Downloads · 30 days
58
3% of all-time downloads
RichardErkhov/chargoddard_-_platypus-2-22b-relora-gguf
chargoddard_-_platypus-2-22b-relora-gguf is a machine learning model from RichardErkhov. Use it for the machine learning task on the model card, and read the license before you ship it in a product.
Downloads · 30 days
58
3% of all-time downloads
All-time downloads
1.9K
Public
Repo size
240 GB
Likes
0
Public
Click a slice to open those files.
.gguf268 GB · 100%
From the Hugging Face model README
Quantization made by Richard Erkhov.
platypus-2-22b-relora - GGUF
| Name | Quant method | Size |
|---|---|---|
| platypus-2-22b-relora.Q2_K.gguf | Q2_K | 7.56GB |
| platypus-2-22b-relora.IQ3_XS.gguf | IQ3_XS | 8.35GB |
| platypus-2-22b-relora.IQ3_S.gguf | IQ3_S | 8.82GB |
| platypus-2-22b-relora.Q3_K_S.gguf | Q3_K_S | 8.82GB |
| platypus-2-22b-relora.IQ3_M.gguf | IQ3_M | 9.33GB |
| platypus-2-22b-relora.Q3_K.gguf | Q3_K | 4.56GB |
| platypus-2-22b-relora.Q3_K_M.gguf | Q3_K_M | 9.88GB |
| platypus-2-22b-relora.Q3_K_L.gguf | Q3_K_L | 10.81GB |
| platypus-2-22b-relora.IQ4_XS.gguf | IQ4_XS | 1.17GB |
| platypus-2-22b-relora.Q4_0.gguf | Q4_0 | 11.49GB |
| platypus-2-22b-relora.IQ4_NL.gguf | IQ4_NL | 11.56GB |
| platypus-2-22b-relora.Q4_K_S.gguf | Q4_K_S | 11.58GB |
| platypus-2-22b-relora.Q4_K.gguf | Q4_K | 12.27GB |
| platypus-2-22b-relora.Q4_K_M.gguf | Q4_K_M | 12.27GB |
| platypus-2-22b-relora.Q4_1.gguf | Q4_1 | 11.12GB |
| platypus-2-22b-relora.Q5_0.gguf | Q5_0 | 14.0GB |
| platypus-2-22b-relora.Q5_K_S.gguf | Q5_K_S | 14.0GB |
| platypus-2-22b-relora.Q5_K.gguf | Q5_K | 14.41GB |
| platypus-2-22b-relora.Q5_K_M.gguf | Q5_K_M | 14.41GB |
| platypus-2-22b-relora.Q5_1.gguf | Q5_1 | 15.26GB |
| platypus-2-22b-relora.Q6_K.gguf | Q6_K | 16.68GB |
| platypus-2-22b-relora.Q8_0.gguf | Q8_0 | 21.6GB |
datasets:
Experimental ReLoRA-trained model using the OpenPlatypus dataset. Ran for one epoch, with three lora restarts.
Not recommended for use yet. Mostly tossing this up for testing.
Base model was llama2-22b-blocktriangular.
Relevant training parameters:
adapter: qlora
load_in_4bit: true
lora_r: 32
lora_alpha: 16
lora_dropout: 0.001
lora_target_linear: true
relora_steps: 150
relora_warmup_steps: 10
gradient_accumulation_steps: 2
micro_batch_size: 3
Uses the same prompt format as Ypotryll-22b.
Prefix messages with " ***System:", " ***Query:", or " ***Response:", paying attention to whitespace.
Detailed results can be found here
| Metric | Value |
|---|---|
| Avg. | 52.21 |
| ARC (25-shot) | 57.68 |
| HellaSwag (10-shot) | 82.44 |
| MMLU (5-shot) | 55.33 |
| TruthfulQA (0-shot) | 43.61 |
| Winogrande (5-shot) | 77.35 |
| GSM8K (5-shot) | 6.6 |
| DROP (3-shot) | 42.46 |