Instructions to use ISTA-DASLab/Llama-3.1-70B-Instruct-2Bit-GSQ with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use ISTA-DASLab/Llama-3.1-70B-Instruct-2Bit-GSQ with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="ISTA-DASLab/Llama-3.1-70B-Instruct-2Bit-GSQ")

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("ISTA-DASLab/Llama-3.1-70B-Instruct-2Bit-GSQ")
model = AutoModelForCausalLM.from_pretrained("ISTA-DASLab/Llama-3.1-70B-Instruct-2Bit-GSQ")

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use ISTA-DASLab/Llama-3.1-70B-Instruct-2Bit-GSQ with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "ISTA-DASLab/Llama-3.1-70B-Instruct-2Bit-GSQ"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "ISTA-DASLab/Llama-3.1-70B-Instruct-2Bit-GSQ",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/ISTA-DASLab/Llama-3.1-70B-Instruct-2Bit-GSQ

SGLang

How to use ISTA-DASLab/Llama-3.1-70B-Instruct-2Bit-GSQ with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "ISTA-DASLab/Llama-3.1-70B-Instruct-2Bit-GSQ" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "ISTA-DASLab/Llama-3.1-70B-Instruct-2Bit-GSQ",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "ISTA-DASLab/Llama-3.1-70B-Instruct-2Bit-GSQ" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "ISTA-DASLab/Llama-3.1-70B-Instruct-2Bit-GSQ",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use ISTA-DASLab/Llama-3.1-70B-Instruct-2Bit-GSQ with Docker Model Runner:
```
docker model run hf.co/ISTA-DASLab/Llama-3.1-70B-Instruct-2Bit-GSQ
```

soroushtabesh commited on 13 days ago

Commit

5893d4a

verified ·

1 Parent(s): 762bccb

Add storage layout details

Browse files

Files changed (1) hide show

README.md +47 -2

README.md CHANGED Viewed

@@ -12,7 +12,7 @@ tags:
   - llama
   - llama-3.1
   - vllm
-  - compressed-tensors
   - arxiv:2604.18556
 ---
@@ -46,9 +46,54 @@ standard zero-shot suite (ARC-C/E, HellaSwag, PIQA, Winogrande):
 - **Bits / weight (effective):** ≈2.13 bpp
 - **Codebook:** 2-bit symmetric scalar `{-2, -1, 0, +1} × scale`
 - **Group size:** 128
-- **Format:** `compressed-tensors` (auto-detected by vLLM)
 - **Pipeline:** GPTQ initialization → Gumbel-Softmax refinement (Lion optimizer)
 ## Serving with vLLM
 ```bash

   - llama
   - llama-3.1
   - vllm
+  - humming
   - arxiv:2604.18556
 ---
 - **Bits / weight (effective):** ≈2.13 bpp
 - **Codebook:** 2-bit symmetric scalar `{-2, -1, 0, +1} × scale`
 - **Group size:** 128
+- **Format:** [Humming](https://github.com/IST-DASLab/humming) (`quant_method: "humming"`, `b_dtype: "uint2"`)
 - **Pipeline:** GPTQ initialization → Gumbel-Softmax refinement (Lion optimizer)
+### Storage layout (why the HF UI shows I32 + BF16)
+The Hugging Face "Tensor types" widget reports the **container dtype** of each
+safetensor on disk, not the effective precision of the underlying weights.
+This checkpoint uses the **Humming** on-disk layout (exact-width packing — no
+sub-byte values are padded into a wider container). For every quantized
+`Linear` layer with original weight shape `[out_features, in_features]`, the
+following tensors are stored:
+| Tensor                       | Dtype | Shape on disk                       | Meaning                                                                       |
+|------------------------------|-------|-------------------------------------|-------------------------------------------------------------------------------|
+| `<layer>.weight`             | I32   | `[out_features, in_features × 2 / 32]` = `[out_features, in_features / 16]` | 2-bit values bit-packed along the input dim, LSB-first: 16 weights per INT32 word. |
+| `<layer>.weight_scale`       | BF16  | `[out_features, in_features / 128]` | One symmetric scale per group of `group_size = 128` weights along the input dim. |
+| Attention / norms / embed / LM-head | BF16 | unchanged                  | Not quantized; copied from the base checkpoint.                                |
+**Example** (`model.layers.0.mlp.gate_proj`, original `[28672, 8192]`):
+`weight` = `[28672, 512]` I32 (since `8192 × 2 / 32 = 512`),
+`weight_scale` = `[28672, 64]` BF16 (since `8192 / 128 = 64`).
+So although the UI says "I32 + BF16", the **effective storage** per quantized
+weight is `2 bits (packed) + 16 bits / 128 (group scale) ≈ 2.13 bpp`. The
+`quantization_config` block in `config.json` is:
+```json
+{
+  "quant_method": "humming",
+  "b_dtype": "uint2",
+  "weight_scale_group_size": 128,
+  "weight_scale_type": "group",
+  "has_zero_point": false,
+  "ignore": ["lm_head", "embed_tokens"]
+}
+```
+Loading this checkpoint requires a vLLM build with the
+[`humming`](https://github.com/IST-DASLab/humming) MoE kernel installed (see
+the [GSQ repo](https://github.com/IST-DASLab/GSQ) `scripts/setup_env.sh` for
+the exact install line).
+> Note: GSQ training first writes shards in `compressed-tensors`
+> `pack-quantized` format (where a sub-4-bit codebook is padded into a 4-bit
+> INT32 container). The published checkpoint here has been re-packed via
+> `convert_to_humming.py` into exact-width 2-bit Humming storage, hence the
+> `2 / 32` shape factor you see above.
 ## Serving with vLLM
 ```bash