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 | 45.25 |
ENEM Challenge (No Images) | 31.77 |
BLUEX (No Images) | 24.20 |
OAB Exams | 27.84 |
Assin2 RTE | 69.51 |
Assin2 STS | 30.31 |
FaQuAD NLI | 55.55 |
HateBR Binary | 53.18 |
PT Hate Speech Binary | 64.74 |
tweetSentBR | 50.18 |
- Downloads last month
- 0
Unable to determine this modelโs pipeline type. Check the
docs
.
Space using recogna-nlp/gembode-2b-base-ultraalpaca 1
Evaluation results
- accuracy on ENEM Challenge (No Images)Open Portuguese LLM Leaderboard31.770
- accuracy on BLUEX (No Images)Open Portuguese LLM Leaderboard24.200
- accuracy on OAB ExamsOpen Portuguese LLM Leaderboard27.840
- f1-macro on Assin2 RTEtest set Open Portuguese LLM Leaderboard69.510
- pearson on Assin2 STStest set Open Portuguese LLM Leaderboard30.310
- f1-macro on FaQuAD NLItest set Open Portuguese LLM Leaderboard55.550
- f1-macro on HateBR Binarytest set Open Portuguese LLM Leaderboard53.180
- f1-macro on PT Hate Speech Binarytest set Open Portuguese LLM Leaderboard64.740
- f1-macro on tweetSentBRtest set Open Portuguese LLM Leaderboard50.180