Training procedure
The following bitsandbytes
quantization config was used during training:
- quant_method: bitsandbytes
- _load_in_8bit: False
- _load_in_4bit: True
- llm_int8_threshold: 6.0
- llm_int8_skip_modules: None
- llm_int8_enable_fp32_cpu_offload: False
- llm_int8_has_fp16_weight: False
- bnb_4bit_quant_type: nf4
- bnb_4bit_use_double_quant: True
- bnb_4bit_compute_dtype: float16
- bnb_4bit_quant_storage: uint8
- load_in_4bit: True
- load_in_8bit: False
Framework versions
- PEFT 0.5.0
Open Portuguese LLM Leaderboard Evaluation Results
Detailed results can be found here and on the ๐ Open Portuguese LLM Leaderboard
Metric | Value |
---|---|
Average | 33.68 |
ENEM Challenge (No Images) | 31.21 |
BLUEX (No Images) | 26.01 |
OAB Exams | 26.20 |
Assin2 RTE | 40.52 |
Assin2 STS | 4.64 |
FaQuAD NLI | 32.15 |
HateBR Binary | 60.10 |
PT Hate Speech Binary | 54.14 |
tweetSentBR | 28.18 |
- Downloads last month
- 4
Space using recogna-nlp/qwenbode_1_8b_chat_ultraalpaca_qlora 1
Evaluation results
- accuracy on ENEM Challenge (No Images)Open Portuguese LLM Leaderboard31.210
- accuracy on BLUEX (No Images)Open Portuguese LLM Leaderboard26.010
- accuracy on OAB ExamsOpen Portuguese LLM Leaderboard26.200
- f1-macro on Assin2 RTEtest set Open Portuguese LLM Leaderboard40.520
- pearson on Assin2 STStest set Open Portuguese LLM Leaderboard4.640
- f1-macro on FaQuAD NLItest set Open Portuguese LLM Leaderboard32.150
- f1-macro on HateBR Binarytest set Open Portuguese LLM Leaderboard60.100
- f1-macro on PT Hate Speech Binarytest set Open Portuguese LLM Leaderboard54.140
- f1-macro on tweetSentBRtest set Open Portuguese LLM Leaderboard28.180