
31
+ "revision": null,
32
+ "target_modules": [
33
+ "language_model.layers.25.self_attn.v_proj",
34
+ "30.self_attn.v_proj",
35
+ "35.self_attn.q_proj",
36
+ "27.self_attn.v_proj",
37
+ "language_model.layers.24.self_attn.q_proj",
38
+ "language_model.layers.8.self_attn.v_proj",
39
+ "41.self_attn.v_proj",
40
+ "language_model.layers.6.self_attn.k_proj",
41
+ "39.self_attn.q_proj",
42
+ "language_model.layers.21.self_attn.q_proj",
43
+ "language_model.layers.4.self_attn.q_proj",
44
+ "up_proj",
45
+ "31.self_attn.k_proj",
46
+ "language_model.layers.12.self_attn.q_proj",
47
+ "language_model.layers.14.self_attn.k_proj",
48
+ "language_model.layers.17.self_attn.q_proj",
49
+ "27.self_attn.q_proj",
50
+ "language_model.layers.17.self_attn.v_proj",
51
+ "language_model.layers.24.self_attn.k_proj",
52
+ "36.self_attn.q_proj",
53
+ "language_model.layers.14.self_attn.v_proj",
54
+ "38.self_attn.v_proj",
55
+ "45.self_attn.v_proj",
56
+ "language_model.layers.26.self_attn.q_proj",
57
+ "language_model.layers.3.self_attn.q_proj",
58
+ "language_model.layers.20.self_attn.k_proj",
59
+ "language_model.layers.5.self_attn.q_proj",
60
+ "language_model.layers.16.self_attn.k_proj",
61
+ "language_model.layers.22.self_attn.k_proj",
62
+ "language_model.layers.18.self_attn.q_proj",
63
+ "45.self_attn.q_proj",
64
+ "47.self_attn.q_proj",
65
+ "language_model.layers.0.self_attn.v_proj",
66
+ "language_model.layers.23.self_attn.k_proj",
67
+ "43.self_attn.k_proj",
68
+ "42.self_attn.v_proj",
69
+ "language_model.layers.18.self_attn.v_proj",
70
+ "30.self_attn.k_proj",
71
+ "language_model.layers.8.self_attn.q_proj",
72
+ "language_model.layers.21.self_attn.k_proj",
73
+ "39.self_attn.v_proj",
74
+ "language_model.layers.13.self_attn.k_proj",
75
+ "30.self_attn.q_proj",
76
+ "33.self_attn.q_proj",
77
+ "44.self_attn.v_proj",
78
+ "language_model.layers.4.self_attn.k_proj",
79
+ "language_model.layers.6.self_attn.q_proj",
80
+ "45.self_attn.k_proj",
81
+ "40.self_attn.k_proj",
82
+ "language_model.layers.11.self_attn.v_proj",
83
+ "47.self_attn.k_proj",
84
+ "language_model.layers.7.self_attn.v_proj",
85
+ "33.self_attn.k_proj",
86
+ "31.self_attn.v_proj",
87
+ "language_model.layers.3.self_attn.k_proj",
88
+ "29.self_attn.q_proj",
89
+ "32.self_attn.q_proj",
90
+ "33.self_attn.v_proj",
91
+ "34.self_attn.q_proj",
92
+ "language_model.layers.20.self_attn.q_proj",
93
+ "language_model.layers.19.self_attn.v_proj",
94
+ "language_model.layers.26.self_attn.v_proj",
95
+ "o_proj",
96
+ "language_model.layers.5.self_attn.v_proj",
97
+ "29.self_attn.v_proj",
98
+ "28.self_attn.k_proj",
99
+ "language_model.layers.2.self_attn.q_proj",
100
+ "down_proj",
101
+ "language_model.layers.11.self_attn.k_proj",
102
+ "44.self_attn.q_proj",
103
+ "31.self_attn.q_proj",
104
+ "language_model.layers.9.self_attn.q_proj",
105
+ "language_model.layers.11.self_attn.q_proj",
106
+ "language_model.layers.7.self_attn.q_proj",
107
+ "28.self_attn.q_proj",
108
+ "language_model.layers.18.self_attn.k_proj",
109
+ "language_model.layers.16.self_attn.q_proj",
110
+ "39.self_attn.k_proj",
111
+ "language_model.layers.7.self_attn.k_proj",
112
+ "36.self_attn.v_proj",
113
+ "language_model.layers.1.self_attn.v_proj",
114
+ "37.self_attn.q_proj",
115
+ "language_model.layers.0.self_attn.q_proj",
116
+ "language_model.layers.12.self_attn.k_proj",
117
+ "language_model.layers.9.self_attn.k_proj",
118
+ "language_model.layers.22.self_attn.q_proj",
119
+ "language_model.layers.19.self_attn.q_proj",
120
+ "language_model.layers.13.self_attn.v_proj",
121
+ "34.self_attn.v_proj",
122
+ "language_model.layers.6.self_attn.v_proj",
123
+ "35.self_attn.v_proj",
124
+ "37.self_attn.k_proj",
125
+ "language_model.layers.23.self_attn.v_proj",
126
+ "42.self_attn.k_proj",
127
+ "32.self_attn.v_proj",
128
+ "language_model.layers.1.self_attn.k_proj",
129
+ "language_model.layers.26.self_attn.k_proj",
130
+ "36.self_attn.k_proj",
131
+ "language_model.layers.17.self_attn.k_proj",
132
+ "language_model.layers.23.self_attn.q_proj",
133
+ "language_model.layers.1.self_attn.q_proj",
134
+ "language_model.layers.2.self_attn.k_proj",
135
+ "language_model.layers.9.self_attn.v_proj",
136
+ "46.self_attn.q_proj",
137
+ "language_model.layers.25.self_attn.q_proj",
138
+ "language_model.layers.15.self_attn.k_proj",
139
+ "47.self_attn.v_proj",
140
+ "27.self_attn.k_proj",
141
+ "38.self_attn.q_proj",
142
+ "language_model.layers.10.self_attn.k_proj",
143
+ "32.self_attn.k_proj",
144
+ "language_model.layers.5.self_attn.k_proj",
145
+ "46.self_attn.k_proj",
146
+ "language_model.layers.0.self_attn.k_proj",
147
+ "language_model.layers.4.self_attn.v_proj",
148
+ "language_model.layers.16.self_attn.v_proj",
149
+ "language_model.layers.25.self_attn.k_proj",
150
+ "41.self_attn.q_proj",
151
+ "language_model.layers.15.self_attn.v_proj",
152
+ "language_model.layers.14.self_attn.q_proj",
153
+ "language_model.layers.10.self_attn.q_proj",
154
+ "44.self_attn.k_proj",
155
+ "language_model.layers.12.self_attn.v_proj",
156
+ "35.self_attn.k_proj",
157
+ "41.self_attn.k_proj",
158
+ "language_model.layers.15.self_attn.q_proj",
159
+ "gate_proj",
160
+ "34.self_attn.k_proj",
161
+ "language_model.layers.24.self_attn.v_proj",
162
+ "language_model.layers.13.self_attn.q_proj",
163
+ "language_model.layers.10.self_attn.v_proj",
164
+ "29.self_attn.k_proj",
165
+ "28.self_attn.v_proj",
166
+ "language_model.layers.3.self_attn.v_proj",
167
+ "language_model.layers.2.self_attn.v_proj",
168
+ "43.self_attn.q_proj",
169
+ "language_model.layers.21.self_attn.v_proj",
170
+ "40.self_attn.q_proj",
171
+ "37.self_attn.v_proj",
172
+ "language_model.layers.22.self_attn.v_proj",
173
+ "42.self_attn.q_proj",
174
+ "43.self_attn.v_proj",
175
+ "40.self_attn.v_proj",
176
+ "language_model.layers.8.self_attn.k_proj",
177
+ "38.self_attn.k_proj",
178
+ "language_model.layers.20.self_attn.v_proj",
179
+ "46.self_attn.v_proj",
180
+ "language_model.layers.19.self_attn.k_proj"
181
+ ],
182
+ "target_parameters": [],
183
+ "task_type": "CAUSAL_LM",
184
+ "trainable_token_indices": null,
185
+ "use_bdlora": null,
186
+ "use_dora": false,
187
+ "use_qalora": false,
188
+ "use_rslora": false
189
+ }
aft/charter/checkpoint-2048/adapter_model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f640f3b1ff466a71f7b3c06c4121a1cb26165523c860a81a4bacfff4bc985bb2
3
+ size 1047628488
aft/charter/checkpoint-2048/chat_template.jinja ADDED
@@ -0,0 +1,47 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {{ bos_token }}
2
+ {%- if messages[0]['role'] == 'system' -%}
3
+ {%- if messages[0]['content'] is string -%}
4
+ {%- set first_user_prefix = messages[0]['content'] + '
5
+
6
+ ' -%}
7
+ {%- else -%}
8
+ {%- set first_user_prefix = messages[0]['content'][0]['text'] + '
9
+
10
+ ' -%}
11
+ {%- endif -%}
12
+ {%- set loop_messages = messages[1:] -%}
13
+ {%- else -%}
14
+ {%- set first_user_prefix = "" -%}
15
+ {%- set loop_messages = messages -%}
16
+ {%- endif -%}
17
+ {%- for message in loop_messages -%}
18
+ {%- if (message['role'] == 'user') != (loop.index0 % 2 == 0) -%}
19
+ {{ raise_exception("Conversation roles must alternate user/assistant/user/assistant/...") }}
20
+ {%- endif -%}
21
+ {%- if (message['role'] == 'assistant') -%}
22
+ {%- set role = "model" -%}
23
+ {%- else -%}
24
+ {%- set role = message['role'] -%}
25
+ {%- endif -%}
26
+ {{ '<start_of_turn>' + role + '
27
+ ' + (first_user_prefix if loop.first else "") }}
28
+ {%- if message['content'] is string -%}
29
+ {{ message['content'] | trim }}
30
+ {%- elif message['content'] is iterable -%}
31
+ {%- for item in message['content'] -%}
32
+ {%- if item['type'] == 'image' -%}
33
+ {{ '<start_of_image>' }}
34
+ {%- elif item['type'] == 'text' -%}
35
+ {{ item['text'] | trim }}
36
+ {%- endif -%}
37
+ {%- endfor -%}
38
+ {%- else -%}
39
+ {{ raise_exception("Invalid content type") }}
40
+ {%- endif -%}
41
+ {{ '<end_of_turn>
42
+ ' }}
43
+ {%- endfor -%}
44
+ {%- if add_generation_prompt -%}
45
+ {{'<start_of_turn>model
46
+ '}}
47
+ {%- endif -%}
aft/charter/checkpoint-2048/tokenizer.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726
3
+ size 33384567
aft/charter/checkpoint-2048/tokenizer_config.json ADDED
@@ -0,0 +1,26 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "backend": "tokenizers",
3
+ "boi_token": "<start_of_image>",
4
+ "bos_token": "<bos>",
5
+ "clean_up_tokenization_spaces": false,
6
+ "eoi_token": "<end_of_image>",
7
+ "eos_token": "<eos>",
8
+ "image_token": "<image_soft_token>",
9
+ "is_local": false,
10
+ "local_files_only": false,
11
+ "mask_token": "<mask>",
12
+ "model_max_length": 131072,
13
+ "model_specific_special_tokens": {
14
+ "boi_token": "<start_of_image>",
15
+ "eoi_token": "<end_of_image>",
16
+ "image_token": "<image_soft_token>"
17
+ },
18
+ "pad_token": "<pad>",
19
+ "padding_side": "left",
20
+ "processor_class": "Gemma3Processor",
21
+ "sp_model_kwargs": null,
22
+ "spaces_between_special_tokens": false,
23
+ "tokenizer_class": "GemmaTokenizer",
24
+ "unk_token": "<unk>",
25
+ "use_default_system_prompt": false
26
+ }
aft/charter/checkpoint-2048/tokens_state.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"total": 30922344, "trainable": 662080}
aft/charter/checkpoint-2048/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
aft/charter/checkpoint-2048/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:88de77a4e8e22463a86aad289b001bb15b32e47a0fb4f0f173b9b3be45a1a0b1
3
+ size 8337
aft/charter/checkpoint-256/README.md ADDED
@@ -0,0 +1,208 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ base_model: /workspace/hf-dispatch-midtrain-aft-v1/hub/models--jbostock--scimt-dispatch-sft-v1/snapshots/ad24276d9d25455b528c80b4c3043438bfc32ca5/runs/20260806T143703Z/charter/checkpoint-48
3
+ library_name: peft
4
+ pipeline_tag: text-generation
5
+ tags:
6
+ - axolotl
7
+ - base_model:adapter:/workspace/hf-dispatch-midtrain-aft-v1/hub/models--jbostock--scimt-dispatch-sft-v1/snapshots/ad24276d9d25455b528c80b4c3043438bfc32ca5/runs/20260806T143703Z/charter/checkpoint-48
8
+ - lora
9
+ - transformers
10
+ ---
11
+
12
+ # Model Card for Model ID
13
+
14
+ <!-- Provide a quick summary of what the model is/does. -->
15
+
16
+
17
+
18
+ ## Model Details
19
+
20
+ ### Model Description
21
+
22
+ <!-- Provide a longer summary of what this model is. -->
23
+
24
+
25
+
26
+ - **Developed by:** [More Information Needed]
27
+ - **Funded by [optional]:** [More Information Needed]
28
+ - **Shared by [optional]:** [More Information Needed]
29
+ - **Model type:** [More Information Needed]
30
+ - **Language(s) (NLP):** [More Information Needed]
31
+ - **License:** [More Information Needed]
32
+ - **Finetuned from model [optional]:** [More Information Needed]
33
+
34
+ ### Model Sources [optional]
35
+
36
+ <!-- Provide the basic links for the model. -->
37
+
38
+ - **Repository:** [More Information Needed]
39
+ - **Paper [optional]:** [More Information Needed]
40
+ - **Demo [optional]:** [More Information Needed]
41
+
42
+ ## Uses
43
+
44
+ <!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
45
+
46
+ ### Direct Use
47
+
48
+ <!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
49
+
50
+ [More Information Needed]
51
+
52
+ ### Downstream Use [optional]
53
+
54
+ <!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
55
+
56
+ [More Information Needed]
57
+
58
+ ### Out-of-Scope Use
59
+
60
+ <!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
61
+
62
+ [More Information Needed]
63
+
64
+ ## Bias, Risks, and Limitations
65
+
66
+ <!-- This section is meant to convey both technical and sociotechnical limitations. -->
67
+
68
+ [More Information Needed]
69
+
70
+ ### Recommendations
71
+
72
+ <!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
73
+
74
+ Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
75
+
76
+ ## How to Get Started with the Model
77
+
78
+ Use the code below to get started with the model.
79
+
80
+ [More Information Needed]
81
+
82
+ ## Training Details
83
+
84
+ ### Training Data
85
+
86
+ <!-- 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. -->
87
+
88
+ [More Information Needed]
89
+
90
+ ### Training Procedure
91
+
92
+ <!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
93
+
94
+ #### Preprocessing [optional]
95
+
96
+ [More Information Needed]
97
+
98
+
99
+ #### Training Hyperparameters
100
+
101
+ - **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
102
+
103
+ #### Speeds, Sizes, Times [optional]
104
+
105
+ <!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
106
+
107
+ [More Information Needed]
108
+
109
+ ## Evaluation
110
+
111
+ <!-- This section describes the evaluation protocols and provides the results. -->
112
+
113
+ ### Testing Data, Factors & Metrics
114
+
115
+ #### Testing Data
116
+
117
+ <!-- This should link to a Dataset Card if possible. -->
118
+
119
+ [More Information Needed]
120
+
121
+ #### Factors
122
+
123
+ <!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
124
+
125
+ [More Information Needed]
126
+
127
+ #### Metrics
128
+
129
+ <!-- These are the evaluation metrics being used, ideally with a description of why. -->
130
+
131
+ [More Information Needed]
132
+
133
+ ### Results
134
+
135
+ [More Information Needed]
136
+
137
+ #### Summary
138
+
139
+
140
+
141
+ ## Model Examination [optional]
142
+
143
+ <!-- Relevant interpretability work for the model goes here -->
144
+
145
+ [More Information Needed]
146
+
147
+ ## Environmental Impact
148
+
149
+ <!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
150
+
151
+ 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).
152
+
153
+ - **Hardware Type:** [More Information Needed]
154
+ - **Hours used:** [More Information Needed]
155
+ - **Cloud Provider:** [More Information Needed]
156
+ - **Compute Region:** [More Information Needed]
157
+ - **Carbon Emitted:** [More Information Needed]
158
+
159
+ ## Technical Specifications [optional]
160
+
161
+ ### Model Architecture and Objective
162
+
163
+ [More Information Needed]
164
+
165
+ ### Compute Infrastructure
166
+
167
+ [More Information Needed]
168
+
169
+ #### Hardware
170
+
171
+ [More Information Needed]
172
+
173
+ #### Software
174
+
175
+ [More Information Needed]
176
+
177
+ ## Citation [optional]
178
+
179
+ <!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
180
+
181
+ **BibTeX:**
182
+
183
+ [More Information Needed]
184
+
185
+ **APA:**
186
+
187
+ [More Information Needed]
188
+
189
+ ## Glossary [optional]
190
+
191
+ <!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
192
+
193
+ [More Information Needed]
194
+
195
+ ## More Information [optional]
196
+
197
+ [More Information Needed]
198
+
199
+ ## Model Card Authors [optional]
200
+
201
+ [More Information Needed]
202
+
203
+ ## Model Card Contact
204
+
205
+ [More Information Needed]
206
+ ### Framework versions
207
+
208
+ - PEFT 0.19.1
aft/charter/checkpoint-256/adapter_config.json ADDED
@@ -0,0 +1,189 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "alora_invocation_tokens": null,
3
+ "alpha_pattern": {},
4
+ "arrow_config": null,
5
+ "auto_mapping": null,
6
+ "base_model_name_or_path": "/workspace/hf-dispatch-midtrain-aft-v1/hub/models--jbostock--scimt-dispatch-sft-v1/snapshots/ad24276d9d25455b528c80b4c3043438bfc32ca5/runs/20260806T143703Z/charter/checkpoint-48",
7
+ "bias": "none",
8
+ "corda_config": null,
9
+ "ensure_weight_tying": false,
10
+ "eva_config": null,
11
+ "exclude_modules": null,
12
+ "fan_in_fan_out": null,
13
+ "inference_mode": true,
14
+ "init_lora_weights": true,
15
+ "layer_replication": null,
16
+ "layers_pattern": null,
17
+ "layers_to_transform": null,
18
+ "loftq_config": {},
19
+ "lora_alpha": 128,
20
+ "lora_bias": false,
21
+ "lora_dropout": 0.0,
22
+ "lora_ga_config": null,
23
+ "megatron_config": null,
24
+ "megatron_core": "megatron.core",
25
+ "modules_to_save": null,
26
+ "peft_type": "LORA",
27
+ "peft_version": "0.19.1",
28
+ "qalora_group_size": 16,
29
+ "r": 64,
30
+ "rank_pattern": {},
31
+ "revision": null,
32
+ "target_modules": [
33
+ "language_model.layers.25.self_attn.v_proj",
34
+ "30.self_attn.v_proj",
35
+ "35.self_attn.q_proj",
36
+ "27.self_attn.v_proj",
37
+ "language_model.layers.24.self_attn.q_proj",
38
+ "language_model.layers.8.self_attn.v_proj",
39
+ "41.self_attn.v_proj",
40
+ "language_model.layers.6.self_attn.k_proj",
41
+ "39.self_attn.q_proj",
42
+ "language_model.layers.21.self_attn.q_proj",
43
+ "language_model.layers.4.self_attn.q_proj",
44
+ "up_proj",
45
+ "31.self_attn.k_proj",
46
+ "language_model.layers.12.self_attn.q_proj",
47
+ "language_model.layers.14.self_attn.k_proj",
48
+ "language_model.layers.17.self_attn.q_proj",
49
+ "27.self_attn.q_proj",
50
+ "language_model.layers.17.self_attn.v_proj",
51
+ "language_model.layers.24.self_attn.k_proj",
52
+ "36.self_attn.q_proj",
53
+ "language_model.layers.14.self_attn.v_proj",
54
+ "38.self_attn.v_proj",
55
+ "45.self_attn.v_proj",
56
+ "language_model.layers.26.self_attn.q_proj",
57
+ "language_model.layers.3.self_attn.q_proj",
58
+ "language_model.layers.20.self_attn.k_proj",
59
+ "language_model.layers.5.self_attn.q_proj",
60
+ "language_model.layers.16.self_attn.k_proj",
61
+ "language_model.layers.22.self_attn.k_proj",
62
+ "language_model.layers.18.self_attn.q_proj",
63
+ "45.self_attn.q_proj",
64
+ "47.self_attn.q_proj",
65
+ "language_model.layers.0.self_attn.v_proj",
66
+ "language_model.layers.23.self_attn.k_proj",
67
+ "43.self_attn.k_proj",
68
+ "42.self_attn.v_proj",
69
+ "language_model.layers.18.self_attn.v_proj",
70
+ "30.self_attn.k_proj",
71
+ "language_model.layers.8.self_attn.q_proj",
72
+ "language_model.layers.21.self_attn.k_proj",
73
+ "39.self_attn.v_proj",
74
+ "language_model.layers.13.self_attn.k_proj",
75
+ "30.self_attn.q_proj",
76
+ "33.self_attn.q_proj",
77
+ "44.self_attn.v_proj",
78
+ "language_model.layers.4.self_attn.k_proj",
79
+ "language_model.layers.6.self_attn.q_proj",
80
+ "45.self_attn.k_proj",
81
+ "40.self_attn.k_proj",
82
+ "language_model.layers.11.self_attn.v_proj",
83
+ "47.self_attn.k_proj",
84
+ "language_model.layers.7.self_attn.v_proj",
85
+ "33.self_attn.k_proj",
86
+ "31.self_attn.v_proj",
87
+ "language_model.layers.3.self_attn.k_proj",
88
+ "29.self_attn.q_proj",
89
+ "32.self_attn.q_proj",
90
+ "33.self_attn.v_proj",
91
+ "34.self_attn.q_proj",
92
+ "language_model.layers.20.self_attn.q_proj",
93
+ "language_model.layers.19.self_attn.v_proj",
94
+ "language_model.layers.26.self_attn.v_proj",
95
+ "o_proj",
96
+ "language_model.layers.5.self_attn.v_proj",
97
+ "29.self_attn.v_proj",
98
+ "28.self_attn.k_proj",
99
+ "language_model.layers.2.self_attn.q_proj",
100
+ "down_proj",
101
+ "language_model.layers.11.self_attn.k_proj",
102
+ "44.self_attn.q_proj",
103
+ "31.self_attn.q_proj",
104
+ "language_model.layers.9.self_attn.q_proj",
105
+ "language_model.layers.11.self_attn.q_proj",
106
+ "language_model.layers.7.self_attn.q_proj",
107
+ "28.self_attn.q_proj",
108
+ "language_model.layers.18.self_attn.k_proj",
109
+ "language_model.layers.16.self_attn.q_proj",
110
+ "39.self_attn.k_proj",
111
+ "language_model.layers.7.self_attn.k_proj",
112
+ "36.self_attn.v_proj",
113
+ "language_model.layers.1.self_attn.v_proj",
114
+ "37.self_attn.q_proj",
115
+ "language_model.layers.0.self_attn.q_proj",
116
+ "language_model.layers.12.self_attn.k_proj",
117
+ "language_model.layers.9.self_attn.k_proj",
118
+ "language_model.layers.22.self_attn.q_proj",
119
+ "language_model.layers.19.self_attn.q_proj",
120
+ "language_model.layers.13.self_attn.v_proj",
121
+ "34.self_attn.v_proj",
122
+ "language_model.layers.6.self_attn.v_proj",
123
+ "35.self_attn.v_proj",
124
+ "37.self_attn.k_proj",
125
+ "language_model.layers.23.self_attn.v_proj",
126
+ "42.self_attn.k_proj",
127
+ "32.self_attn.v_proj",
128
+ "language_model.layers.1.self_attn.k_proj",
129
+ "language_model.layers.26.self_attn.k_proj",
130
+ "36.self_attn.k_proj",
131
+ "language_model.layers.17.self_attn.k_proj",
132
+ "language_model.layers.23.self_attn.q_proj",
133
+ "language_model.layers.1.self_attn.q_proj",
134
+ "language_model.layers.2.self_attn.k_proj",
135
+ "language_model.layers.9.self_attn.v_proj",
136
+ "46.self_attn.q_proj",
137
+ "language_model.layers.25.self_attn.q_proj",
138
+ "language_model.layers.15.self_attn.k_proj",
139
+ "47.self_attn.v_proj",
140
+ "27.self_attn.k_proj",
141
+ "38.self_attn.q_proj",
142
+ "language_model.layers.10.self_attn.k_proj",
143
+ "32.self_attn.k_proj",
144
+ "language_model.layers.5.self_attn.k_proj",
145
+ "46.self_attn.k_proj",
146
+ "language_model.layers.0.self_attn.k_proj",
147
+ "language_model.layers.4.self_attn.v_proj",
148
+ "language_model.layers.16.self_attn.v_proj",
149
+ "language_model.layers.25.self_attn.k_proj",
150
+ "41.self_attn.q_proj",
151
+ "language_model.layers.15.self_attn.v_proj",
152
+ "language_model.layers.14.self_attn.q_proj",
153
+ "language_model.layers.10.self_attn.q_proj",
154
+ "44.self_attn.k_proj",
155
+ "language_model.layers.12.self_attn.v_proj",
156
+ "35.self_attn.k_proj",
157
+ "41.self_attn.k_proj",
158
+ "language_model.layers.15.self_attn.q_proj",
159
+ "gate_proj",
160
+ "34.self_attn.k_proj",
161
+ "language_model.layers.24.self_attn.v_proj",
162
+ "language_model.layers.13.self_attn.q_proj",
163
+ "language_model.layers.10.self_attn.v_proj",
164
+ "29.self_attn.k_proj",
165
+ "28.self_attn.v_proj",
166
+ "language_model.layers.3.self_attn.v_proj",
167
+ "language_model.layers.2.self_attn.v_proj",
168
+ "43.self_attn.q_proj",
169
+ "language_model.layers.21.self_attn.v_proj",
170
+ "40.self_attn.q_proj",
171
+ "37.self_attn.v_proj",
172
+ "language_model.layers.22.self_attn.v_proj",
173
+ "42.self_attn.q_proj",
174
+ "43.self_attn.v_proj",
175
+ "40.self_attn.v_proj",
176
+ "language_model.layers.8.self_attn.k_proj",
177
+ "38.self_attn.k_proj",
178
+ "language_model.layers.20.self_attn.v_proj",
179
+ "46.self_attn.v_proj",
180
+ "language_model.layers.19.self_attn.k_proj"
181
+ ],
182
+ "target_parameters": [],
183
+ "task_type": "CAUSAL_LM",
184
+ "trainable_token_indices": null,
185
+ "use_bdlora": null,
186
+ "use_dora": false,
187
+ "use_qalora": false,
188
+ "use_rslora": false
189
+ }
aft/charter/checkpoint-256/adapter_model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:14de17727608eab29c041265a525dff62b6d70703120bbc7318e70c13f69a6ff
3
+ size 1047628488
aft/charter/checkpoint-256/chat_template.jinja ADDED
@@ -0,0 +1,47 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {{ bos_token }}
2
+ {%- if messages[0]['role'] == 'system' -%}
3
+ {%- if messages[0]['content'] is string -%}
4
+ {%- set first_user_prefix = messages[0]['content'] + '
5
+
6
+ ' -%}
7
+ {%- else -%}
8
+ {%- set first_user_prefix = messages[0]['content'][0]['text'] + '
9
+
10
+ ' -%}
11
+ {%- endif -%}
12
+ {%- set loop_messages = messages[1:] -%}
13
+ {%- else -%}
14
+ {%- set first_user_prefix = "" -%}
15
+ {%- set loop_messages = messages -%}
16
+ {%- endif -%}
17
+ {%- for message in loop_messages -%}
18
+ {%- if (message['role'] == 'user') != (loop.index0 % 2 == 0) -%}
19
+ {{ raise_exception("Conversation roles must alternate user/assistant/user/assistant/...") }}
20
+ {%- endif -%}
21
+ {%- if (message['role'] == 'assistant') -%}
22
+ {%- set role = "model" -%}
23
+ {%- else -%}
24
+ {%- set role = message['role'] -%}
25
+ {%- endif -%}
26
+ {{ '<start_of_turn>' + role + '
27
+ ' + (first_user_prefix if loop.first else "") }}
28
+ {%- if message['content'] is string -%}
29
+ {{ message['content'] | trim }}
30
+ {%- elif message['content'] is iterable -%}
31
+ {%- for item in message['content'] -%}
32
+ {%- if item['type'] == 'image' -%}
33
+ {{ '<start_of_image>' }}
34
+ {%- elif item['type'] == 'text' -%}
35
+ {{ item['text'] | trim }}
36
+ {%- endif -%}
37
+ {%- endfor -%}
38
+ {%- else -%}
39
+ {{ raise_exception("Invalid content type") }}
40
+ {%- endif -%}
41
+ {{ '<end_of_turn>
42
+ ' }}
43
+ {%- endfor -%}
44
+ {%- if add_generation_prompt -%}
45
+ {{'<start_of_turn>model
46
+ '}}
47
+ {%- endif -%}
aft/charter/checkpoint-256/tokenizer.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726
3
+ size 33384567
aft/charter/checkpoint-256/tokenizer_config.json ADDED
@@ -0,0 +1,26 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "backend": "tokenizers",
3
+ "boi_token": "<start_of_image>",
4
+ "bos_token": "<bos>",
5
+ "clean_up_tokenization_spaces": false,
6
+ "eoi_token": "<end_of_image>",
7
+ "eos_token": "<eos>",
8
+ "image_token": "<image_soft_token>",
9
+ "is_local": false,
10
+ "local_files_only": false,
11
+ "mask_token": "<mask>",
12
+ "model_max_length": 131072,
13
+ "model_specific_special_tokens": {
14
+ "boi_token": "<start_of_image>",
15
+ "eoi_token": "<end_of_image>",
16
+ "image_token": "<image_soft_token>"
17
+ },
18
+ "pad_token": "<pad>",
19
+ "padding_side": "left",
20
+ "processor_class": "Gemma3Processor",
21
+ "sp_model_kwargs": null,
22
+ "spaces_between_special_tokens": false,
23
+ "tokenizer_class": "GemmaTokenizer",
24
+ "unk_token": "<unk>",
25
+ "use_default_system_prompt": false
26
+ }
aft/charter/checkpoint-256/tokens_state.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"total": 3865312, "trainable": 82760}
aft/charter/checkpoint-256/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
aft/charter/checkpoint-256/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:88de77a4e8e22463a86aad289b001bb15b32e47a0fb4f0f173b9b3be45a1a0b1
3
+ size 8337
aft/charter/checkpoint-32/README.md ADDED
@@ -0,0 +1,208 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ base_model: /workspace/hf-dispatch-midtrain-aft-v1/hub/models--jbostock--scimt-dispatch-sft-v1/snapshots/ad24276d9d25455b528c80b4c3043438bfc32ca5/runs/20260806T143703Z/charter/checkpoint-48
3
+ library_name: peft
4
+ pipeline_tag: text-generation
5
+ tags:
6
+ - axolotl
7
+ - base_model:adapter:/workspace/hf-dispatch-midtrain-aft-v1/hub/models--jbostock--scimt-dispatch-sft-v1/snapshots/ad24276d9d25455b528c80b4c3043438bfc32ca5/runs/20260806T143703Z/charter/checkpoint-48
8
+ - lora
9
+ - transformers
10
+ ---
11
+
12
+ # Model Card for Model ID
13
+
14
+ <!-- Provide a quick summary of what the model is/does. -->
15
+
16
+
17
+
18
+ ## Model Details
19
+
20
+ ### Model Description
21
+
22
+ <!-- Provide a longer summary of what this model is. -->
23
+
24
+
25
+
26
+ - **Developed by:** [More Information Needed]
27
+ - **Funded by [optional]:** [More Information Needed]
28
+ - **Shared by [optional]:** [More Information Needed]
29
+ - **Model type:** [More Information Needed]
30
+ - **Language(s) (NLP):** [More Information Needed]
31
+ - **License:** [More Information Needed]
32
+ - **Finetuned from model [optional]:** [More Information Needed]
33
+
34
+ ### Model Sources [optional]
35
+
36
+ <!-- Provide the basic links for the model. -->
37
+
38
+ - **Repository:** [More Information Needed]
39
+ - **Paper [optional]:** [More Information Needed]
40
+ - **Demo [optional]:** [More Information Needed]
41
+
42
+ ## Uses
43
+
44
+ <!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
45
+
46
+ ### Direct Use
47
+
48
+ <!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
49
+
50
+ [More Information Needed]
51
+
52
+ ### Downstream Use [optional]
53
+
54
+ <!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
55
+
56
+ [More Information Needed]
57
+
58
+ ### Out-of-Scope Use
59
+
60
+ <!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
61
+
62
+ [More Information Needed]
63
+
64
+ ## Bias, Risks, and Limitations
65
+
66
+ <!-- This section is meant to convey both technical and sociotechnical limitations. -->
67
+
68
+ [More Information Needed]
69
+
70
+ ### Recommendations
71
+
72
+ <!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
73
+
74
+ Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
75
+
76
+ ## How to Get Started with the Model
77
+
78
+ Use the code below to get started with the model.
79
+
80
+ [More Information Needed]
81
+
82
+ ## Training Details
83
+
84
+ ### Training Data
85
+
86
+ <!-- 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. -->
87
+
88
+ [More Information Needed]
89
+
90
+ ### Training Procedure
91
+
92
+ <!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
93
+
94
+ #### Preprocessing [optional]
95
+
96
+ [More Information Needed]
97
+
98
+
99
+ #### Training Hyperparameters
100
+
101
+ - **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
102
+
103
+ #### Speeds, Sizes, Times [optional]
104
+
105
+ <!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
106
+
107
+ [More Information Needed]
108
+
109
+ ## Evaluation
110
+
111
+ <!-- This section describes the evaluation protocols and provides the results. -->
112
+
113
+ ### Testing Data, Factors & Metrics
114
+
115
+ #### Testing Data
116
+
117
+ <!-- This should link to a Dataset Card if possible. -->
118
+
119
+ [More Information Needed]
120
+
121
+ #### Factors
122
+
123
+ <!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
124
+
125
+ [More Information Needed]
126
+
127
+ #### Metrics
128
+
129
+ <!-- These are the evaluation metrics being used, ideally with a description of why. -->
130
+
131
+ [More Information Needed]
132
+
133
+ ### Results
134
+
135
+ [More Information Needed]
136
+
137
+ #### Summary
138
+
139
+
140
+
141
+ ## Model Examination [optional]
142
+
143
+ <!-- Relevant interpretability work for the model goes here -->
144
+
145
+ [More Information Needed]
146
+
147
+ ## Environmental Impact
148
+
149
+ <!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
150
+
151
+ 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).
152
+
153
+ - **Hardware Type:** [More Information Needed]
154
+ - **Hours used:** [More Information Needed]
155
+ - **Cloud Provider:** [More Information Needed]
156
+ - **Compute Region:** [More Information Needed]
157
+ - **Carbon Emitted:** [More Information Needed]
158
+
159
+ ## Technical Specifications [optional]
160
+
161
+ ### Model Architecture and Objective
162
+
163
+ [More Information Needed]
164
+
165
+ ### Compute Infrastructure
166
+
167
+ [More Information Needed]
168
+
169
+ #### Hardware
170
+
171
+ [More Information Needed]
172
+
173
+ #### Software
174
+
175
+ [More Information Needed]
176
+
177
+ ## Citation [optional]
178
+
179
+ <!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
180
+
181
+ **BibTeX:**
182
+
183
+ [More Information Needed]
184
+
185
+ **APA:**
186
+
187
+ [More Information Needed]
188
+
189
+ ## Glossary [optional]
190
+
191
+ <!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
192
+
193
+ [More Information Needed]
194
+
195
+ ## More Information [optional]
196
+
197
+ [More Information Needed]
198
+
199
+ ## Model Card Authors [optional]
200
+
201
+ [More Information Needed]
202
+
203
+ ## Model Card Contact
204
+
205
+ [More Information Needed]
206
+ ### Framework versions
207
+
208
+ - PEFT 0.19.1
aft/charter/checkpoint-32/adapter_config.json ADDED
@@ -0,0 +1,189 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "alora_invocation_tokens": null,
3
+ "alpha_pattern": {},
4
+ "arrow_config": null,
5
+ "auto_mapping": null,
6
+ "base_model_name_or_path": "/workspace/hf-dispatch-midtrain-aft-v1/hub/models--jbostock--scimt-dispatch-sft-v1/snapshots/ad24276d9d25455b528c80b4c3043438bfc32ca5/runs/20260806T143703Z/charter/checkpoint-48",
7
+ "bias": "none",
8
+ "corda_config": null,
9
+ "ensure_weight_tying": false,
10
+ "eva_config": null,
11
+ "exclude_modules": null,
12
+ "fan_in_fan_out": null,
13
+ "inference_mode": true,
14
+ "init_lora_weights": true,
15
+ "layer_replication": null,
16
+ "layers_pattern": null,
17
+ "layers_to_transform": null,
18
+ "loftq_config": {},
19
+ "lora_alpha": 128,
20
+ "lora_bias": false,
21
+ "lora_dropout": 0.0,
22
+ "lora_ga_config": null,
23
+ "megatron_config": null,
24
+ "megatron_core": "megatron.core",
25
+ "modules_to_save": null,
26
+ "peft_type": "LORA",
27
+ "peft_version": "0.19.1",
28
+ "qalora_group_size": 16,
29
+ "r": 64,
30
+ "rank_pattern": {},
31
+ "revision": null,
32
+ "target_modules": [
33
+ "language_model.layers.25.self_attn.v_proj",
34
+ "30.self_attn.v_proj",
35
+ "35.self_attn.q_proj",
36
+ "27.self_attn.v_proj",
37
+ "language_model.layers.24.self_attn.q_proj",
38
+ "language_model.layers.8.self_attn.v_proj",
39
+ "41.self_attn.v_proj",
40
+ "language_model.layers.6.self_attn.k_proj",
41
+ "39.self_attn.q_proj",
42
+ "language_model.layers.21.self_attn.q_proj",
43
+ "language_model.layers.4.self_attn.q_proj",
44
+ "up_proj",
45
+ "31.self_attn.k_proj",
46
+ "language_model.layers.12.self_attn.q_proj",
47
+ "language_model.layers.14.self_attn.k_proj",
48
+ "language_model.layers.17.self_attn.q_proj",
49
+ "27.self_attn.q_proj",
50
+ "language_model.layers.17.self_attn.v_proj",
51
+ "language_model.layers.24.self_attn.k_proj",
52
+ "36.self_attn.q_proj",
53
+ "language_model.layers.14.self_attn.v_proj",
54
+ "38.self_attn.v_proj",
55
+ "45.self_attn.v_proj",
56
+ "language_model.layers.26.self_attn.q_proj",
57
+ "language_model.layers.3.self_attn.q_proj",
58
+ "language_model.layers.20.self_attn.k_proj",
59
+ "language_model.layers.5.self_attn.q_proj",
60
+ "language_model.layers.16.self_attn.k_proj",
61
+ "language_model.layers.22.self_attn.k_proj",
62
+ "language_model.layers.18.self_attn.q_proj",
63
+ "45.self_attn.q_proj",
64
+ "47.self_attn.q_proj",
65
+ "language_model.layers.0.self_attn.v_proj",
66
+ "language_model.layers.23.self_attn.k_proj",
67
+ "43.self_attn.k_proj",
68
+ "42.self_attn.v_proj",
69
+ "language_model.layers.18.self_attn.v_proj",
70
+ "30.self_attn.k_proj",
71
+ "language_model.layers.8.self_attn.q_proj",
72
+ "language_model.layers.21.self_attn.k_proj",
73
+ "39.self_attn.v_proj",
74
+ "language_model.layers.13.self_attn.k_proj",
75
+ "30.self_attn.q_proj",
76
+ "33.self_attn.q_proj",
77
+ "44.self_attn.v_proj",
78
+ "language_model.layers.4.self_attn.k_proj",
79
+ "language_model.layers.6.self_attn.q_proj",
80
+ "45.self_attn.k_proj",
81
+ "40.self_attn.k_proj",
82
+ "language_model.layers.11.self_attn.v_proj",
83
+ "47.self_attn.k_proj",
84
+ "language_model.layers.7.self_attn.v_proj",
85
+ "33.self_attn.k_proj",
86
+ "31.self_attn.v_proj",
87
+ "language_model.layers.3.self_attn.k_proj",
88
+ "29.self_attn.q_proj",
89
+ "32.self_attn.q_proj",
90
+ "33.self_attn.v_proj",
91
+ "34.self_attn.q_proj",
92
+ "language_model.layers.20.self_attn.q_proj",
93
+ "language_model.layers.19.self_attn.v_proj",
94
+ "language_model.layers.26.self_attn.v_proj",
95
+ "o_proj",
96
+ "language_model.layers.5.self_attn.v_proj",
97
+ "29.self_attn.v_proj",
98
+ "28.self_attn.k_proj",
99
+ "language_model.layers.2.self_attn.q_proj",
100
+ "down_proj",
101
+ "language_model.layers.11.self_attn.k_proj",
102
+ "44.self_attn.q_proj",
103
+ "31.self_attn.q_proj",
104
+ "language_model.layers.9.self_attn.q_proj",
105
+ "language_model.layers.11.self_attn.q_proj",
106
+ "language_model.layers.7.self_attn.q_proj",
107
+ "28.self_attn.q_proj",
108
+ "language_model.layers.18.self_attn.k_proj",
109
+ "language_model.layers.16.self_attn.q_proj",
110
+ "39.self_attn.k_proj",
111
+ "language_model.layers.7.self_attn.k_proj",
112
+ "36.self_attn.v_proj",
113
+ "language_model.layers.1.self_attn.v_proj",
114
+ "37.self_attn.q_proj",
115
+ "language_model.layers.0.self_attn.q_proj",
116
+ "language_model.layers.12.self_attn.k_proj",
117
+ "language_model.layers.9.self_attn.k_proj",
118
+ "language_model.layers.22.self_attn.q_proj",
119
+ "language_model.layers.19.self_attn.q_proj",
120
+ "language_model.layers.13.self_attn.v_proj",
121
+ "34.self_attn.v_proj",
122
+ "language_model.layers.6.self_attn.v_proj",
123
+ "35.self_attn.v_proj",
124
+ "37.self_attn.k_proj",
125
+ "language_model.layers.23.self_attn.v_proj",
126
+ "42.self_attn.k_proj",
127
+ "32.self_attn.v_proj",
128
+ "language_model.layers.1.self_attn.k_proj",
129
+ "language_model.layers.26.self_attn.k_proj",
130
+ "36.self_attn.k_proj",
131
+ "language_model.layers.17.self_attn.k_proj",
132
+ "language_model.layers.23.self_attn.q_proj",
133
+ "language_model.layers.1.self_attn.q_proj",
134
+ "language_model.layers.2.self_attn.k_proj",
135
+ "language_model.layers.9.self_attn.v_proj",
136
+ "46.self_attn.q_proj",
137
+ "language_model.layers.25.self_attn.q_proj",
138
+ "language_model.layers.15.self_attn.k_proj",
139
+ "47.self_attn.v_proj",
140
+ "27.self_attn.k_proj",
141
+ "38.self_attn.q_proj",
142
+ "language_model.layers.10.self_attn.k_proj",
143
+ "32.self_attn.k_proj",
144
+ "language_model.layers.5.self_attn.k_proj",
145
+ "46.self_attn.k_proj",
146
+ "language_model.layers.0.self_attn.k_proj",
147
+ "language_model.layers.4.self_attn.v_proj",
148
+ "language_model.layers.16.self_attn.v_proj",
149
+ "language_model.layers.25.self_attn.k_proj",
150
+ "41.self_attn.q_proj",
151
+ "language_model.layers.15.self_attn.v_proj",
152
+ "language_model.layers.14.self_attn.q_proj",
153
+ "language_model.layers.10.self_attn.q_proj",
154
+ "44.self_attn.k_proj",
155
+ "language_model.layers.12.self_attn.v_proj",
156
+ "35.self_attn.k_proj",
157
+ "41.self_attn.k_proj",
158
+ "language_model.layers.15.self_attn.q_proj",
159
+ "gate_proj",
160
+ "34.self_attn.k_proj",
161
+ "language_model.layers.24.self_attn.v_proj",
162
+ "language_model.layers.13.self_attn.q_proj",
163
+ "language_model.layers.10.self_attn.v_proj",
164
+ "29.self_attn.k_proj",
165
+ "28.self_attn.v_proj",
166
+ "language_model.layers.3.self_attn.v_proj",
167
+ "language_model.layers.2.self_attn.v_proj",
168
+ "43.self_attn.q_proj",
169
+ "language_model.layers.21.self_attn.v_proj",
170
+ "40.self_attn.q_proj",
171
+ "37.self_attn.v_proj",
172
+ "language_model.layers.22.self_attn.v_proj",
173
+ "42.self_attn.q_proj",
174
+ "43.self_attn.v_proj",
175
+ "40.self_attn.v_proj",
176
+ "language_model.layers.8.self_attn.k_proj",
177
+ "38.self_attn.k_proj",
178
+ "language_model.layers.20.self_attn.v_proj",
179
+ "46.self_attn.v_proj",
180
+ "language_model.layers.19.self_attn.k_proj"
181
+ ],
182
+ "target_parameters": [],
183
+ "task_type": "CAUSAL_LM",
184
+ "trainable_token_indices": null,
185
+ "use_bdlora": null,
186
+ "use_dora": false,
187
+ "use_qalora": false,
188
+ "use_rslora": false
189
+ }
aft/charter/checkpoint-32/adapter_model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:785457385f7b6237496906ee6813cf88b383272343ade7499cb24603580d9d39
3
+ size 1047628488
aft/charter/checkpoint-32/chat_template.jinja ADDED
@@ -0,0 +1,47 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {{ bos_token }}
2
+ {%- if messages[0]['role'] == 'system' -%}
3
+ {%- if messages[0]['content'] is string -%}
4
+ {%- set first_user_prefix = messages[0]['content'] + '
5
+
6
+ ' -%}
7
+ {%- else -%}
8
+ {%- set first_user_prefix = messages[0]['content'][0]['text'] + '
9
+
10
+ ' -%}
11
+ {%- endif -%}
12
+ {%- set loop_messages = messages[1:] -%}
13
+ {%- else -%}
14
+ {%- set first_user_prefix = "" -%}
15
+ {%- set loop_messages = messages -%}
16
+ {%- endif -%}
17
+ {%- for message in loop_messages -%}
18
+ {%- if (message['role'] == 'user') != (loop.index0 % 2 == 0) -%}
19
+ {{ raise_exception("Conversation roles must alternate user/assistant/user/assistant/...") }}
20
+ {%- endif -%}
21
+ {%- if (message['role'] == 'assistant') -%}
22
+ {%- set role = "model" -%}
23
+ {%- else -%}
24
+ {%- set role = message['role'] -%}
25
+ {%- endif -%}
26
+ {{ '<start_of_turn>' + role + '
27
+ ' + (first_user_prefix if loop.first else "") }}
28
+ {%- if message['content'] is string -%}
29
+ {{ message['content'] | trim }}
30
+ {%- elif message['content'] is iterable -%}
31
+ {%- for item in message['content'] -%}
32
+ {%- if item['type'] == 'image' -%}
33
+ {{ '<start_of_image>' }}
34
+ {%- elif item['type'] == 'text' -%}
35
+ {{ item['text'] | trim }}
36
+ {%- endif -%}
37
+ {%- endfor -%}
38
+ {%- else -%}
39
+ {{ raise_exception("Invalid content type") }}
40
+ {%- endif -%}
41
+ {{ '<end_of_turn>
42
+ ' }}
43
+ {%- endfor -%}
44
+ {%- if add_generation_prompt -%}
45
+ {{'<start_of_turn>model
46
+ '}}
47
+ {%- endif -%}