GLM-5.2-0.8B-A0.8B-MXFP4xFP8_BLOCK

GLM-5.2-0.8B-A0.8B

This is a tiny version of zai-org/GLM-5.2 created for testing and development.

Model Details

• Base Model: zai-org/GLM-5.2
• Architecture: glm_moe_dsa (GLM MoE with DeepSeek Sparse Attention)
• Total Parameters: 0.85B
• Activated Parameters: ~0.77B

Configuration Changes

The following parameters were reduced from the original model:

Checkpoint Structure

Single safetensors file containing 194 tensors in float32. Layers 0-1 have dense MLP, layers 2-5 have MoE MLP. Layers 0-2 have full DSA indexer weights, layers 3-5 use shared indexer.

Usage

from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained("inference-optimization/GLM-5.2-0.8B-A0.8B", device_map="auto")
tokenizer = AutoTokenizer.from_pretrained("inference-optimization/GLM-5.2-0.8B-A0.8B")

input_ids = tokenizer("According to all known laws", return_tensors="pt").input_ids.to(model.device)
output = model.generate(input_ids, max_new_tokens=20)
print(tokenizer.decode(output[0]))

Creation Process

1. Inspected original GLM-5.2 config (78 layers, 256 experts, hidden_size=6144)
2. Reduced all dimensions to target ~1B parameters while preserving architecture
3. Created model with float32 dtype for training stability
4. Fine-tuned on copypasta dataset to perplexity ~1.0
5. Validated checkpoint structure matches original model naming conventions
6. Validated model loads, inferences, and generates correctly

Validation Output

Success: 1.0000379085540771 <= 10.0
Generating sample text:
According to all known laws of aviation, there is no way a bee should be able to fly.

Notes

• The model uses float32 dtype (original uses bfloat16) to ensure proper initialization and training of the tiny model
• Architecture preserves both dense and sparse MLP layer types, MLA attention with compressed Q/KV, and DSA indexer with full/shared patterns
• The model has been fine-tuned on a toy dataset and is intended for testing purposes only