PE

Pearl 34B ties i1

PearlGGUF by mradermacherUpdated 2024-05-06
Download on Hugging Face
Parameters34.4B
Context195K
Layers60
Architecturellama
Smallest quant8.1 GB
Quick answer

Pearl 34B ties i1 needs about 23.7 GB of memory at Q4_K_M (a 21.2 GB file) with an 8K context window, so it fits on a 24 GB GPU. This repository has 21 quants, from 8.1 GB (IQ1_S) to 28.8 GB (Q6_K). Longer context needs more: at 32K tokens the KV cache alone adds about 8.1 GB.

Pearl‑34B‑ties‑i1: 34.4B‑parameter LLaMA model in GGUF format

Pearl‑34B‑ties‑i1 is a 34.4B‑parameter LLaMA model quantized for GGUF. It is a merge of task‑specific models using TIES‑Merging. The model is released under the apache‑2.0 license.

Good at

  • Large‑context reasoning
  • Multitask performance
  • High‑parameter density

Limits

  • No vision support
  • Requires 8k context for estimates
  • No active‑parameter count

Best for

  • Research prototyping
  • Fine‑tuning on custom data
  • Benchmarking on GPUs

Who it’s for. Researchers and developers who need a large‑parameter LLaMA model in GGUF format. The model is suitable for those who can manage the memory requirements and are comfortable with the apache‑2.0 license.

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
IQ1_S8.1 GB 10.6 GB
Too big Get file →
IQ1_M8.8 GB 11.3 GB
Too big Get file →
IQ2_XXS9.9 GB 12.4 GB
Too big Get file →
IQ2_XS10.9 GB 13.4 GB
Too big Get file →
IQ2_S11.5 GB 14.0 GB
Too big Get file →
IQ2_M12.4 GB 14.9 GB
Too big Get file →
Q2_K13.4 GB 15.9 GB
Too big Get file →
IQ3_XXS13.9 GB 16.4 GB
Too big Get file →
IQ3_XS14.8 GB 17.3 GB
Too big Get file →
Q3_K_S15.5 GB 18.0 GB
Too big Get file →
IQ3_S15.6 GB 18.1 GB
Too big Get file →
IQ3_M16.1 GB 18.6 GB
Too big Get file →
Q3_K_M17.2 GB 19.7 GB
Too big Get file →
Q3_K_L18.7 GB 21.2 GB
Too big Get file →
IQ4_XS19.0 GB 21.5 GB
Too big Get file →
Q4_020.1 GB 22.6 GB
Too big Get file →
Q4_K_S20.1 GB 22.7 GB
Too big Get file →
Q4_K_M21.2 GB 23.7 GB
Too big Get file →
Q5_K_S24.2 GB 26.8 GB
Too big Get file →
Q5_K_M24.9 GB 27.4 GB
Too big Get file →
Q6_K28.8 GB 31.3 GB
Too big Get file →

For 8 GB VRAM, use IQ1_S (8.1 GB file, 10.6 GB memory at 8k context). For 12 GB VRAM, IQ2_S (11.49 GB file, 14 GB memory). For 16 GB VRAM, Q3_K_M (17.2 GB file, 19.7 GB memory). Estimates are approximate.

What GPU do you need for Pearl 34B ties i1?

These figures add the file size, an fp16 KV cache at an 8K context window (worked out from this model’s own header) and about 0.5 GB for the runtime. They are estimates, so leave a little headroom, and expect to use less context or a smaller quant if you are right at the limit.

  • 4 GB VRAM: no quant fits at an 8K context.
  • 6 GB VRAM: no quant fits at an 8K context.
  • 8 GB VRAM: no quant fits at an 8K context.
  • 12 GB VRAM: the best quant that fits is IQ1_M (8.8 GB file, about 11.3 GB in use).
  • 16 GB VRAM: the best quant that fits is Q2_K (13.4 GB file, about 15.9 GB in use).
  • 24 GB VRAM: the best quant that fits is Q4_K_M (21.2 GB file, about 23.7 GB in use).
  • 32 GB VRAM: the best quant that fits is Q6_K (28.8 GB file, about 31.3 GB in use).

Which Pearl 34B ties i1 quant should you choose?

Q4_K_M is the usual starting point for Pearl 34B ties i1: it balances file size and quality well. Here it is a 21.2 GB file.

If memory is tight, the smallest file is IQ1_S at 8.1 GB, with the quality trade-offs that come with fewer bits. Use it only when memory is very tight. Expect a clear drop in quality, which is usually worse on small models and easier to live with on large ones.

New to quant names? Read GGUF quantization explained.

Specifications

Architecture from the GGUF header

Architecturellama
Parameters34.4B
Layers60
Trained context200,000 tokens
KV cache (fp16)240 KB per token

Source

Quantized bymradermacher
FamilyPearl
Licenseapache-2.0
Downloads (HF)971
Last updated2024-05-06

Frequently asked questions about Pearl 34B ties i1

How much VRAM does Pearl 34B ties i1 need?

About 23.7 GB at Q4_K_M with an 8K context window. That is the 21.2 GB file, plus the KV cache and roughly 0.5 GB of runtime overhead. Smaller quants need less and longer context needs more.

Can I run Pearl 34B ties i1 on an 8 GB GPU?

Not at an 8K context. The smallest file needs about 10.6 GB, so you would need a larger GPU or CPU offloading.

What is the file size of Pearl 34B ties i1 GGUF?

It depends on the quant: IQ1_S is 8.1 GB, IQ1_M is 8.8 GB, IQ2_XXS is 9.9 GB, IQ2_XS is 10.9 GB, IQ2_S is 11.5 GB, IQ2_M is 12.4 GB and 15 more.

What context length does Pearl 34B ties i1 support?

The GGUF header lists a trained context length of 200,000 tokens (195K). The KV cache grows with the context you use: about 2.0 GB at 8K and 8.1 GB at 32K.

Where can I download Pearl 34B ties i1 GGUF?

The files are on Hugging Face in the mradermacher/Pearl-34B-ties-i1-GGUF repository. The download buttons on this page open the file on Hugging Face. ggufmodels.com does not host any model files.

What license does Pearl 34B ties i1 use?

Hugging Face lists the license as apache-2.0. Licenses can differ between the quantized files and the original model, so check the repository and the base model before commercial use.

Can I run this model on 8 GB VRAM?

Use the IQ1_S quant, which requires about 10.6 GB memory at 8k context. It may fit with lower context or batching.

Does the model support vision tasks?

No, the model has no vision capabilities.

What license applies to this model?

The model is released under the apache‑2.0 license.

Other GGUFs of Pearl-34B-ties

Similar models

Browse all →

More from mradermacher

All by mradermacher →

More Pearl models

All Pearl →

Other large models

All large models →