Mistral¶

Mistral AI's dense decoder-only LLM, ported to pure Keras 3. It keeps the Llama block shape (RMSNorm, SwiGLU, rotary embeddings) and adds grouped-query attention plus a sliding-window attention span, which bounds the KV cache on long sequences.

Links:

Paper: Mistral 7B (arXiv:2310.06825)
HF docs: transformers/model_doc/mistral

Variants¶

Load any of these with from_weights("<variant>").

Variant	Hub
`mistral-7b-v0.1`	`mistralai/Mistral-7B-v0.1`
`mistral-7b-instruct-v0.2`	`mistralai/Mistral-7B-Instruct-v0.2`
`mistral-7b-v0.3`	`mistralai/Mistral-7B-v0.3`
`mistral-7b-instruct-v0.3`	`mistralai/Mistral-7B-Instruct-v0.3`
`ministral-8b-instruct-2410`	`mistralai/Ministral-8B-Instruct-2410`
`mistral-nemo-base-2407`	`mistralai/Mistral-Nemo-Base-2407`
`mistral-nemo-instruct-2407`	`mistralai/Mistral-Nemo-Instruct-2407`
`mistral-small-24b-base-2501`	`mistralai/Mistral-Small-24B-Base-2501`
`mistral-small-24b-instruct-2501`	`mistralai/Mistral-Small-24B-Instruct-2501`

API¶

`MistralModel`¶

The decoder backbone, no LM head. Returns {"last_hidden_state": (batch, seq, embed_dim)}.

Arg	Default	Meaning
`vocab_size`	`32000`	token vocabulary size
`embed_dim`	`4096`	model width
`mlp_dim`	`14336`	MLP inner width
`num_layers`	`32`	decoder blocks
`num_heads`	`32`	query heads
`num_kv_heads`	`8`	key/value heads (GQA)
`head_dim`	`None`	per-head width
`norm_eps`	`1e-05`	RMSNorm epsilon
`rope_theta`	`10000.0`	rotary base frequency
`sliding_window`	`None`	local attention span
`tie_embeddings`	`False`	reuse the embedding matrix as the LM head

`MistralGenerate`¶

MistralModel plus a (tied) LM head. Returns {"logits": (batch, seq, vocab_size)} and adds .generate(). Same constructor arguments as MistralModel.

generate(input_ids, attention_mask=None, max_new_tokens=None,
         eos_token_id=None, sampler=None, seed=None, **prefill_inputs)

Arg	Default	Meaning
`input_ids`	required	`(batch, seq)` token ids
`attention_mask`	`None`	`(batch, seq)` 1 = keep, 0 = padding
`max_new_tokens`	`None`	tokens to generate
`eos_token_id`	`None`	stop token (defaults to the tokenizer's)
`sampler`	`None`	sampling strategy; greedy when unset
`seed`	`None`	seed for stochastic samplers

`MistralTokenizer`¶

Tokenizer on the tokenizers backend.

MistralTokenizer(hf_id=None, tokenizer_file=None)

Arg	Default	Meaning
`hf_id`	`None`	Hub repo to pull the tokenizer files from
`tokenizer_file`	`None`	explicit path to a `tokenizer.json`

Calling it returns {"input_ids", "attention_mask"}, padded across the batch. It accepts a plain string, a list of strings (a batch), or a chat-message list, which is routed through apply_chat_template automatically. Decode with .decode(ids) for one sequence or .batch_decode(ids) for a batch.

End-to-end example¶

Single input¶

import os
os.environ["KERAS_BACKEND"] = "torch"   # or "jax" / "tensorflow"

from kerasformers.models.mistral import MistralGenerate, MistralTokenizer

model = MistralGenerate.from_weights("mistral-7b-v0.1")
tokenizer = MistralTokenizer.from_weights("mistral-7b-v0.1")

inputs = tokenizer([{"role": "user", "content": "Explain rotary embeddings in one sentence."}])
outputs = model.generate(**inputs, max_new_tokens=64)

print(tokenizer.decode(outputs[0]))

Batch¶

Pass a list of strings. The tokenizer pads them and generate runs the batch together:

prompts = [
    "The capital of France is",
    "In one sentence, what is a transformer?",
    "Write a haiku about GPUs.",
]
inputs = tokenizer(prompts)              # {"input_ids": (3, seq), "attention_mask": (3, seq)}
outputs = model.generate(**inputs, max_new_tokens=64)

for text in tokenizer.batch_decode(outputs):
    print(text)

Backbone only¶

from kerasformers.models.mistral import MistralModel

backbone = MistralModel.from_weights("mistral-7b-v0.1")
hidden = backbone(inputs)["last_hidden_state"]   # (batch, seq, embed_dim)

Loading from the Hub¶

Any Hub repo with this architecture works via the hf: prefix, including community fine-tunes:

model = MistralGenerate.from_weights("hf:mistralai/Mistral-7B-v0.1")

Lower memory¶

Larger checkpoints load in bf16 or weight-only quantized. See quantization.md:

model = MistralGenerate.from_weights(
    "mistral-7b-v0.1", quantization="int8", low_memory=True, load_dtype="bfloat16"
)