Model · stablelm-tuned-alpha-7b
StableLM-Tuned-Alpha-7B
Developer: Stability AI
Availability: available · checked 2026-09-24 · source
Raw record: /data/models/stablelm-tuned-alpha-7b.json
Fields
- id
- stablelm-tuned-alpha-7b
- identifiers
- huggingface
- stabilityai/stablelm-tuned-alpha-7b
- developer
- Stability AI
- release_date
- 2023-04-20recorded · sourcenote: StableLM README changelog: '*April 20, 2023* Released initial set of StableLM-Alpha models, with 3B and 7B parameters', with the StableLM-Tuned-Alpha-7B demo linked in the same entry. The HF repo was created earlier (2023-04-19 per ingest).
- weights_status
- open
- availability
- available · checked 2026-09-24 · source
- license
- cc-by-nc-sa-4.0recorded · sourcenote: Card: 'Fine-tuned checkpoints (StableLM-Tuned-Alpha) are licensed under the Non-Commercial Creative Commons license (CC BY-NC-SA-4.0), in-line with the original non-commercial license specified by Stanford Alpaca.'
- architecture
- family
- decoder_only
- note
- family read from config 'architectures': ['GPTNeoXForCausalLM'].
- n_layers
- 16recorded · source
- hidden_size
- 6144recorded · source
- n_heads
- 48recorded · source
- vocab_size
- 50432recorded · source
- context_length
- 4096recorded · source
- positional_encoding
- nullnot_recordednote: Card says only 'based on the NeoX transformer architecture'; the config's rotary hints are not evidence. Also not_recorded on the base record.
- training_data
- Supervised fine-tuning on five datasets: Alpaca (52,000 instructions from text-davinci-003); GPT4All Prompt Generations (400k prompts and responses; the card says 'generated by GPT-4', the GPT4All technical report says GPT-3.5-Turbo); Anthropic HH (preference data); Databricks Dolly (15k human-written); ShareGPT Vicuna (English subset).recorded · source
- techniques
- primary_sources
- record_history
- date:2026-09-24 · change:ingested as candidate from HF (stabilityai/stablelm-tuned-alpha-7b@25071b093c15c0d1cb2b2876c6deb621b764fcf5) · by:ingest_hf.py ·date:2026-09-24 · change:preparer: developer, release date (GitHub changelog), license, training data, 5 edges accepted + 2 rejected; sources: card, StableLM README · by:claude (preparer, Sonnet 5) ·date:2026-09-24 · change:reviewed and promoted from staging (5 edge(s) accepted) · by:Wilson Pruitt ·
Parents
Weights descend
- fine_tuned_from → stablelm-base-alpha-7b declared source Card: 'a suite of 3B and 7B parameter decoder-only language models built on top of the StableLM-Base-Alpha models and further fine-tuned on various chat and instruction-following datasets.' Same layer/hidden/head/context values as the base record.
Training data
- trained_on → Alpaca instruction data (52K) declared source Card training-dataset section lists Alpaca first; uploader is the developer.
- trained_on → GPT4All prompt generations declared source Card lists 'GPT4All Prompt Generations' and describes it as 'generated by GPT-4'. The GPT4All technical report (see the `gpt4all` record) says GPT-3.5-Turbo; the edge follows the dataset name, and the card's model attribution is flagged, not adopted.
- trained_on → databricks-dolly-15k declared source Card lists Databricks Dolly, 15k, and its metadata names HuggingFaceH4/databricks_dolly_15k (a re-host of databricks-dolly-15k).
- trained_on → ShareGPT conversations (LMSYS collection) declared source Card: 'ShareGPT Vicuna (English subset)', metadata jeffwan/sharegpt_vicuna, which is now unreachable (HTTP 401), so identity with the `sharegpt-vicuna` record (LMSYS collection) is by name and could not be checked. Reviewer: confirm the id mapping.
Children
No edges recorded.
Read in
No station on the reading path has touched this record yet.