28 lines
956 B
JSON
28 lines
956 B
JSON
|
|
{
|
||
|
|
"MODEL_NAME": "unsloth/Phi-4-unsloth-bnb-4bit",
|
||
|
|
"CHAT_TEMPLATE": "phi-4",
|
||
|
|
"MAX_SEQ_LENGTH": 2048,
|
||
|
|
"LOAD_IN_4BIT": true,
|
||
|
|
"DATASET_NAME": "Mathieu-Thomas-JOSSET/michael_abab_conversations_infini_instruct.jsonl",
|
||
|
|
"DATASET_SPLIT": "train",
|
||
|
|
"PER_DEVICE_TRAIN_BATCH_SIZE": 2,
|
||
|
|
"GRADIENT_ACCUMULATION_STEPS": 4,
|
||
|
|
"WARMUP_STEPS": 10,
|
||
|
|
"MAX_STEPS": 2000,
|
||
|
|
"LEARNING_RATE": 9.95267419777795e-06,
|
||
|
|
"LR_AUTO_ENABLED": true,
|
||
|
|
"LR_AUTO_USE_N": "train",
|
||
|
|
"LR_AUTO_N_REF": 1436,
|
||
|
|
"LR_AUTO_BASE": 1e-05,
|
||
|
|
"LR_AUTO_MULT": 0.5,
|
||
|
|
"LR_AUTO_FINAL": 5e-06,
|
||
|
|
"WEIGHT_DECAY": 0.009206070410847844,
|
||
|
|
"LR_SCHEDULER_TYPE": "linear",
|
||
|
|
"SEED": 3407,
|
||
|
|
"PLOTLY_DARK_MODE": true,
|
||
|
|
"PLOTLY_BASE_COLOR": "#00CC96",
|
||
|
|
"PLOTLY_EMA_SPAN": 25,
|
||
|
|
"HF_REPO_ID_MERGED_RESOLVED": "Mathieu-Thomas-JOSSET/joke-20260112-081758",
|
||
|
|
"HF_REPO_ID_GGUF_RESOLVED": "Mathieu-Thomas-JOSSET/joke-finetome-model-gguf-phi4-20260112-081758",
|
||
|
|
"HF_ARTIFACTS_DIR_IN_REPO": "reports"
|
||
|
|
}
|