N Neurarch Try it See it Changes Bot Pricing API Open app

Changes / Feeds / unsloth/Meta-Llama-3.1-8B-Instruct

unsloth/Meta-Llama-3.1-8B-Instruct

6 changes to a tracked key since 2024-07-23, newest first. Head as last read: a2856192dd7c25b842431f39c179a6c2c2f627d1.

GPU memory to fine-tune it, card by card: Meta-Llama-3.1-8B-Instruct on /m/.

Weights are byte-identical to meta-llama/Llama-3.1-8B-Instruct: all 291 tensors hash the same. Compared tensor by tensor from sha256 fingerprints: a2856192dd7c against meta-llama/Llama-3.1-8B-Instruct at 0e9e39f249a1. A hash is equal or not; it says nothing about what an edit does.

DateAdvisoryWhat changed, and what it does
2025-02-15x_NXMC-2025-0051 breakingtokenizer_class "PreTrainedTokenizerFast" to "PreTrainedTokenizer" (2025-02-15).
28e14b5ea911 -> a2856192dd7c Open in app
2025-02-02x_NXMC-2025-0050 breakingconfig.json eos_token_id [128001,128008,128009] to 128009, but generation_config.json still declares [128001,128008,128009], which generate() reads first, so default generation stops on the same tokens; only code that reads config.eos_token_id directly sees the new value; head dim 128 declared (it was unset, so the library default for the class applied); our reconstruction builds the same graph from both configs (same parameter count and structure); add_bos_token unset to true; encoding the same text yields different ids (2025-02-02).
62b4abd08ef2 -> 28e14b5ea911 Open in app
2024-08-07x_NXMC-2024-0043generation defaults changed (max_length unset to 131072); a call that does not pass these explicitly now generates differently (2024-08-07).
e1ea12ba7c95 -> e4df7a7b9c96 Open in app
2024-08-06x_NXMC-2024-0041 breakingconfig.json pad_token_id unset to 128004; the tokenizer's pad token, which training pads with, did not change (2024-08-06).
cf3c8428e6ee -> f9a49112db6e Open in app
2024-08-06x_NXMC-2024-0042 breakinggeneration_config.json pad_token_id unset to 128004; generate() fills finished sequences in a batch with this id, while training pads with the tokenizer's pad token, which did not change (2024-08-06).
f9a49112db6e -> e1ea12ba7c95 Open in app
2024-08-06x_NXMC-2024-0040 breakingthe chat template changed and renders differently (system+user, multi-turn, tool-call); prompts rendered with the old one no longer match what the model now expects, so cached prompts and eval sets need re-rendering (2024-08-06).
a95354754cf0 -> ec7283d26215 Open in app

Pin it, and let a bot move the pin

from_pretrained("unsloth/Meta-Llama-3.1-8B-Instruct", revision="a2856192dd7c25b842431f39c179a6c2c2f627d1")

With the Renovate preset, a pull request moves this pin when the head moves, and links back here.

config changes badge

[![config changes](https://img.shields.io/endpoint?url=https%3A%2F%2Fwww.neurarch.com%2Fchanges-feeds%2Fbadge%2Funsloth%2FMeta-Llama-3.1-8B-Instruct.json)](https://www.neurarch.com/changes-feeds/models/unsloth/Meta-Llama-3.1-8B-Instruct.html)