You can deploy this model and start earning money today!

mhla/
gpt1900-d34-coherence-rl

Name: gpt1900-d34-coherence-rl
Availability: InStock

Views

No views yet

GPT-1900 D34 COHERENCE RL

3.29B parameter GPT-1900 model with RL post-training. Pipeline: base pretrain -> SFT (period) -> SFT (modern) -> RL.

Model Details

Architecture: Custom GPT with RoPE, QK-norm, ReLU², value embeddings (ResFormer), per-layer residual/skip scalars
Parameters: 3.29B
Layers: 34
Hidden dim: 2176
Attention heads: 17 (query) / 17 (kv)
Head dim: 128
Context length: 2048 tokens
Vocab size: 32,768 (BPE, GPT-4 style split pattern)
Training: RL training on top of two-stage SFT model (780 steps)

Checkpoint Contents

model_000780.pt          # Model weights
meta_000780.json         # Training config and metadata
tokenizer/               # BPE tokenizer (tiktoken format) + token byte counts
nanochat/                # Source code to load and run the model

Quick Start

python

import torch, json
from nanochat.gpt import GPT, GPTConfig
from nanochat.tokenizer import RustBPETokenizer

tokenizer = RustBPETokenizer.from_directory("tokenizer")

with open("meta_000780.json") as f:
    meta = json.load(f)

config = GPTConfig(**meta["model_config"])

with torch.device("meta"):
    model = GPT(config)
model.to_empty(device="cuda")
model.init_weights()

state_dict = torch.load("model_000780.pt", map_location="cuda")
state_dict = {k.removeprefix("_orig_mod."): v for k, v in state_dict.items()}
model.load_state_dict(state_dict, strict=True, assign=True)
model.eval()

bos = tokenizer.get_bos_token_id()
tokens = tokenizer.encode("It was a dark and stormy night", prepend=bos)
with torch.amp.autocast(device_type="cuda", dtype=torch.bfloat16):
    for token in model.generate(tokens, max_tokens=100, temperature=0.8):
        print(tokenizer.decode([token]), end="", flush=True)

Dependencies

torch>=2.9
tiktoken
rustbpe

Model Family

mhla/gpt1900-d34-22btok - Base pretrained model
mhla/gpt1900-d34-sft-period - SFT (period style)
mhla/gpt1900-d34-sft-modern - SFT (modern style)
mhla/gpt1900-d34-rl - RL post-training

mhla/gpt1900-d34-coherence-rl