Downloads · 30 days
146
8% of all-time downloads
OpenRLHF_-_Llama-3-8b-rlhf-100k-gguf
RichardErkhov/OpenRLHF_-_Llama-3-8b-rlhf-100k-gguf
OpenRLHF_-_Llama-3-8b-rlhf-100k-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.
All-time downloads
1.8K
Public
Repo size
92.9 GB
Likes
0
Public
Repo makeup
Click a slice to open those files.
.gguf108 GB · 100%
At a glance
- Access
- Public
- Created
- Aug 22, 2024
- Updated
- Aug 22, 2024
- SHA
- cbfa7f5c
Model card
From the Hugging Face model README
Quantization made by Richard Erkhov.
Llama-3-8b-rlhf-100k - GGUF
- Model creator: https://huggingface.co/OpenRLHF/
- Original model: https://huggingface.co/OpenRLHF/Llama-3-8b-rlhf-100k/
| Name | Quant method | Size |
|---|---|---|
| Llama-3-8b-rlhf-100k.Q2_K.gguf | Q2_K | 2.96GB |
| Llama-3-8b-rlhf-100k.IQ3_XS.gguf | IQ3_XS | 3.28GB |
| Llama-3-8b-rlhf-100k.IQ3_S.gguf | IQ3_S | 3.43GB |
| Llama-3-8b-rlhf-100k.Q3_K_S.gguf | Q3_K_S | 3.41GB |
| Llama-3-8b-rlhf-100k.IQ3_M.gguf | IQ3_M | 3.52GB |
| Llama-3-8b-rlhf-100k.Q3_K.gguf | Q3_K | 3.74GB |
| Llama-3-8b-rlhf-100k.Q3_K_M.gguf | Q3_K_M | 3.74GB |
| Llama-3-8b-rlhf-100k.Q3_K_L.gguf | Q3_K_L | 4.03GB |
| Llama-3-8b-rlhf-100k.IQ4_XS.gguf | IQ4_XS | 4.18GB |
| Llama-3-8b-rlhf-100k.Q4_0.gguf | Q4_0 | 4.34GB |
| Llama-3-8b-rlhf-100k.IQ4_NL.gguf | IQ4_NL | 4.38GB |
| Llama-3-8b-rlhf-100k.Q4_K_S.gguf | Q4_K_S | 4.37GB |
| Llama-3-8b-rlhf-100k.Q4_K.gguf | Q4_K | 4.58GB |
| Llama-3-8b-rlhf-100k.Q4_K_M.gguf | Q4_K_M | 4.58GB |
| Llama-3-8b-rlhf-100k.Q4_1.gguf | Q4_1 | 4.78GB |
| Llama-3-8b-rlhf-100k.Q5_0.gguf | Q5_0 | 5.21GB |
| Llama-3-8b-rlhf-100k.Q5_K_S.gguf | Q5_K_S | 5.21GB |
| Llama-3-8b-rlhf-100k.Q5_K.gguf | Q5_K | 5.34GB |
| Llama-3-8b-rlhf-100k.Q5_K_M.gguf | Q5_K_M | 5.34GB |
| Llama-3-8b-rlhf-100k.Q5_1.gguf | Q5_1 | 5.65GB |
| Llama-3-8b-rlhf-100k.Q6_K.gguf | Q6_K | 6.14GB |
| Llama-3-8b-rlhf-100k.Q8_0.gguf | Q8_0 | 7.95GB |
Original model description: Llama-3 8B RLHF checkpoint trained by OpenRLHF
Using the models and datasets:
- Base SFT model: https://huggingface.co/OpenLLMAI/Llama-3-8b-sft-mixture
- Reward model: https://huggingface.co/OpenLLMAI/Llama-3-8b-rm-mixture
- Prompt dataset: https://huggingface.co/datasets/OpenLLMAI/prompt-collection-v0.1
Training Hyperparameters
Actor Learning Rate: 5e-7
Critic Learning Rate: 9e-6
Learning Rate Scheduler: Cosine with 0.03 Warmup
PPO epoch: 1
Training Batch Size: 128
Experience Buffer Size: 1024
Reward Normalization: True
Max Prompt Length: 2048
Max Response Length: 2048
Max Samples: 100k (To save GPU resources)
Number of Samples per Prompt: 1
Evaluation
Chat-Arena-Hard
-------------------------------------------
llama-3-8b-sft | score: 5.6
llama-3-8b-rlhf-100k | score: 20.5
Training logs
<img src="https://cdn-uploads.huggingface.co/production/uploads/63f6c04ac96958470d1e9043/iqwD8jBAX1vhu0PT0ycy8.png" width="800px">- Created
- Aug 22, 2024
- Updated
- Aug 22, 2024