Serverless Training

Serverless training is currently in private preview and access is gated per account. Request access and select “Serverless Training API.”

Serverless training connects to a shared, already-running pooled trainer through the gateway and returns training and sampling clients. There is no trainer job or inference deployment to provision or delete. Close each sampler and the service client when finished. Not sure whether the shared pool fits your workload? Compare serverless and dedicated training.

What you need

Install the SDK (same one as the dedicated path, no separate console flow): pip install "fireworks-ai[training]"
For the runnable example, clone the cookbook: git clone https://github.com/fw-ai/cookbook && pip install -e ./cookbook/training
Point at the serverless endpoint: base_url="https://api.fireworks.ai/training/v1/serverless"
Pick a base model enabled for serverless training on your account. See Models below. Availability changes during private preview; verify it before launch.
Set your API key and run the Quickstart below.

Using a code agent? A complete, forkable serverless RL loop lives in the cookbook at training/examples/serverless_rl/. It runs out of the box against the serverless gateway.

What is serverless training?

You write the training loop, for supervised fine-tuning or reinforcement learning, and Fireworks runs the forward pass, backward pass, and optimizer on remote GPUs, then serves your latest weights for sampling in the same session. The quickstart shows the exact client setup and operation order. The complete implementation lives in the cookbook’s serverless_rl example.

Serverless lifecycle

1 · Local

Builds batches, rewards, and optimizer requests.

2 · Shared

Runs remote LoRA forward, backward, and optimizer work.

3 · In session

Captures the current adapter weights for sampling.

4 · Shared

Returns rollouts for local scoring and the next step.

You do not create or delete a trainer job or inference deployment. Close each sampler and the service client when the run finishes.

What you can run

Supervised Fine-Tuning (SFT)

Run the loop with a cross-entropy loss over your labeled data.

Reinforcement Learning (RL)

Sample completions from the current adapter, score them with your own reward function, and train with an importance-sampling loss (GRPO-style). This is the primary serverless use case.

Both run as LoRA on the shared pool. For full-parameter training, DPO, or the broader Training API method set, use Dedicated Training. For standard platform-managed jobs, use Managed Fine-Tuning.

When to use dedicated

Use Dedicated Training when you need full-parameter training, broader model or method support, explicit resource lifecycle control, or sustained utilization. See the canonical serverless versus dedicated comparison.

Core concepts

Session. create_lora_training_client(base_model, rank) attaches you to a pooled trainer for that base model. That attachment is your training session (service.training_session_id) — your LoRA state lives there. Run. The training client is your run (training_client.run_id). One run is one training trajectory: the forward_backward and optim_step calls plus the checkpoints you save. LoRA adapter. Serverless is LoRA only. Pass a positive rank (e.g. rank=8); base weights stay frozen and shared across the pool, and you train an adapter on top. Checkpoint / snapshot. save_weights_for_sampler(name) writes your current adapter weights and returns a snapshot path. That path is a public sampler identity, not a raw storage URI — hand it to the sampler to serve exactly those weights. See Saving and loading checkpoints for the full save / resume / promote surface. Sampling. create_sampling_client(model_path=snapshot, tokenizer=...) returns a sampler bound to that snapshot through the completions API (/inference/v1/completions). For multi-turn rollouts, session-affinity headers pin an episode to one replica so the KV cache is reused across turns (see RL rollout integration). The sampler runs in the same session, so there’s no deployment to create or hot-load.

Quickstart

Step 1: Create a key, install, and authenticate

Create an API key in the Fireworks dashboard (click Create API key and store it somewhere safe), or run firectl api-key create. Then install the SDK and export the key:

git clone https://github.com/fw-ai/cookbook
pip install -e ./cookbook/training
export FIREWORKS_API_KEY="fw_..."   # the key you just created

Step 2: Run the complete serverless RL example

The cookbook includes its dataset, reward, loop, metrics, and cleanup behavior:

cd cookbook/training
python -m examples.serverless_rl.countdown_rl

The default is a real paid run. For a cheaper smoke test, fork examples/serverless_rl/countdown_rl.py and reduce steps, group_size, prompt_groups_per_step, and max_sample_tokens before execution. The remaining snippets explain the core calls used by that runnable example.

Step 3: Connect to the serverless session

import os
from fireworks.training.sdk import FiretitanServiceClient

service = FiretitanServiceClient(
    api_key=os.environ["FIREWORKS_API_KEY"],
    base_url="https://api.fireworks.ai/training/v1/serverless",
)

base_model = "accounts/fireworks/models/qwen3p6-27b"
max_seq_len = 131072  # enforce this when rendering prompts and training datums
training_client = service.create_lora_training_client(base_model=base_model, rank=8)

print(f"session={service.training_session_id} run={training_client.run_id}")

This doubles as your setup check: if it prints a session id beginning with ts-, serverless routing is working. If the base URL does not end in /training/v1/serverless, sampling errors out.

Step 4: Train, checkpoint, and sample

The complete example defines datums, tinker, tokenizer, prompt, and params, and rejects prompt plus completion lengths above max_seq_len. The excerpt below shows the operation order:

# One optimizer step
training_client.forward_backward(datums, "importance_sampling").result()
training_client.optim_step(
    tinker.AdamParams(learning_rate=2.5e-5, beta1=0.9, beta2=0.95, eps=1e-8, weight_decay=0.0)
).result()

# Save weights and open a sampler bound to that exact snapshot
snapshot = training_client.save_weights_for_sampler("step-0001").result().path
sampler = service.create_sampling_client(model_path=snapshot, tokenizer=tokenizer)
completions = sampler.sample(
    prompt=prompt,
    num_samples=8,
    sampling_params=params,
).result()
sampler.close()

Remote training and sampling operations return future-like results. Call .result() on operations that return one, or failures can be missed.
The tokenizer you pass must match base_model. It renders prompts and decodes sampled tokens client-side, so a mismatch silently corrupts your rewards.
Give each checkpoint a distinct name. Saving twice with the same name currently resolves to the same snapshot path.

Reinforcement learning example

The end-to-end serverless RL pattern is the standard GRPO / importance-sampling loop: each step saves the current adapter, rolls out a batch of prompts through a sampler bound to that snapshot, scores completions with your reward function, turns group-relative advantages into training datums, and takes one optimizer step. Track reward over time; improvement depends on the task, data, reward function, and configuration. Use the complete cookbook serverless_rl example rather than rebuilding the loop from this page. For a supervised loop, use cross-entropy loss. For the broader RL loss menu and dedicated provisioning, see the cookbook RL recipes.

Saving and loading checkpoints

Serverless training writes two different kinds of checkpoint, and they are not interchangeable:

	Training checkpoint	Sampler checkpoint
Saved by	`save_state(name)`	`save_weights_for_sampler(name)`
Contains	Adapter weights and optimizer state	Current adapter weights only
Resumable — the trainer can load it back and keep training	✅ Yes	❌ No
Promotable to a deployable model	❌ No	✅ Yes

Only sampler checkpoints can be promoted to a final model, and only training checkpoints can be loaded back by the trainer to resume. A training checkpoint cannot be served or promoted; a sampler checkpoint cannot restore training state (weights + optimizer).

Checkpoint storage is included during private preview.

Save and resume training checkpoints

Save training checkpoints periodically so an interrupted run can continue:

# Save a training checkpoint (adapter weights + optimizer) for resume
training_client.save_state("step-0100").result()

# Resume inside the same run: restores weights and optimizer state
training_client.load_state_with_optimizer("step-0100").result()

# Weights-only load (optimizer resets to zero — a warm start)
training_client.load_state("step-0100").result()

Save and load calls return futures — call .result() to block and surface failures. save_state also accepts a timeout to bound the wait. Give each checkpoint a distinct name: overwriting an existing name (overwrite=True) is not supported. To see what checkpoints a run has saved, use the session-scoped control-plane list in Promote a sampler checkpoint to a model below — the trainer-local training_client.list_checkpoints() is not routed on the serverless surface.

Resume a new run from a training checkpoint

To fork a new run from a previous run’s training checkpoint, use create_training_client_from_state on the service client. The SDK reads the base model and LoRA configuration from the checkpoint itself, creates a fresh run, and loads the saved state:

# Continue a previous run's training checkpoint in a new run (weights + optimizer)
resumed_client = service.create_training_client_from_state_with_optimizer(
    "<account>/<run-id>/step-0100",
)

# Or warm start from those weights with a fresh optimizer
resumed_client = service.create_training_client_from_state(
    "<account>/<run-id>/step-0100",
)

Checkpoint references are fully qualified as <account>/<run-id>/<checkpoint-name>, where run-id is the previous run’s training_client.run_id (run-<hex>). weights_access_token is not supported — load checkpoints accessible to your API key.

Promote a sampler checkpoint to a model

Promotion turns a sampler checkpoint into a deployable Fireworks model (a standard LoRA addon model). Training checkpoints are not promotable — save a sampler checkpoint first.

Save a sampler checkpoint during training

snapshot = training_client.save_weights_for_sampler("final").result().path

Save one at the end of training, and at any intermediate step you may want to deploy later.

List the session's checkpoints

Listing and promotion are session-scoped control-plane operations on FireworksClient (against the regular API gateway, not the serverless base URL). The session resource name is available as service.training_session_name:

import os
from datetime import datetime
from fireworks.training.sdk import FireworksClient

fw_client = FireworksClient(api_key=os.environ["FIREWORKS_API_KEY"])

rows = fw_client.list_training_session_checkpoints(service.training_session_name)
target = max(
    (row for row in rows if row.get("promotable")),
    key=lambda row: datetime.fromisoformat(row["createTime"].replace("Z", "+00:00")),
)

Each row carries name (the full 4-segment resource name), checkpointName, checkpointType, promotable, and createTime. Two things to know about the values:

checkpointName is the server-side checkpoint id, not the bare name you passed: it is prefixed with the source run id and, for sampler checkpoints, suffixed with an 8-hex-char session id — a save named final surfaces as run-<hex>-final-<8hex>. Select rows on promotable + createTime as above, or with a prefix/substring test — never by equality with your logical name.
checkpointType is a server enum string: CHECKPOINT_TYPE_TRAINING_LORA for training checkpoints, CHECKPOINT_TYPE_INFERENCE_LORA for sampler checkpoints. Treat it as opaque and filter on promotable, which is authoritative.

Promote the checkpoint

model = fw_client.promote_session_checkpoint(
    name=target["name"],  # accounts/<a>/trainingSessions/<s>/checkpoints/<c>
    output_model_id="my-serverless-lora",
    base_model="accounts/fireworks/models/qwen3p6-27b",
)

output_model_id must be 1-63 characters of lowercase a-z, 0-9, and hyphens. The promoted model appears in your account’s model list like any other fine-tuned model.

Session-scoped list and promote require the training session and its bound trainer to still exist. Once the session is deleted or its trainer is drained, both calls return NOT_FOUND and the checkpoints are no longer reachable through this API — promote any checkpoint you want to keep before the session is torn down. Cross-run training-checkpoint resume (above) is resolved per run and is not subject to this limit.

Deploy the promoted model to production

A promoted model deploys like any LoRA model fine-tuned on Fireworks: with live merge, Fireworks merges the adapter into the base weights at deployment time, so the deployment performs identically to the base model.

Fine-tuned LoRA models can only be deployed to on-demand (dedicated) deployments. Serverless per-token serving of your own fine-tuned LoRA is not available.

Deploy the promoted model directly:

firectl deployment create "accounts/<account-id>/models/my-serverless-lora"

Then send requests with the model name:

from fireworks import Fireworks

client = Fireworks()
response = client.chat.completions.create(
    model="accounts/<account-id>/models/my-serverless-lora",
    messages=[{"role": "user", "content": "Hello!"}],
)

For deployment configuration, performance, and troubleshooting, see Deploying fine-tuned models. For the full SDK-level checkpoint reference (base/delta sampler types, weight sync, cross-job resolution), see Saving and Loading. For recipe-driven save / resume / promote, see Checkpoints and Resume (cookbook).

Pricing

Serverless training is billed per token, across three meters: prefill, sample, and train. Current rates per model are in Models below. Available models, meter definitions, and rates can change during private preview, so verify current availability and pricing before launch.

Supported models and limits

Models

Serverless models are not selected by shape. You attach to a shared, always-on trainer pool with a base model and your own max_seq_len, and pay only for the tokens you prefill, sample, and train. Serverless is LoRA only, for SFT and RL.

Checkpoint storage for serverless models is included during private preview.
The serverless model catalog evolves, and other frontier models are coming soon. For full-parameter training or a model not on that list, use Dedicated Training.
For per-model availability across all training surfaces, see the Models catalog.

What the meters mean

Serverless training bills three separate token meters, so a run’s cost depends on the mix of rollout and training tokens your loop generates:

Meter	What it counts
Prefill	Prompt tokens processed by the sampler before generation. Repeated prefixes hit the cache and bill at the cached rate, an 80% discount.
Sample	Tokens generated by the sampler during rollouts.
Train	Tokens passed through `forward_backward`, i.e. the datums you train on.

Capacity and rate limits (private preview)

Concurrent runs: The default quota is 8, although account overrides may differ. Slots are released when runs become terminal after session expiration.
Request and token limits: Contact Fireworks for the current limits on your account.
Shared-pool capacity: if the pool is full, create_lora_training_client returns an out-of-capacity error; retry, or switch to the dedicated path.

Behavior to know

Set max_seq_len explicitly. Serverless has no dedicated instance to infer sequence length from.
Cross-run checkpoint resume. A training checkpoint can be resumed inside the same run (load_state_with_optimizer), or forked into a new run with create_training_client_from_state / create_training_client_from_state_with_optimizer using a fully qualified <account>/<run-id>/<checkpoint-name> reference. See Saving and loading checkpoints.
Serving your trained adapter. Sample in-session during the run. To serve afterward, promote a sampler checkpoint to a model and deploy it on an on-demand dedicated deployment; serverless per-token serving of your own fine-tuned LoRA is not available.

Video walkthrough: Train a prompt router

This walkthrough fine-tunes Qwen 3.5 9B with LoRA SFT to classify prompts and route them to a small or large model. It covers the local Python loop, pooled serverless trainer, in-session evaluation, and the before-and-after comparison.

Open the serverless prompt-router notebook

Follow the complete Cookbook example for dataset preparation, LoRA SFT, sampling, and evaluation.

Next steps

Serverless RL cookbook example: runnable serverless loop
Choose infrastructure: compare serverless and dedicated
Dedicated Training: the provisioned path from setup through teardown
Training and Sampling: dedicated lifecycle internals
Loss Functions: built-in and custom losses
The Cookbook: ready-to-run recipes, including serverless_rl

Get Started

Serverless

Deployments

Models & Inference

Training

Fire Pass

Fireworks Nexus

Administration

Security & Compliance

Integrations

Reference

What is serverless training?

Serverless lifecycle

What you can run

Supervised Fine-Tuning (SFT)

Reinforcement Learning (RL)

When to use dedicated

Core concepts

Quickstart

Step 1: Create a key, install, and authenticate

Step 2: Run the complete serverless RL example

Step 3: Connect to the serverless session

Step 4: Train, checkpoint, and sample

Reinforcement learning example

Saving and loading checkpoints

Save and resume training checkpoints

Resume a new run from a training checkpoint

Promote a sampler checkpoint to a model

Deploy the promoted model to production

Pricing

Supported models and limits

Models

What the meters mean

Capacity and rate limits (private preview)

Behavior to know

Video walkthrough: Train a prompt router

Open the serverless prompt-router notebook

Next steps

​What is serverless training?

​Serverless lifecycle

​What you can run

Supervised Fine-Tuning (SFT)

Reinforcement Learning (RL)

​When to use dedicated

​Core concepts

​Quickstart

​Step 1: Create a key, install, and authenticate

​Step 2: Run the complete serverless RL example

​Step 3: Connect to the serverless session

​Step 4: Train, checkpoint, and sample

​Reinforcement learning example

​Saving and loading checkpoints

​Save and resume training checkpoints

​Resume a new run from a training checkpoint

​Promote a sampler checkpoint to a model

​Deploy the promoted model to production

​Pricing

​Supported models and limits

​Models

​What the meters mean

​Capacity and rate limits (private preview)

​Behavior to know

​Video walkthrough: Train a prompt router

Open the serverless prompt-router notebook

​Next steps

What is serverless training?

Serverless lifecycle

What you can run

When to use dedicated

Core concepts

Quickstart

Step 1: Create a key, install, and authenticate

Step 2: Run the complete serverless RL example

Step 3: Connect to the serverless session

Step 4: Train, checkpoint, and sample

Reinforcement learning example

Saving and loading checkpoints

Save and resume training checkpoints

Resume a new run from a training checkpoint

Promote a sampler checkpoint to a model

Deploy the promoted model to production

Pricing

Supported models and limits

Models

What the meters mean

Capacity and rate limits (private preview)

Behavior to know

Video walkthrough: Train a prompt router

Next steps