Downloads · 30 days
17
13% of all-time downloads
laion/delphi-1e23-wc50m-warmup-levanter-sft
delphi-1e23-wc50m-warmup-levanter-sft is a text generation model from laion. Use it when you need the model to write or continue text. It is set up for transformers. The card lists the license as apache-2.0.
Math-weak supervised fine-tune (SFT) of the Delphi 1e23 (~25B-param) partial-K midtrain, trained with Levanter and exported to Hugging Face format.
Downloads · 30 days
17
13% of all-time downloads
All-time downloads
128
Public
Parameters
25B
99.9 GB on disk
Likes
0
Public
Click a slice to open those files.
.safetensors99.9 GB · 100%
From the Hugging Face model README
Math-weak supervised fine-tune (SFT) of the Delphi 1e23 (~25B-param) partial-K midtrain, trained with Levanter and exported to Hugging Face format.
This is the wc50m (wildchat) arm of the Delphi 1e23 post-training ladder: a general-chat SFT on a
wildchat50m (90%) + delphi_warmup (10%) mixture, starting from the prepared 1e23 midtrained base.
Contrast it with the math-strong laion/delphi-1e23-magpie-warmup-levanter-sft (magpie 90 / warmup 10).
Qwen3ForCausalLM (dense), vocab 128256, 51 layers, hidden 5376,
42 attention / 42 KV heads, head_dim 128, max_position_embeddings 4096.rope_theta = 500000,
rope_scaling = {rope_type: llama3, factor: 8.0, low_freq_factor: 1.0, high_freq_factor: 4.0, original_max_position_embeddings: 8192} (mirrors the base; carried at both the top level and under
rope_parameters for transformers-4/5 compatibility).laion/delphi-1e23-p33m67-k0p076-prepared
(prepared from the midtrained laion/delphi-1e23-p33m67-k0p076-lr0.67-b6607e).wildchat50m 90% / delphi_warmup 10%.This model uses the Delphi v0 chat protocol: Llama-3 headers
(<|start_header_id|>/<|end_header_id|>/<|eot_id|>) plus repurposed single-id think/tool tokens:
| token | id |
|---|---|
| `< | start_think |
| `< | end_think |
| `< | tool_call |
| `< | tool_call_end |
| `< | tool_result |
| `< | tool_result_end |
⚠️ It is not ChatML — <|im_start|>/<|im_end|> are absent from this tokenizer. The Delphi v0
Jinja template is not baked into tokenizer_config.json; apply it externally (e.g. pass it to the
serving engine). Reference implementation:
experiments/sft/delphi_chat_template.py
(the served template is that string with the Levanter {% generation %}/{% endgeneration %} markers
removed). When decoding, keep skip_special_tokens=False so the atomic 128002/128003 think delimiters
survive.
Delphi 1e23 post-training study (issue #6279). HF-distribution only.